From 20fcb1ab1d32cb53694c95fd4f7fa4df8b952614 Mon Sep 17 00:00:00 2001 From: Sylvia Hang Nguyen <49593977+sylviahangnguyen@users.noreply.github.com> Date: Fri, 25 Jun 2021 18:57:06 +0100 Subject: [PATCH] Add files --- NER_eng_BiLSTM.ipynb | 2116 +++++++++++++++++++++++++++++++++++++++ NER_trivia_BiLSTM.ipynb | 2104 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 4220 insertions(+) create mode 100644 NER_eng_BiLSTM.ipynb create mode 100644 NER_trivia_BiLSTM.ipynb diff --git a/NER_eng_BiLSTM.ipynb b/NER_eng_BiLSTM.ipynb new file mode 100644 index 0000000..e6bf79a --- /dev/null +++ b/NER_eng_BiLSTM.ipynb @@ -0,0 +1,2116 @@ +{ + "nbformat": 4, + "nbformat_minor": 0, + "metadata": { + "colab": { + "name": "NER_eng_BiLSTM.ipynb", + "provenance": [], + "collapsed_sections": [ + "YsSB3CDt2NxF", + "WP5M3GnHxY-9", + "MRofcZfAUd9g", + "4p6nLSAePE9U", + "Hm9UnKBqUeZA", + "vfswr-bAXwl5" + ], + "toc_visible": true + }, + "kernelspec": { + "name": "python3", + "display_name": "Python 3" + }, + "accelerator": "GPU" + }, + "cells": [ + { + "cell_type": "markdown", + "metadata": { + "id": "YsSB3CDt2NxF" + }, + "source": [ + "## Outline\n", + "- [Introduction](#0)\n", + " - [Import libraries](#0.1)\n", + "- [Part 1: Explore the data](#1)\n", + " - [1.1 Import the datasets](#1.1)\n", + " - [1.2 Exploratory Analysis](#1.2)\n", + " - [Conclusion after analysis](#1.3)\n", + " \n", + "- [Part 2: Pre-process the data](#2)\n", + " - [2.1 Stemming](#2.1)\n", + " - [2.2 Lemmatization](#2.2)\n", + " - [2.3 Replacement](#2.3)\n", + " - [2.4 Pre-processing pipeline](#2.4)\n", + " - [2.5 Split to train/val datasets](#2.5)\n", + " - [2.6 Tokenization and Padding](#2.6)\n", + " - [2.7 Check the Imbalance in train dataset](#2.7)\n", + " - [2.8 One-hot encoding](#2.8)\n", + "\n", + "- [Part 3: Build the model](#3)\n", + " - [3.1 Glove Embedding](#3.1)\n", + " - [3.2 Define the model](#3.2)\n", + " - [3.3 Callbacks](#3.3)\n", + " \n", + "\n", + "- [Part 4: Train the model](#4)\n", + "- [Part 5: Test the model](#5)\n", + "- [Part 6: Test with your own sentence](#6)\n", + "\n", + "- [Part 7: Analyse the incorrect predictions](#7)\n", + " - [Potential improvement](#7.1)\n", + "\n", + "- [Export result to .tsv file](#8)\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "Uj9t5Iav2nzR" + }, + "source": [ + "\n", + "# Introduction\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "pZDTBRDfi6u7" + }, + "source": [ + "\n", + "## Import libraries" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "Ntqkdg4N3HW4", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "026c8545-ec38-4d57-9ca9-e7d508510898" + }, + "source": [ + "!python --version\n", + "import os\n", + "\n", + "%tensorflow_version 2.x\n", + "import tensorflow as tf\n", + "print(tf.__version__)\n", + "\n", + "# build the tokenized sentences and tags\n", + "from tensorflow.keras.preprocessing.text import Tokenizer\n", + "from tensorflow.keras.preprocessing.sequence import pad_sequences\n", + "\n", + "from tensorflow.keras.utils import to_categorical\n", + "from tensorflow.keras.initializers import Constant\n", + "from tensorflow.keras import Model\n", + "from tensorflow.keras.layers import Input, Embedding, Bidirectional, LSTM, \\\n", + "TimeDistributed, Dense, Dropout\n", + "from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping\n", + "\n", + "import numpy as np # linear algebra\n", + "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n", + "import matplotlib.pyplot as plt\n", + "%matplotlib inline\n", + "import seaborn as sns\n", + "from collections import Counter\n", + "import random as rnd\n", + "from nltk.corpus import stopwords\n", + "import nltk\n", + "nltk.download('stopwords')\n", + "from nltk.stem import WordNetLemmatizer \n", + "from nltk.stem import PorterStemmer\n", + "\n", + "!pip install sklearn_crfsuite\n", + "from sklearn_crfsuite.metrics import flat_classification_report\n", + "!pip install seqeval\n", + "from seqeval.metrics import precision_score, recall_score, f1_score, classification_report\n", + "import csv\n" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Python 3.7.10\n", + "2.5.0\n", + "[nltk_data] Downloading package stopwords to /root/nltk_data...\n", + "[nltk_data] Unzipping corpora/stopwords.zip.\n", + "Collecting sklearn_crfsuite\n", + " Downloading https://files.pythonhosted.org/packages/25/74/5b7befa513482e6dee1f3dd68171a6c9dfc14c0eaa00f885ffeba54fe9b0/sklearn_crfsuite-0.3.6-py2.py3-none-any.whl\n", + "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from sklearn_crfsuite) (1.15.0)\n", + "Requirement already satisfied: tabulate in /usr/local/lib/python3.7/dist-packages (from sklearn_crfsuite) (0.8.9)\n", + "Collecting python-crfsuite>=0.8.3\n", + "\u001b[?25l Downloading https://files.pythonhosted.org/packages/79/47/58f16c46506139f17de4630dbcfb877ce41a6355a1bbf3c443edb9708429/python_crfsuite-0.9.7-cp37-cp37m-manylinux1_x86_64.whl (743kB)\n", + "\u001b[K |████████████████████████████████| 747kB 12.5MB/s \n", + "\u001b[?25hRequirement already satisfied: tqdm>=2.0 in /usr/local/lib/python3.7/dist-packages (from sklearn_crfsuite) (4.41.1)\n", + "Installing collected packages: python-crfsuite, sklearn-crfsuite\n", + "Successfully installed python-crfsuite-0.9.7 sklearn-crfsuite-0.3.6\n", + "Collecting seqeval\n", + "\u001b[?25l Downloading https://files.pythonhosted.org/packages/9d/2d/233c79d5b4e5ab1dbf111242299153f3caddddbb691219f363ad55ce783d/seqeval-1.2.2.tar.gz (43kB)\n", + "\u001b[K |████████████████████████████████| 51kB 5.0MB/s \n", + "\u001b[?25hRequirement already satisfied: numpy>=1.14.0 in /usr/local/lib/python3.7/dist-packages (from seqeval) (1.19.5)\n", + "Requirement already satisfied: scikit-learn>=0.21.3 in /usr/local/lib/python3.7/dist-packages (from seqeval) (0.22.2.post1)\n", + "Requirement already satisfied: scipy>=0.17.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.21.3->seqeval) (1.4.1)\n", + "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.21.3->seqeval) (1.0.1)\n", + "Building wheels for collected packages: seqeval\n", + " Building wheel for seqeval (setup.py) ... \u001b[?25l\u001b[?25hdone\n", + " Created wheel for seqeval: filename=seqeval-1.2.2-cp37-none-any.whl size=16184 sha256=ae07dcf27893686fb86ad23dc99961e6c43db9b655a080d92441256538067ba7\n", + " Stored in directory: /root/.cache/pip/wheels/52/df/1b/45d75646c37428f7e626214704a0e35bd3cfc32eda37e59e5f\n", + "Successfully built seqeval\n", + "Installing collected packages: seqeval\n", + "Successfully installed seqeval-1.2.2\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "T58oiLSU25Za" + }, + "source": [ + "\n", + "# Part 1: Explore the data" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "WP5M3GnHxY-9" + }, + "source": [ + "\n", + "## 1.1 Import the datasets" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "iHhxVylamxCG", + "outputId": "50607de3-273f-432a-e13b-af3344732e19" + }, + "source": [ + "# Create new directories\n", + "!mkdir -p /data/eng\n", + "!mkdir -p /data/trivia10k13\n", + "!mkdir -p /model\n", + "\n", + "# Download data\n", + "!wget --no-check-certificate \\\n", + "https://groups.csail.mit.edu/sls/downloads/movie/engtrain.bio \\\n", + "-O /data/eng/train.tsv\n", + "\n", + "!wget --no-check-certificate \\\n", + "https://groups.csail.mit.edu/sls/downloads/movie/engtest.bio \\\n", + "-O /data/eng/test.tsv\n", + "\n", + "!wget --no-check-certificate \\\n", + "https://groups.csail.mit.edu/sls/downloads/movie/trivia10k13train.bio \\\n", + "-O /data/trivia10k13/train.tsv\n", + "\n", + "!wget --no-check-certificate \\\n", + "https://groups.csail.mit.edu/sls/downloads/movie/trivia10k13test.bio \\\n", + "-O /data/trivia10k13/test.tsv" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "--2021-06-03 08:19:10-- https://groups.csail.mit.edu/sls/downloads/movie/engtrain.bio\n", + "Resolving groups.csail.mit.edu (groups.csail.mit.edu)... 128.30.2.44\n", + "Connecting to groups.csail.mit.edu (groups.csail.mit.edu)|128.30.2.44|:443... connected.\n", + "HTTP request sent, awaiting response... 200 OK\n", + "Length: 1013492 (990K)\n", + "Saving to: ‘/data/eng/train.tsv’\n", + "\n", + "/data/eng/train.tsv 100%[===================>] 989.74K 1.85MB/s in 0.5s \n", + "\n", + "2021-06-03 08:19:11 (1.85 MB/s) - ‘/data/eng/train.tsv’ saved [1013492/1013492]\n", + "\n", + "--2021-06-03 08:19:11-- https://groups.csail.mit.edu/sls/downloads/movie/engtest.bio\n", + "Resolving groups.csail.mit.edu (groups.csail.mit.edu)... 128.30.2.44\n", + "Connecting to groups.csail.mit.edu (groups.csail.mit.edu)|128.30.2.44|:443... connected.\n", + "HTTP request sent, awaiting response... 200 OK\n", + "Length: 252636 (247K)\n", + "Saving to: ‘/data/eng/test.tsv’\n", + "\n", + "/data/eng/test.tsv 100%[===================>] 246.71K 714KB/s in 0.3s \n", + "\n", + "2021-06-03 08:19:12 (714 KB/s) - ‘/data/eng/test.tsv’ saved [252636/252636]\n", + "\n", + "--2021-06-03 08:19:12-- https://groups.csail.mit.edu/sls/downloads/movie/trivia10k13train.bio\n", + "Resolving groups.csail.mit.edu (groups.csail.mit.edu)... 128.30.2.44\n", + "Connecting to groups.csail.mit.edu (groups.csail.mit.edu)|128.30.2.44|:443... connected.\n", + "HTTP request sent, awaiting response... 200 OK\n", + "Length: 1785558 (1.7M)\n", + "Saving to: ‘/data/trivia10k13/train.tsv’\n", + "\n", + "/data/trivia10k13/t 100%[===================>] 1.70M 2.61MB/s in 0.7s \n", + "\n", + "2021-06-03 08:19:13 (2.61 MB/s) - ‘/data/trivia10k13/train.tsv’ saved [1785558/1785558]\n", + "\n", + "--2021-06-03 08:19:13-- https://groups.csail.mit.edu/sls/downloads/movie/trivia10k13test.bio\n", + "Resolving groups.csail.mit.edu (groups.csail.mit.edu)... 128.30.2.44\n", + "Connecting to groups.csail.mit.edu (groups.csail.mit.edu)|128.30.2.44|:443... connected.\n", + "HTTP request sent, awaiting response... 200 OK\n", + "Length: 438729 (428K)\n", + "Saving to: ‘/data/trivia10k13/test.tsv’\n", + "\n", + "/data/trivia10k13/t 100%[===================>] 428.45K 995KB/s in 0.4s \n", + "\n", + "2021-06-03 08:19:14 (995 KB/s) - ‘/data/trivia10k13/test.tsv’ saved [438729/438729]\n", + "\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "-D8_xEVDN-dj", + "outputId": "fe057421-cf99-47e0-cdae-d56f342434b9" + }, + "source": [ + "def get_sentence(file_path):\n", + " '''\n", + " Input:\n", + " file_path - path to the tsv file\n", + " Output:\n", + " sentences - list of sentences in string format\n", + " tags - list associated tags in string format\n", + " '''\n", + " sentences = []\n", + " tags = []\n", + " with open(file_path) as f:\n", + " contents = f.read()\n", + " sens_tags = contents.split(\"\\n\\n\")\n", + " for sen_tag in sens_tags:\n", + " words_tags = sen_tag.split(\"\\n\")\n", + " while (\"\" in words_tags):\n", + " words_tags.remove(\"\")\n", + " sen = ' '.join([word_tag.split(\"\\t\")[1] for word_tag in words_tags])\n", + " tag = ' '.join([word_tag.split(\"\\t\")[0] for word_tag in words_tags])\n", + " sentences.append(sen)\n", + " tags.append(tag)\n", + "\n", + " return sentences, tags\n", + "\n", + "\n", + "train_path = \"/data/eng/train.tsv\"\n", + "test_path = \"/data/eng/test.tsv\"\n", + "\n", + "sentences, tags = get_sentence(train_path)\n", + "test_sentences, test_tags = get_sentence(test_path)\n", + "\n", + "print(\"The train dataset has {} sentences.\".format(len(sentences)))\n", + "print(\"The test dataset has {} sentences.\".format(len(test_sentences)))" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "The train dataset has 9776 sentences.\n", + "The test dataset has 2444 sentences.\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "SJyYqFkQNCoW" + }, + "source": [ + "\n", + "## 1.2 Exploratory Analysis" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 198 + }, + "id": "Q6KzO6mq8326", + "outputId": "282abcf1-586b-4c59-87ce-dddcd4c0b66d" + }, + "source": [ + "# Take a look at the data\n", + "df = pd.read_csv(train_path, delimiter=\"\\t\", names=[\"Tag\", \"Word\"])\n", + "df.head()" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
TagWord
0Owhat
1Omovies
2Ostar
3B-ACTORbruce
4I-ACTORwillis
\n", + "
" + ], + "text/plain": [ + " Tag Word\n", + "0 O what\n", + "1 O movies\n", + "2 O star\n", + "3 B-ACTOR bruce\n", + "4 I-ACTOR willis" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 4 + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "tBVUmUQtFRJ1" + }, + "source": [ + "\n", + "### 1.2.1 Sentence Length \n" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 584 + }, + "id": "h2J38zVa-wkD", + "outputId": "a8c282e7-c675-42e9-a23a-461f0b386d0b" + }, + "source": [ + "plt.style.use(\"dark_background\")\n", + "\n", + "# How long are the sentences?\n", + "def plot_sentence_length_histogram(list_sentences):\n", + " '''\n", + " Input:\n", + " list_sentences - a list of sentences\n", + " Output:\n", + " [print] - Min, Max, Median and Average value of sentence length\n", + " [plot] - Histogram plot of sentence length\n", + " '''\n", + " lengths = [len(sen.split(' ')) for sen in list_sentences]\n", + " a4_dims = (11.7, 8.27)\n", + " fig, ax = plt.subplots(figsize=a4_dims)\n", + " sns.histplot(lengths)\n", + " plt.xlabel(\"Number of tokens in a sentence\")\n", + " plt.ylabel(\"Number of occurrences\")\n", + " print(\"Min: \",np.min(lengths))\n", + " print(\"Max: \",np.max(lengths))\n", + " \n", + " print(\"Median: \",np.median(lengths))\n", + " print(\"Average: \",round(np.mean(lengths),2))\n", + "\n", + "plot_sentence_length_histogram(sentences)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Min: 1\n", + "Max: 47\n", + "Median: 9.0\n", + "Average: 10.18\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "qY8VCLDQhCbJ" + }, + "source": [ + "### 1.2.2 Entity Length" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 584 + }, + "id": "XQrFzlrkhCm3", + "outputId": "310ea49d-586a-4ce8-b009-2aaf519a4c9d" + }, + "source": [ + "# How long are the entities?\n", + "def plot_entity_length_histogram(series):\n", + " '''\n", + " Input:\n", + " series - a pandas series of the tags\n", + " Output:\n", + " [print] - Min, Max, Median and Average value of entity length\n", + " [plot] - Histogram plot of entity length\n", + " '''\n", + " tags_list=[tag for tag in series]\n", + " tag_length = []\n", + " current_length = 0\n", + " for tag in tags_list:\n", + " if tag.startswith(\"B\"):\n", + " tag_length.append(current_length)\n", + " current_length = 1\n", + " elif tag.startswith(\"I\"):\n", + " current_length += 1\n", + " tag_length = tag_length[1:]\n", + " \n", + " a4_dims = (11.7, 8.27)\n", + " fig, ax = plt.subplots(figsize=a4_dims)\n", + " sns.histplot(tag_length)\n", + " plt.xlabel(\"Number of tokens in a tag\")\n", + " plt.ylabel(\"Number of occurrences\")\n", + " print(\"Min: \",np.min(tag_length))\n", + " print(\"Max: \",np.max(tag_length))\n", + " print(\"Median: \",np.median(tag_length))\n", + " print(\"Average: \",round(np.mean(tag_length),2))\n", + "\n", + "plot_entity_length_histogram(df[\"Tag\"])" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Min: 1\n", + "Max: 16\n", + "Median: 2.0\n", + "Average: 1.81\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAs8AAAHuCAYAAABgauG5AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3df1TUdb7H8RcDoeQPhuAmCaZOYVvd6mqSlpWlpqYllmbuVpKRVtf80VpKluGubquuZR67/SIui2WLoKZ4MwU1a0+tQkKIgQKKKCT+ChHNROB7//A2N/PHfFKZGeL5OOdzlvnOr/d895x67uxnZnwkWQIAAADgks3TAwAAAACNBfEMAAAAGCKeAQAAAEPEMwAAAGCIeAYAAAAM+Xl6AHfbt2+fSktLPT0GAAAAvFT79u11+eWXn/G6JhfPpaWlioyM9PQYAAAA8FJZWVlnvY5tGwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwFCDxXNCQoL27t2rvLw857GgoCClp6ersLBQ6enpstvtzuvmzZunoqIi5ebmqnPnzs7jI0aMUGFhoQoLCzVixAjn8S5dumjz5s0qKirSvHnzGuplAAAAAKewGmLdcccdVufOna28vDznsVmzZlmTJ0+2JFmTJ0+2Zs6caUmy7r33XmvlypWWJKtbt27Whg0bLElWUFCQtX37disoKMiy2+3W9u3bLbvdbkmyNm7caHXr1s2SZK1cudLq37+/0VxZWVkN8npZLBaLxWKxWL+Nda5ebLB3nv/5z3/q+++/P+VYVFSUkpKSJElJSUkaPHiw8/iCBQskSRs3bpTdbldoaKj69eunjIwMVVZW6tChQ8rIyFD//v0VGhqq1q1ba+PGjZKkBQsWOB8LAAAAaChu/Z7nNm3aqKKiQpJUUVGhNm3aSJLCwsK0e/du5+3KysoUFhZ2zuNlZWWnHT+bUaNGafTo0ZKkkJCQi/qaAAAA0HR49AODlmW55Xni4+MVGRmpyMhIHThwwC3PCQAAgN8et8bz3r17FRoaKkkKDQ3Vvn37JEnl5eVq166d83bh4eEqLy8/5/Hw8PDTjgMAAAANya3xnJaWpujoaElSdHS0li9f7jz+0zdpdOvWTVVVVaqoqNDq1avVt29f2e122e129e3bV6tXr1ZFRYUOHz6sbt26STr5jRw/PRYAAADQkBrkU4offfSR9d1331k1NTXW7t27rSeeeMK67LLLrDVr1liFhYVWRkaGFRQU5Lz9m2++aRUXF1ubN2+2br75ZufxkSNHWkVFRVZRUZH1+OOPO4/ffPPNVl5enlVcXGzNnz//onx6ksVisVgsFovFOlcv+vzfH01GVlaWIiMjPT0GAAAAvNS5epFfGAQAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAY8vP0AHA/m80mh8PhvLxjxw7V19d7cCIAAIDGgXhughwOh8bOn6ODR6oV3LKV5o99XsXFxZ4eCwAAwOsRz03UwSPV2n+4ytNjAAAANCrseQYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABgingEAAABDxDMAAABgiHgGAAAADBHPAAAAgCHiGQAAADBEPAMAAACGiGcAAADAEPEMAAAAGCKeAQAAAEPEMwAAAGCIeAYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABgingEAAABDxDMAAABgiHgGAAAADPl5eoDfOpvNJofD4by8Y8cO1dfXe3AiAAAAnC/iuYE5HA6NnT9HB49UK7hlK80f+7yKi4s9PRYAAADOA/HsBgePVGv/4SpPjwEAAIALxJ5nAAAAwBDxDAAAABgingEAAABDxDMAAABgiHgGAAAADBHPAAAAgCHiGQAAADBEPAMAAACGiGcAAADAEPEMAAAAGCKeAQAAAEPEMwAAAGCIeAYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABjySDxPmDBBW7ZsUV5enj766CM1a9ZMHTp00IYNG1RUVKTk5GRdcsklkiR/f38lJyerqKhIGzZsUPv27Z2PExsbq6KiIm3dulV9+/b1xEsBAABAE+L2eG7btq3GjRunrl276oYbbpCvr6+GDx+uWbNmae7cuYqIiFBlZaViYmIkSTExMaqsrFRERITmzp2rWbNmSZKuvfZaDR8+XNdff7369++vt956SzYbb6QDAACg4XikNv38/BQQECBfX19deuml2rNnj3r16qXFixdLkpKSkjR48GBJUlRUlJKSkiRJixcvVu/evZ3Hk5OTVVNTo507d6q4uFi33HKLJ14OAAAAmgi3x/N3332nOXPmaNeuXdqzZ4+qqqq0adMmHTp0SHV1dZKksrIyhYWFSZLCwsK0e/duSVJdXZ2qqqoUHBx8yvFf3ueXRo0apaysLGVlZSkkJKSBXyEAAAB+q9wez3a7XVFRUerYsaPatm2rFi1aqH///g36nPHx8YqMjFRkZKQOHDjQoM8FAACA3y63x3OfPn1UUlKiAwcOqLa2VkuXLlWPHj1kt9vl6+srSQoPD1d5ebkkqby8XO3atZMk+fr6KjAwUAcPHjzl+C/vAwAAADQEt8fzrl271L17dwUEBEiSevfurfz8fH322WcaOnSoJCk6OlrLly+XJKWlpSk6OlqSNHToUK1bt855fPjw4fL391eHDh0UERGhzMxMd78cAAAANCF+7n7CzMxMLV68WNnZ2aqtrVVOTo7ee+89ffLJJ0pOTtaMGTOUk5OjhIQESVJCQoI++OADFRUV6fvvv9fw4cMlSfn5+UpJSVF+fr5qa2s1ZswY1dfXu/vlAAAAoAlxezxL0rRp0zRt2rRTjpWUlKhbt26n3fb48eMaNmzYGR/n1Vdf1auvvtoQIwIAAACn4YuRAQAAAEPEMwAAAGCIeAYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABgingEAAABDxDMAAABgiHgGAAAADBHPAAAAgCHiGQAAADBEPAMAAACGiGcAAADAEPEMAAAAGCKeAQAAAEPEMwAAAGCIeAYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABgingEAAABDxDMAAABgiHgGAAAADBHPAAAAgCHiGQAAADBEPAMAAACGiGcAAADAEPEMAAAAGCKeAQAAAEPEMwAAAGCIeAYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABgingEAAABDxDMAAABgyGU8z5o1S61atZKfn5/WrFmjffv26ZFHHnHHbAAAAIBXcRnPffv2VXV1te677z7t3LlTV199tV544QV3zAYAAAB4FZfx7OfnJ0kaOHCgUlNTdfjw4QYfCgAAAPBGfq5u8D//8z8qKCjQsWPH9MwzzygkJEQ//vijO2YDAAAAvIrLd55ffPFF3Xbbberatatqa2v1ww8/KCoqyh2zAQAAAF7FZTwHBAToP//zP/X2229Lktq2bauuXbs2+GAAAACAt3EZz4mJiaqpqdFtt90mSSovL9eMGTMafDAAAADA27iM56uuukp/+9vfdOLECUnSsWPH5OPj0+CDAQAAAN7GZTzX1NSoefPmsixLkuRwOHT8+PEGHwwAAADwNi6/bSMuLk6rVq1Su3bt9OGHH6pHjx56/PHH3TAaAAAA4F1cxvOaNWuUnZ2t7t27y8fHR+PHj9fBgwfdMRsAAADgVVxu2xg8eLBqa2u1cuVKffLJJ6qtreWr6gAAANAkuYznuLi4U35VsKqqSnFxcQ06FAAAAOCNXMazzXb6TX76yW4AAACgKXEZz19//bVee+01ORwOORwOvfbaa9q0adMFPWlgYKBSU1NVUFCg/Px8de/eXUFBQUpPT1dhYaHS09Nlt9udt583b56KioqUm5urzp07O4+PGDFChYWFKiws1IgRIy5oJgAAAMAVl/E8duxY1dTUaNGiRVq0aJGOHz+uMWPGXNCTzps3T6tWrdK1116rm266SQUFBYqNjdXatWvVqVMnrV27VrGxsZKke++9VxEREYqIiNDo0aOdv3QYFBSkuLg4devWTbfccovi4uJOCW4AAADgYnMZzz/88INefPFFRUZGKjIyUlOmTNEPP/xw3k/YunVr3XnnnUpISJAknThxQlVVVYqKilJSUpIkKSkpSYMHD5YkRUVFacGCBZKkjRs3ym63KzQ0VP369VNGRoYqKyt16NAhZWRkqH///uc9FwAAAOCKy83LERERev7559WhQ4dT9jr37t37vJ6wY8eO2r9/vxITE3XTTTdp06ZNGj9+vNq0aaOKigpJUkVFhdq0aSNJCgsL0+7du533LysrU1hY2FmPn8moUaM0evRoSVJISMh5zQ0AAAC4jOfU1FS98847ev/991VXV3fhT+jnpy5dumjs2LHKzMzUG2+84dyi8XM//aLhxRAfH6/4+HhJUlZW1kV7XAAAADQtLuO5trZW77zzzkV7wrKyMpWVlSkzM1OStHjxYsXGxmrv3r0KDQ1VRUWFQkNDtW/fPklSeXm52rVr57x/eHi4ysvLVV5errvuuuuU4+vXr79ocwIAAAC/5HLP84oVK/TMM88oNDRUQUFBznW+9u7dq927d6tTp06STm7/yM/PV1pamqKjoyVJ0dHRWr58uSQpLS3N+U0a3bp1U1VVlSoqKrR69Wr17dtXdrtddrtdffv21erVq897LgAAAMAVl+88/xS0L7zwgvOYZVm66qqrzvtJx44dq4ULF8rf3187duzQyJEjZbPZlJKSopiYGJWWlmrYsGGSpJUrV2rAgAEqLi7WDz/8oJEjR0qSKisrNX36dOc2jD//+c+qrKw875kAAAAAV1zGs8PhuOhPmpubq8jIyNOO9+nT54y3f/bZZ894PDExUYmJiRd1NgAAAOBsXG7bCAgI0EsvvaR3331XknT11Vdr4MCBDT4YAAAA4G1cxnNiYqJqamp02223STr5Ab4ZM2Y0+GAAAACAt3EZz1dddZX+9re/6cSJE5KkY8eOycfHp8EHAwAAALyNy3iuqalR8+bNnd+77HA4dPz48QYfDAAAAPA2Lj8wGBcXp1WrVqldu3b68MMP1aNHDz3++ONuGA0AAADwLueMZx8fHwUFBenBBx9U9+7d5ePjo/Hjx+vgwYPumg8AAADwGueMZ8uyNGnSJKWmpmrlypXumgkAAADwSi73PK9Zs0YTJ05UeHj4RfmFQQAAAKCxcrnn+eGHH5YkjRkzxnnsQn9hEAAAAGiMXO55jo2NVUpKirvmAQAAALzWObdtWJalF154wV2zAAAAAF6NPc8AAACAIfY8AwAAAIZcxrPD4XDHHAAAAIDXcxnPjz322BmPf/DBBxd9GAAAAMCbuYznyMhI59/NmzdX7969lZ2dTTwDAACgyXEZz+PGjTvlcmBgoJKTkxtsIAAAAMBbufy2jV86evSoOnbs2BCzAAAAAF7N5TvPaWlpsixLkmSz2XTdddfxoykAAABoklzG85w5c5x/19bWqrS0VOXl5Q06FAAAAOCNXMbzrl27tGfPHh0/flzSyQ8Ntm/fXqWlpQ0+HAAAAOBNXO55Tk1NVX19vfNyXV2dUlNTG3QoAAAAwBu5jGc/Pz+dOHHCefnEiRPy9/dv0KEAAAAAb+Qynvfv36/777/feXnQoEE6cOBAgw4FAAAAeCOXe56ffvppLVy4UG+++aYkqaysTCNGjGjwwQAAAABv4zKed+zYoVtvvVUtWrSQdPJ7ngEAAICmyOW2jb/85S8KDAzU0aNHdfToUdntdk2fPt0dswEAAABexWU833vvvaqqqnJePnTokAYMGNCgQwEAAADeyGU8+/r6nvLtGs2bN1ezZs0adCgAAADAG7nc87xw4UKtXbtWiYmJkqSRI0cqKSmpwQcDAAAAvI3LeJ49e7Zyc3PVp08fSdL06dOVnp7e4IMBAAAA3sZlPEtSTk6OLrnkElmWpZycnIaeCQAAAPBKLvc8P/TQQ8rMzNTQoUM1bNgwbdy4UUOGDHHHbAAAAIBXcfnO80svvaTIyEjt379fkhQSEqI1a9ZoyZIlDT4cAAAA4E1cvvNss9mc4SxJBw8elM3m8m4AAADAb47Ld55XrVqlVatW6R//+Ick6eGHH9bKlSsbfDAAAADA27iM50mTJumBBx7Q7bffLkl67733tGzZsgYfDAAAAPA2Rt+28fHHH+vjjz9u6FkAAAAAr8bmZQAAAMAQ8QwAAAAYOms8r1mzRpI0c+ZMtw0DAAAAeLOz7nm+4oordOutt2rQoEFKTk6Wj4/PKdfzS4MAAABoas4az6+88oqmTp2q8PBwvf7666dcZ1mWevfu3eDDAQAAAN7krPG8ZMkSLVmyRC+//LJmzJjhzpkAAAAAr+Tyq+pmzJih+++/X3feeackaf369frkk08afDAAAADA27j8to1XX31V48ePV35+vvLz8zV+/Hj95S9/ccdsAAAAgFdx+c7zwIED9R//8R+yLEuSlJSUpJycHL300ksNPhwAAADgTYy+59lutzv/DgwMbLBhAAAAAG/m8p3nv/71r8rJydFnn30mHx8f3XnnnYqNjXXHbAAAAIBXcRnPycnJWr9+vSIjIyVJkydP1t69ext8MAAAAMDbuIxnSaqoqNCKFSsaehYAAADAqxnteQYAAABAPAMAAADGzhnPNptNBQUF7poFAAAA8GrnjOf6+npt27ZN7dq1c9c8AAAAgNdy+YHBoKAgffvtt8rMzNTRo0edx6Oiohp0MAAAAMDbuIznqVOnumMOAAAAwOu5jOcvvvhCV155pSIiIrR27VoFBATI19fXHbMBAAAAXsXlt208+eSTWrx4sd59911JUlhYmJYtW9bggwEAAADexmU8jxkzRj169NDhw4clScXFxbr88ssbfDAAAADA27iM5+PHj+vEiRPOy76+vrIsq0GHAgAAALyRy3j+/PPP9eKLLyogIEB9+vRRamoqP9UNAACAJsllPMfGxmr//v3Ky8vTU089pZUrV+rll192x2wAAACAV3H5bRuWZSkpKUkbN26UZVnatm2bO+YCAAAAvI7LeB4wYIDeeecdbd++XT4+PurYsaOeeuoprVq1yh3zAQAAAF7DZTy/9tpruvvuu7V9+3ZJksPh0CeffEI8AwAAoMlxuee5urraGc6StGPHDlVXVzfoUAAAAIA3Ous7zw888IAk6euvv9Ynn3yilJQUWZalhx56SFlZWW4bEAAAAPAWZ43n+++/3/n33r171bNnT0nS/v37FRAQ0PCTAQAAAF7mrPH8xBNPuHMOAAAAwOu5/MBghw4dNHbsWHXo0EF+fv9/86ioqAYdDAAAAPA2LuN52bJlSkhI0IoVK1RfX++OmQAAAACv5DKef/zxR82fP98dswAAAABezWU8z5s3T6+88orS09N1/Phx5/GcnJwGHQwAAADwNi7j+YYbbtBjjz2mXr16ObdtWJal3r17N/hwAAAAgDdxGc8PPfSQHA6HTpw44Y55AAAAAK/l8hcGt2zZIrvd7o5ZAAAAAK/m8p1nu92urVu3Kisr65Q9z3xVHQAAAJoal/EcFxfnjjkAAAAAr+cynr/44osGeWKbzaavv/5a5eXluv/++9WhQwclJycrODhYmzZt0mOPPaYTJ07I399fCxYs0M0336yDBw/q4YcfVmlpqSQpNjZWMTExqqur07hx45Sent4gswIAAACSwZ7nw4cPq6qqSlVVVTp27Jhqa2tVVVV1wU88fvx4FRQUOC/PmjVLc+fOVUREhCorKxUTEyNJiomJUWVlpSIiIjR37lzNmjVLknTttddq+PDhuv7669W/f3+99dZbstlcvhwAAADgvLmszdatWyswMFCBgYEKCAjQkCFD9NZbb13Qk4aFhWngwIF6//33ncd69eqlxYsXS5KSkpI0ePBgSSf3ViclJUmSFi9e7PyKvKioKCUnJ6umpkY7d+5UcXGxbrnllguaCwAAADiXX/1W7fLly9WvX78LetI33nhDkyZNcn5vdHBwsA4dOqS6ujpJUllZmcLCwiSdDO3du3dLkurq6lRVVaXg4OBTjv/yPr80atQoZWVlKSsrSyEhIRc0OwAAAJoul3ueH3jgAeffNptNXbt21Y8//njeTzhw4EDt27dP2dnZ6tmz53k/zq8RHx+v+Ph4SVJWVpZbnhMAAAC/PS7j+f7773f+XVtbq507d17Q19T16NFDgwYN0oABA9S8eXO1bt1a8+bNk91ul6+vr+rq6hQeHq7y8nJJUnl5udq1a6fy8nL5+voqMDBQBw8edB7/yc/vAwAAADQEl/H8xBNPXNQnnDJliqZMmSJJ6tmzp55//nk9+uijSklJ0dChQ7Vo0SJFR0dr+fLlkqS0tDRFR0drw4YNGjp0qNatW+c8/tFHH+n1119X27ZtFRERoczMzIs6KwAAAPBzZ43nqVOnnvVOlmVpxowZF3WQyZMnKzk5WTNmzFBOTo4SEhIkSQkJCfrggw9UVFSk77//XsOHD5ck5efnKyUlRfn5+aqtrdWYMWOce6gBAACAhnDWeD569Ohpx1q0aKGYmBgFBwdflHj+/PPP9fnnn0uSSkpK1K1bt9Nuc/z4cQ0bNuyM93/11Vf16quvXvAcAAAAgImzxvPrr7/u/Ltly5YaP368Ro4cqeTkZL322mtuGQ4AAADwJuf8qrqgoCBNnz5dmzdvlp+fn7p06aLY2Fjt37/fXfMBAAAAXuOs7zzPnj1bDz74oN577z3dcMMNZ9zGAQAAADQlZ33neeLEiWrbtq1efvllfffdd86f6P7p57oBAACApuas7zz7+vq6cw4AAADA6/3qn+cGAAAAmiriGQAAADBEPAMAAACGiGcAAADAEPEMAAAAGCKeAQAAAEPEMwAAAGCIeAYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABgingEAAABDxDMAAABgiHgGAAAADBHPAAAAgCHiGQAAADBEPAMAAACGiGcAAADAEPEMAAAAGCKeAQAAAEPEMwAAAGCIeAYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABgingEAAABDxDMAAABgiHgGAAAADBHPAAAAgCHiGQAAADBEPAMAAACGiGcAAADAEPEMAAAAGCKeAQAAAEPEMwAAAGCIeAYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABgingEAAABDxDMAAABgiHgGAAAADBHPAAAAgCHiGQAAADDk9ngODw/XunXr9O2332rLli0aN26cJCkoKEjp6ekqLCxUenq67Ha78z7z5s1TUVGRcnNz1blzZ+fxESNGqLCwUIWFhRoxYoS7XwoAAACaGLfHc21trSZOnKjrr79e3bt315gxY3TttdcqNjZWa9euVadOnbR27VrFxsZKku69915FREQoIiJCo0eP1ttvvy3pZGzHxcWpW7duuuWWWxQXF3dKcAMAAAAXm9vjuaKiQjk5OZKkI0eOqKCgQGFhYYqKilJSUpIkKSkpSYMHD5YkRUVFacGCBZKkjRs3ym63KzQ0VP369VNGRoYqKyt16NAhZWRkqH///u5+OQAAAGhC/Dz55O3bt1fnzp21ceNGtWnTRhUVFZJOBnabNm0kSWFhYdq9e7fzPmVlZQoLCzvr8TMZNWqURo8eLUkKCQlpqJcDAACA3ziPfWCwRYsWWrJkiSZMmKDq6urTrrcs66I9V3x8vCIjIxUZGakDBw5ctMcFAABA0+KRePbz89OSJUu0cOFCffzxx5KkvXv3KjQ0VJIUGhqqffv2SZLKy8vVrl07533Dw8NVXl5+1uNo/Gw2m66++mrnstn4UhgAAOAdPFIlCQkJKigo0Ny5c53H0tLSFB0dLUmKjo7W8uXLncd/+iaNbt26qaqqShUVFVq9erX69u0ru90uu92uvn37avXq1e5/MbjoHA6Hxs6foz/8NU5j58+Rw+Hw9EgAAACSPLDnuUePHhoxYoQ2b97s/ODglClTNHPmTKWkpCgmJkalpaUaNmyYJGnlypUaMGCAiouL9cMPP2jkyJGSpMrKSk2fPl1ZWVmSpD//+c+qrKx098tBAzl4pFr7D1d5egwAAIBTuD2ev/zyS/n4+Jzxuj59+pzx+LPPPnvG44mJiUpMTLxoswEAAADnwmZSAAAAwBDxDAAAABgingEAAABDxDMAAABgiHgGAAAADBHPAAAAgCHiGQAAADBEPAMAAACGiGcAAADAEPEMAAAAGCKeAQAAAEPEMwAAAGCIeAYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDKtxvKMAABEOSURBVAAAABgingEAAABDxDMAAABgiHgGAAAADBHPAAAAgCHiGQAAADBEPAMAAACGiGcAAADAEPEMAAAAGCKeAQAAAEPEMwAAAGCIeAYAAAAMEc8AAACAIeIZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABgingEAAABDxDMAAABgiHgGAAAADBHPAAAAgCHiGQAAADBEPAMAAACGiGcAAADAEPEMAAAAGPLz9ABAY2Oz2eRwOJyXd+zYofr6eg9OBAAA3IV4Bn4lh8OhsfPn6OCRagW3bKX5Y59XcXGxp8cCAABuQDwD5+HgkWrtP1zl6TEAAICbsecZAAAAMEQ8AwAAAIaIZwAAAMAQ8QwAAAAYIp4BAAAAQ8QzAAAAYIh4BgAAAAwRzwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwJCfpwcAcHHZbDY5HA7n5R07dqi+vt6DEwEA8NtBPAO/MQ6HQ2Pnz9HBI9UKbtlK88c+r+LiYk+PBQDAbwLxDPwGHTxSrf2Hqzw9BgAAvznseQYAAAAMEc8AAACAIbZtAHArPtAIAGjMiGcAbsUHGgEAjVmj37bRr18/bd26VUVFRZo8ebKnxwFg4KcPNB48Uu3pUU5js9l09dVXO5fN1uj/MQkAuIga9TvPNptN//Vf/6V77rlHZWVlysrKUlpamgoKCjw9GoBGytvfGffGbS/eONPPeft8ABqXRh3Pt9xyi4qLi1VSUiJJSk5OVlRUlNfFc3DLVs7/bN++vYenkdq3b+91M/0c810Y5rswv5zHG+cbOe0lHfrhqOyXtlDitL+otLSUmc7B2+cDcG7e9AaGJPlIsjw9xPkaMmSI+vfvr1GjRkmSHn30UXXr1k1jx4495XajRo3S6NGjJUnXXHONtm3b5vZZvVlISIgOHDjg6TEaJc7d+ePcnT/O3fnj3J0/zt2F4fydP0+cu/bt2+vyyy8/6/VWY11Dhgyx4uPjnZcfffRRa/78+R6fq7GtrKwsj8/QWBfnjnPHuWtci3PHueP8Nb7lbeeuUX8Spry8XO3atXNeDg8PV3l5uQcnAgAAwG9Zo47nrKwsRUREqEOHDrrkkks0fPhwpaWleXosAAAA/Eb5Sprm6SHOl2VZKioq0sKFCzV27Fh9+OGHWrp0qafHapSys7M9PUKjxbk7f5y788e5O3+cu/PHubswnL/z503nrlF/YBAAAABwp0a9bQMAAABwJ+IZAAAAMEQ8N1Hh4eFat26dvv32W23ZskXjxo3z9EiNjs1mU3Z2tlasWOHpURqdwMBApaamqqCgQPn5+erevbunR2o0JkyYoC1btigvL08fffSRmjVr5umRvFZCQoL27t2rvLw857GgoCClp6ersLBQ6enpstvtHpzQe53p3M2ePVsFBQXKzc3V0qVLFRgY6MEJvdeZzt1P/vjHP8qyLAUHB3tgssbhbOfv2WefVUFBgbZs2aJZs2Z5aLr/5/Hvy2O5f4WGhlqdO3e2JFktW7a0tm3bZl177bUen6sxreeee85auHChtWLFCo/P0tjW3//+dysmJsaSZF1yySVWYGCgx2dqDKtt27bWjh07rObNm1uSrEWLFlnR0dEen8tb1x133GF17tzZysvLcx6bNWuWNXnyZEuSNXnyZGvmzJken9Mb15nO3T333GP5+vpakqyZM2dy7n7FuZNkhYeHW6tWrbJ27txpBQcHe3xOb11nOn933XWXlZGRYfn7+1uSrH/7t3/z9JyeP1Esz69ly5ZZffr08fgcjWWFhYVZa9asse6++27i+Veu1q1bWzt27PD4HI1xtW3b1tq1a5cVFBRk+fr6WitWrLDuuecej8/lzat9+/an/Et469atVmhoqCWdfBNh69atHp/RW9cvz93P1+DBg60PP/zQ4zN66zrTuUtNTbVuvPFGq6SkhHj+ledv0aJFVu/evT0+10+LbRtQ+/bt1blzZ23cuNHTozQab7zxhiZNmqT6+npPj9LodOzYUfv371diYqKys7MVHx+vSy+91NNjNQrfffed5syZo127dmnPnj2qqqpSRkaGp8dqVNq0aaOKigpJUkVFhdq0aePhiRqnJ554Qp9++qmnx2g0Bg0apPLycm3evNnTozRKnTp10h133KENGzZo/fr16tq1q0fnIZ6buBYtWmjJkiWaMGGCqqurPT1OozBw4EDt27fPq75zsjHx8/NTly5d9Pbbb6tLly46evSoYmNjPT1Wo2C32xUVFaWOHTuqbdu2atGihR555BFPj9WoWZbl6REanSlTpqi2tlYLFy709CiNQkBAgKZMmaJXXnnF06M0Wn5+frrsssvUvXt3vfDCC0pJSfHoPMRzE+bn56clS5Zo4cKF+vjjjz09TqPRo0cPDRo0SCUlJUpOTlavXr30wQcfeHqsRqOsrExlZWXKzMyUJC1evFhdunTx8FSNQ58+fVRSUqIDBw6otrZWS5cu1W233ebpsRqVvXv3KjQ0VJIUGhqqffv2eXiixiU6Olr33Xcf/6PtV7jqqqvUsWNH5ebmqqSkROHh4crOzub/9fgVysrKnD+Cl5WVpfr6eoWEhHhsHuK5CUtISFBBQYHmzp3r6VEalSlTpqhdu3bq2LGjhg8frnXr1umxxx7z9FiNxt69e7V792516tRJktS7d2/l5+d7eKrGYdeuXerevbsCAgIknTx3BQUFHp6qcUlLS1N0dLSkkyG4fPlyD0/UePTr10+TJk3SoEGDdOzYMU+P02hs2bJFbdq0UceOHdWxY0eVlZWpS5cu2rt3r6dHazSWLVumu+++W5IUEREhf39/HThwwKMzeXzjNcv9q0ePHpZlWVZubq6Vk5Nj5eTkWPfee6/H52psq2fPnnxg8DzWTTfdZGVlZVm5ubnWxx9/bNntdo/P1FjWtGnTrIKCAisvL89asGCB89PnrNPXRx99ZH333XdWTU2NtXv3buuJJ56wLrvsMmvNmjVWYWGhlZGRYQUFBXl8Tm9cZzp3RUVF1q5du5z/znj77bc9Pqc3rjOdu59fzwcGf/35u+SSS6wPPvjAysvLszZt2mTdfffdHp2Rn+cGAAAADLFtAwAAADBEPAMAAACGiGcAAADAEPEMAAAAGCKeAQAAAEPEM4Amz7IszZkzx3l54sSJiouLuyiPnZiYqCFDhlyUxzqXoUOHKj8/X+vWrTvlePv27fX73//e5f2jo6M1f/78hhrvFF9++eVFf8yePXvq1ltvveiPCwC/RDwDaPJ+/PFHPfjggwoODvb0KKfw9fU1vm1MTIxGjRqlXr16nXK8Q4cO+sMf/nCxR7sgPXr0uOiPedddd/FriwDcgngG0OTV1tbqvffe03PPPXfadb9857i6ulrSyXc6169fr2XLlmn79u3661//qj/84Q/auHGjNm/eLIfD4bxPnz59lJWVpW3btmngwIGSJJvNptmzZyszM1O5ubkaPXq083G/+OILLV++/Iy/vDh8+HBt3rxZeXl5mjlzpiRp6tSpuv3225WQkKDZs2efcvuZM2fqjjvuUE5OjiZMmKBmzZrpv//7v7V582ZlZ2frrrvuOu05BgwYoK+++krBwcG655579NVXX2nTpk1KSUlRixYtJEklJSWaNm2aNm3apM2bN+uaa66RJN15553KyclRTk6OsrOz1bJly9Me/+fn8LPPPlNqaqoKCgr04YcfnvG/nyeffFKZmZn65ptvtHjxYucvLP6kffv2evrpp/Xcc88pJydHt99+u+677z5t2LBB2dnZysjI0OWXXy5JCgkJUXp6urZs2aL4+Hjt3LnT6/5HEwDv5/Ffk2GxWCxPrurqaqtVq1ZWSUmJ1bp1a2vixIlWXFycJclKTEy0hgwZcsptpZO/LllZWWmFhoZa/v7+VllZmTVt2jRLkjVu3Dhr7ty5zvt/+umnlo+Pj3X11Vdbu3fvtpo1a2aNGjXKeumllyxJlr+/v5WVlWV16NDB6tmzp3XkyBGrQ4cOp815xRVXWKWlpVZISIjl6+trrV271oqKirIkWZ999pl18803n3afX/4K5h//+EcrISHBkmRdc801VmlpqdWsWTMrOjramj9/vjV48GDriy++sOx2uxUcHGx9/vnn1qWXXmpJsiZNmmRNnTrVkk7+Stqzzz5rSbKeeeYZKz4+3pJkpaWlWbfddpslyWrRooXl6+t7xvP902yHDh2ywsLCLB8fH+urr76yevTocdrtL7vsMuff06dPdz7vz1dcXJw1ceJE5+Wf/2plTEyMNWfOHEuSNX/+fCs2NtaSZPXr18+yLItfe2OxWL9q+QkAoOrqai1YsEDjxo3TsWPHjO6TlZWliooKSdL27duVnp4uScrLy9Pdd9/tvF1KSoosy1JxcbF27Nih3/3ud+rbt69uvPFGDR06VJIUGBioiIgI1dTUKDMzUzt37jzt+SIjI7V+/XodOHBAkrRw4ULdeeedWr58ufHrvP322517m7dt26bS0lJ16tRJktSrVy917dpVffv2VXV1tQYOHKjrrrvOuUfZ399f//rXv5yPtXTpUknSpk2b9OCDD0o6uZ/59ddf18KFC7V06VKVl5efc57MzEznbb755ht16NDhtD3R//7v/64ZM2bIbrerZcuWWr16tcvXGR4erkWLFumKK66Qv7+/SkpKnK//gQcekCStXr1a33//vcvHAoCfY9sGAPyfN954QzExMc6tCdLJLR0228l/VPr4+Mjf39953fHjx51/19fXOy/X19fLz+//35uwLOuU57EsSz4+Pho7dqw6d+6szp07y+FwKCMjQ5J09OjRi//iDGzfvl2tWrVyxrSPj48yMjKcM15//fV68sknnbf/6fXW1dU5X++sWbP05JNPKiAgQF9++aVzO8fZ/Pwc/vxxfu7vf/+7nn32Wd14443605/+pObNm7t8LfPnz9ebb76pG2+8UU899ZTRfQDABPEMAP+nsrJSKSkpiomJcR7buXOnbr75ZknSoEGDTolnUw899JB8fHzkcDjkcDi0bds2rV69Ws8884wzFiMiInTppZee83EyMzPVs2dPBQcHy2az6fe//70+//zzc96nurparVq1cl7+5z//qUceecT5nFdeeaW2bdsmSSotLdWQIUO0YMECXXfdddqwYYN69Oihq666SpJ06aWXKiIi4pzP53A4tGXLFs2ePVtZWVn63e9+d+6TY6BVq1bas2eP/Pz8nLO7ep2BgYHOd7Sjo6Odx7/88ksNGzZMknTPPffosssuu+D5ADQtxDMA/Mxrr72mkJAQ5+X4+Hj17NlT33zzjW699VYdOXLkVz/mrl27lJmZqU8//VRPP/20jh8/rvfff1/5+fnKzs5WXl6e3n333TO+6/pzFRUVio2N1Weffabc3Fxt2rRJaWlp57zP5s2bVVdXp2+++UYTJkzQW2+9JZvNps2bN2vRokV6/PHHVVNT47z9tm3b9Mgjjyg1NVWtW7fW448/rn/84x/Kzc3Vv/71L5cxPGHCBOXl5Sk3N1cnTpzQp59+an6izmLq1KnauHGjvvzyS23duvWMt1mxYoUeeOAB5wcGp02bptTUVH399dfObS6S9Kc//Ul9+/ZVXl6eHnroIe3Zs8f5AUYAMOGjk5ufAQD4zfP391ddXZ3q6urUvXt3vf322+rcubOnxwLQiPCBQQBAk3HllVcqJSVFNptNNTU1GjVqlKdHAtDI8M4zAAAAYIg9zwAAAIAh4hkAAAAwRDwDAAAAhohnAAAAwBDxDAAAABj6Xwk6RYWMQuMmAAAAAElFTkSuQmCC\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "OcoUQIfwfOAe" + }, + "source": [ + "### 1.2.3 Token frequency" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 567 + }, + "id": "b2kvMJJ3KvXu", + "outputId": "757556df-681a-4624-c19b-94bf69835f53" + }, + "source": [ + "def plot_top_non_stopwords_barchart(series, top=20, word=True):\n", + " '''\n", + " Input:\n", + " series - a pd.Series of words or tags\n", + " top - number of most common words to plot\n", + " Output:\n", + " [print] - No of distinct words in train dataset\n", + " [plot] - Barchart of most common words' occurrence\n", + " '''\n", + " stop=set(stopwords.words('english'))\n", + " value = 'words' if word == True else 'tags'\n", + " corpus=[word for word in series]\n", + " counter=Counter(corpus)\n", + " print(\"There are {} distinct {} in dataset\".format(len(dict(counter)), value))\n", + " print(dict(counter))\n", + "\n", + " most=counter.most_common()\n", + " x, y=[], []\n", + " for word,count in most:\n", + " if (word not in stop):\n", + " x.append(count)\n", + " y.append(word)\n", + " if len(x) == top:\n", + " break\n", + " a4_dims = (11.7, 8.27)\n", + " fig, ax = plt.subplots(figsize=a4_dims)\n", + " sns.barplot(x=x,y=y)\n", + " plt.xlabel(\"Number of {} occurrences in a sentence\".format(value))\n", + " plt.ylabel(\"Most common {}s\".format(value))\n", + " return dict(counter)\n", + "\n", + "word_counter = plot_top_non_stopwords_barchart(df[\"Word\"], top=40)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "There are 6710 distinct words in dataset\n", + "{'what': 2938, 'movies': 1702, 'star': 303, 'bruce': 37, 'willis': 24, 'show': 588, 'me': 958, 'films': 510, 'with': 1302, 'drew': 12, 'barrymore': 8, 'from': 895, 'the': 5331, '1980s': 74, 'starred': 245, 'both': 8, 'al': 23, 'pacino': 17, 'and': 777, 'robert': 86, 'deniro': 11, 'find': 810, 'all': 278, 'of': 1414, 'that': 1640, 'harold': 3, 'ramis': 2, 'bill': 33, 'murray': 14, 'a': 4224, 'movie': 3533, 'quote': 53, 'about': 1182, 'baseball': 11, 'in': 2789, 'it': 197, 'have': 335, 'mississippi': 3, 'title': 133, 'science': 163, 'fiction': 162, 'directed': 1122, 'by': 940, 'steven': 50, 'spielberg': 34, 'do': 231, 'you': 413, 'any': 738, 'thrillers': 21, 'sofia': 10, 'coppola': 26, 'leonard': 8, 'cohen': 4, 'songs': 18, 'been': 178, 'used': 16, 'elvis': 10, 'set': 39, 'hawaii': 2, 'is': 2214, 'references': 1, 'zydrate': 1, 'are': 765, 'there': 1062, 'musical': 107, 'patrick': 23, 'dempsey': 1, 'list': 795, 'westerns': 21, 'starring': 759, 'john': 155, 'wayne': 22, 'military': 56, 'related': 4, 'demi': 5, 'moore': 19, 'did': 770, 'stephen': 12, 'made': 300, 'sex': 6, 'or': 42, 'horror': 242, 'must': 90, 'be': 76, 'to': 413, 'watch': 53, 'ang': 2, 'lee': 56, 'direct': 268, 'costume': 2, 'drama': 199, '1800s': 1, 'who': 599, 'first': 100, 'james': 129, 'bond': 37, 'r': 653, 'rated': 1546, 'best': 147, 'picture': 24, 'oscar': 28, 'winners': 3, 'where': 213, 'can': 187, 'i': 433, 'buy': 10, 'sucker': 1, 'punch': 2, 'soundtrack': 64, 'british': 8, 'came': 189, 'out': 384, '1990s': 85, 'royal': 1, 'family': 97, 'camerons': 3, 'titanic': 21, 'was': 1559, 'last': 562, 'judy': 13, 'garland': 11, 'spoof': 1, 'charlie': 20, 'sheen': 10, 'id': 69, 'like': 174, 'documentary': 150, 'doctors': 1, 'chicago': 3, 'hospital': 3, 'action': 270, 'woo': 2, 'romantic': 129, 'film': 1201, 'protagonist': 1, 'doesnt': 4, 'fall': 2, 'love': 68, 'does': 66, 'original': 14, 'little': 23, 'mermaid': 4, 'hans': 5, 'christian': 23, 'anderson': 21, 'come': 127, 'on': 264, 'dvd': 8, 'disney': 48, 'beyonce': 1, 'sang': 7, 'shawn': 5, 'levy': 7, '2010s': 2, 'new': 46, 'david': 85, 'dhawan': 1, 'hannibal': 2, 'lector': 2, 'childrens': 15, 'billy': 39, 'crystal': 8, 'reviewed': 11, 'rock': 15, 'ballads': 1, 'were': 235, 'free': 1, 'willy': 9, 'filmed': 26, 'jennifer': 23, 'aniston': 9, 'main': 36, 'character': 99, 'wars': 32, 'im': 139, 'looking': 254, 'for': 612, 'song': 168, 'day': 46, 'has': 561, 'shortest': 1, 'trailer': 90, 'will': 70, 'always': 8, 'want': 121, 'see': 188, 'clint': 49, 'eastwood': 47, 'comedy': 322, 'whitney': 6, 'houstons': 1, 'u': 2, 'had': 149, 'fast': 9, 'furious': 8, 'jack': 53, 'nicholson': 11, 'an': 657, 'insane': 5, 'asylum': 1, 'dramatic': 10, 'donald': 6, 'sutherland': 9, 'rocky': 14, 'aaron': 4, 'sorkin': 1, 'trent': 1, 'reznor': 1, 'michael': 105, 'jackson': 32, '1970s': 38, 'political': 16, 'alec': 7, 'baldwin': 4, 'keaton': 13, 'play': 81, 'batman': 12, 'ewan': 7, 'macgregor': 3, 'singing': 3, 'superman': 11, 'romance': 110, 'popular': 117, '2002': 6, 'kevin': 45, 'smith': 42, 'jason': 51, 'mews': 1, 'suicide': 5, 'painless': 2, 'their': 7, 'world': 40, 'warcraft': 2, 'wes': 15, 'schwartzman': 1, 'savannah': 1, 'smiles': 1, 'owen': 4, 'wilson': 6, 'considered': 41, 'which': 345, 'features': 45, 'blaze': 1, 'glory': 4, 'maker': 1, 'planning': 2, 'produce': 2, 'farely': 1, 'brothers': 19, 'move': 21, 'frank': 36, 'sinatra': 3, 'something': 6, 'one': 44, 'armed': 2, 'man': 41, 'rick': 9, 'springfield': 2, 'steve': 41, 'martin': 52, 'cameron': 43, 'just': 24, 'two': 128, 'stars': 465, 'less': 4, 'angelina': 18, 'jolie': 16, 'released': 347, 'most': 62, '4': 6, 'winning': 22, '1998': 6, 'roadtrip': 1, 'ben': 15, 'stiller': 8, 'how': 190, 'many': 209, 'based': 79, 'texas': 7, 'chainsaw': 3, 'massacre': 4, 'jimmy': 13, 'stewart': 25, 'ever': 204, 'actor': 166, 'called': 82, 'matchstick': 2, 'men': 14, 'good': 542, 'nc': 151, '17': 152, 'not': 23, 'princess': 14, 'them': 57, 'comedies': 64, 'brad': 44, 'pitt': 33, '2000s': 35, 'godfather': 15, 'featured': 47, 'dude': 2, 'g': 240, 'war': 177, 'ii': 12, 'please': 95, 'locate': 7, 'george': 63, 'lucas': 27, '1996': 4, 'actors': 25, 'cameo': 4, 'spongebob': 1, 'year': 380, 'social': 4, 'network': 2, 'plot': 161, '1': 4, '1990': 169, '1991': 3, 'garry': 3, 'marshall': 15, '2011': 35, 'natalie': 9, 'portman': 2, 'dramas': 37, 'latest': 20, 'williams': 33, 'time': 39, 'car': 19, 'chases': 3, 'worst': 9, 'scary': 49, '2010': 184, 'thriller': 154, 'art': 2, 'heist': 3, 'late': 5, 'earl': 2, 'jones': 41, 'at': 61, 'end': 3, 'bronx': 3, 'tale': 17, 'top': 45, 'scorsesy': 1, 'moive': 2, 'pg': 791, '13': 558, 'ghosts': 13, 'dogs': 9, 'musicals': 26, 'antonio': 8, 'banderas': 7, 'hitchcock': 26, 'color': 3, 'animals': 6, 'five': 157, 'youre': 5, 'stories': 5, 'amanda': 7, 'seyfried': 3, 'ill': 11, 'back': 31, 'foreign': 8, 'blue': 8, 'rob': 23, 'reiner': 4, '1992': 2, 'drill': 1, 'team': 8, 'give': 55, 'ice': 15, 'baby': 11, 'vanilla': 1, 'atmospheric': 1, 'bambi': 1, 'teen': 29, 'got': 80, 'operas': 1, 'selma': 3, 'blair': 4, 'cars': 10, 'dragons': 5, 'clooney': 20, 'leonardo': 27, 'dicaprio': 24, 'danny': 25, 'devito': 10, 'krasinski': 1, 'look': 5, 'up': 71, 'could': 58, 'mobster': 4, 'friday': 8, '13th': 4, 'loved': 6, 'tomb': 1, 'raider': 1, 'carrel': 1, 'remake': 6, 'dirty': 8, 'dancing': 7, 'similar': 8, 'gnomeo': 2, 'juliet': 3, 'talking': 16, 'scorsese': 13, 'documentaries': 13, 'pollution': 2, 'china': 1, 'famous': 24, 'name': 486, 'cary': 17, 'grant': 22, 'eva': 3, 'marie': 1, 'saint': 4, 'billing': 1, 'backdraft': 1, 'theme': 32, '8': 3, 'mile': 4, 'america': 4, 'fererra': 1, 'susan': 11, 'sarandon': 7, 'jenna': 1, 'elfman': 4, 'julie': 15, 'andrews': 7, 'grease': 1, 'russel': 4, 'crowe': 9, 'mathematician': 1, '2001': 16, 'played': 99, 'ron': 25, 'weasley': 1, 'harry': 37, 'potter': 31, 'groups': 2, 'robbers': 1, 'robbing': 1, 'same': 6, 'bank': 9, 'whats': 111, 'coolio': 1, 'funniest': 8, 'mobsters': 2, 'recent': 31, 'sequel': 8, 'speed': 4, 'anthony': 28, 'minghella': 1, 'matt': 18, 'damon': 14, 'milos': 2, 'forman': 2, 'life': 26, 'together': 9, 'when': 97, 'hobbit': 4, 'woody': 24, 'allen': 33, 'other': 17, 'people': 62, 'think': 27, 'iii': 6, 'terrible': 2, 'highest': 31, 'kids': 26, 'release': 22, 'involving': 34, 'cartoons': 3, 'get': 52, 'biography': 104, 'real': 9, 'us': 5, 'president': 8, 'portrayed': 4, 'funny': 33, 'wedding': 5, 'toasts': 1, 'britney': 1, 'spears': 1, 'she': 8, 'satisfaction': 1, 'indiana': 19, 'judd': 7, 'apatow': 6, 'since': 5, 'dog': 9, 'chase': 25, 'statum': 1, 'nora': 1, 'ephron': 1, 'write': 5, 'spielbergs': 5, 'award': 15, 'well': 121, 'known': 16, 'julia': 26, 'roberts': 28, '1993': 1, '2012': 19, 'awful': 1, 'teenagers': 3, 'lost': 13, 'woods': 8, 'contained': 2, 'line': 19, 'phone': 3, 'home': 13, 'police': 28, 'pg13': 4, 'darren': 1, 'aranofsky': 1, 'reviews': 49, 'before': 19, '2000': 159, 'giraffe': 1, 'health': 7, 'care': 3, 'lists': 1, 'braddock': 1, 'includes': 5, 'your': 18, 'touch': 2, 'black': 26, 'keys': 1, 'critics': 23, 'non': 6, 'rating': 479, 'next': 12, 'coming': 40, 'courtney': 2, 'foxx': 4, 'looks': 2, 'great': 17, 'this': 106, 'shining': 61, 'boats': 3, 'beautiful': 3, 'mind': 9, 'score': 10, 'shierly': 1, 'temples': 2, 'greatest': 4, 'diaz': 8, 'named': 34, 'hugo': 1, 'high': 23, 'school': 17, 'sean': 35, 'penn': 7, 'horses': 4, 'sci': 114, 'fi': 114, 'laurence': 12, 'olivier': 10, '1940s': 6, 'bourne': 13, 'featuring': 66, 'alien': 26, 'invasion': 4, 'some': 288, 'lewis': 18, '3d': 4, 'animated': 87, 'monsters': 7, 'vs': 2, 'aliens': 11, 'appropriate': 5, 'toddlers': 1, 'somewhere': 6, 'tom': 74, 'hanks': 31, 'keanu': 10, 'reeves': 15, 'need': 20, 'quotes': 9, 'trailers': 4, 'lesbian': 2, 'concert': 2, '1973': 1, 'receive': 19, 'genre': 77, 'walk': 4, 'remember': 7, 'part': 13, 'beatles': 8, 'simon': 4, 'pegg': 1, 'uses': 11, 'abba': 1, 'ridley': 20, 'scott': 39, 'music': 18, 'luke': 7, 'my': 60, 'heart': 14, 'go': 26, 'won': 23, '1999': 9, 'twilight': 9, 'expected': 2, 'appear': 20, 'theatres': 1, 'four': 168, 'higher': 10, '1950s': 9, 'boxing': 8, 'hopkins': 13, 'everythings': 1, 'roses': 1, 'howard': 28, 'meryl': 8, 'streep': 9, 'mad': 13, '1969': 2, 'scarlett': 5, 'johansson': 2, 'spaceships': 2, 'candy': 5, 'helen': 8, 'hunt': 8, '1960s': 30, 'names': 10, 'books': 10, 'fifth': 2, 'freddy': 6, 'krueger': 4, 'gary': 19, 'ross': 2, 'tobey': 1, 'maguire': 1, 'peter': 44, 'than': 12, 'lord': 13, 'rings': 13, 'historical': 39, 'jeff': 10, 'daniels': 3, 'midsummer': 1, 'nights': 6, 'dream': 4, 'romatic': 2, 'johnny': 45, 'depp': 40, 'christopher': 42, 'plummer': 2, 'win': 9, 'selling': 4, 'novel': 4, 'vampires': 14, 'bay': 19, 'psychiatric': 1, 'hospitals': 1, 'pretty': 23, 'pink': 7, '1930s': 6, 'scifi': 3, 'villain': 5, 'wins': 2, 'adventure': 150, 'rolling': 2, 'stones': 2, 'lone': 2, 'ranger': 3, 'moulin': 3, 'rouge': 3, 'awards': 10, 'silent': 9, 'harpo': 1, 'marx': 1, 'scream': 6, 'skating': 2, 'roger': 12, 'mpaa': 32, 'henry': 10, 'thomas': 11, 'quentin': 14, 'tarentino': 2, 'directors': 5, 'only': 11, 'psychological': 28, 'stanly': 1, 'kubrick': 21, '1971': 2, 'ed': 4, 'norton': 12, 'recently': 14, 'babies': 2, 'around': 47, 'presley': 3, 'ma': 4, 'collection': 1, 'van': 13, 'dame': 1, 'lead': 39, 'animal': 7, 'jmes': 1, 'alfred': 25, 'suspense': 19, 'series': 18, 'spiderman': 5, 'moviews': 1, '12': 4, 'gave': 9, 'cruise': 25, 'biggs': 5, 'favorite': 14, 'things': 13, 'hilary': 2, 'duff': 1, 'tell': 88, '2004': 8, 'ad': 1, 'placement': 1, 'arnold': 14, 'schwarzenegger': 11, 'say': 9, 'mw': 1, 'nice': 1, 'enrique': 1, 'toy': 11, 'story': 31, 'robbery': 10, 'now': 14, 'director': 154, 'kung': 8, 'fu': 8, 'ashley': 4, 'olsen': 3, 'hanksmeg': 1, 'ryan': 35, 'cillian': 2, 'murphy': 16, 'role': 58, 'jonah': 7, 'hill': 10, 'suspensful': 1, 'mansell': 1, 'done': 36, 'voice': 36, 'shrek': 20, 'cant': 11, 'wait': 2, 'king': 26, 'characters': 18, '2001s': 1, 'oceans': 5, '11': 2, 'spike': 17, 'vin': 5, 'diesel': 3, 'involve': 7, 'packed': 1, 'should': 5, 'humor': 7, 'too': 5, 'mathematicians': 1, 'three': 101, 'sylvester': 7, 'stallone': 7, 'dancers': 6, 'house': 28, 'sand': 8, 'fog': 4, 'actress': 19, 'kristen': 7, 'would': 97, 'marvel': 4, 'comics': 1, 'through': 1, 'eyes': 4, 'tomorrow': 1, 'never': 4, 'dies': 1, '2003': 1, 'wesley': 5, 'snipes': 3, 'swayze': 5, 'cinderella': 2, 'judi': 5, 'dench': 5, 'adventures': 5, 'tintin': 4, 'mermaids': 1, 'whos': 6, 'as': 127, 'chick': 8, 'flick': 49, 'lawrence': 13, 'turner': 7, 'pirates': 21, 'caribbean': 11, 'explosions': 3, 'robots': 5, 'elise': 1, 'clifton': 2, 'ward': 4, 'mockumentary': 127, 'guest': 2, 'mojave': 1, 'mon': 1, 'his': 25, 'fish': 9, 'says': 12, 'keep': 1, 'swimming': 1, '2': 32, 'years': 378, '1997': 7, 'jacksons': 2, 'mash': 9, 'clown': 4, 'posse': 2, 'mariah': 4, 'carey': 7, 'act': 29, 'jamie': 11, 'curtis': 7, 'decode': 1, 'rosie': 4, 'odonnell': 4, 'duchovnys': 1, 'edward': 18, 'jon': 11, 'bon': 2, 'jovi': 2, 'appeared': 13, 'parker': 9, 'superhero': 4, 'ghost': 7, 'rider': 5, 'la': 6, 'dolce': 2, 'vita': 2, 'wrote': 10, 'avatar': 10, 'interview': 4, 'vampire': 18, 'elizabeth': 15, 'taylor': 18, 'national': 1, 'velvet': 3, 'trying': 16, 'take': 17, 'over': 23, 'breeding': 1, 'humans': 3, 'created': 7, 'soundtrak': 1, 'nominated': 7, 'but': 19, 'havent': 1, 'going': 12, 'sin': 6, 'city': 23, 'joely': 1, 'richardson': 3, 'sharks': 6, 'lowest': 4, 'bad': 8, 'guy': 10, 'tron': 2, 'children': 42, 'ratatouille': 1, 'hunger': 17, 'games': 19, 'preview': 23, 'argonauts': 1, 'stuff': 2, 'dreams': 3, 'tim': 39, 'burtons': 1, 'newest': 15, 'ther': 1, 'gene': 24, 'wilder': 20, 'army': 16, 'darkness': 8, 'kind': 10, 'valentines': 4, 'sports': 22, '99': 1, 'basketball': 5, 'teams': 1, 'marisa': 4, 'tomei': 4, 'hattie': 1, 'mcdaniel': 1, 'avengers': 2, 'kirk': 6, 'douglas': 11, 'adam': 24, 'sandler': 21, 'horrors': 2, 'joe': 21, 'pesci': 12, 'craven': 3, 'breakfast': 1, 'club': 4, 'halle': 9, 'berry': 8, 'obi': 1, 'wan': 1, 'kenobi': 1, 'roman': 15, 'holiday': 2, 'every': 6, 'serenity': 2, 'ages': 1, 'jaws': 11, 'bruc': 1, 'ni': 1, 'battle': 11, 'algiers': 1, 'cabin': 2, 'disaster': 33, 'sandra': 20, 'bullock': 14, 'pullman': 3, 'brendan': 3, 'fehr': 1, 'its': 28, 'highly': 148, 'recommended': 55, 'summer': 7, 'allison': 1, 'hannigan': 5, 'band': 6, 'camp': 11, 'connelly': 2, 'means': 2, 'pirate': 8, 'muir': 1, 'happy': 5, 'feet': 4, 'trek': 15, '10': 12, 'friendly': 4, 'scariest': 4, 'halloween': 6, '1982': 2, 'robin': 27, 'scores': 2, 'grace': 2, 'pixar': 7, '2005': 6, 'lindsay': 3, 'lohan': 5, 'boyscout': 1, 'big': 12, 'kid': 16, 'casey': 1, 'affleck': 4, 'predator': 1, 'malcolm': 4, 'reynolds': 9, 'smurfs': 1, 'ten': 103, 'sellers': 3, 'burger': 1, 'robot': 16, 'adult': 3, '1975': 5, 'female': 13, 'leads': 3, 'wrestler': 2, 'carrie': 4, 'alone': 6, 'wall': 61, 'street': 8, 'financial': 2, 'collapse': 1, 'joss': 2, 'whedon': 2, 'make': 51, 'another': 8, '90s': 34, 'concentration': 3, 'survivor': 1, 'buddha': 1, 'contains': 3, 'bigger': 4, 'boat': 5, 'clooneys': 2, 'oldest': 1, 'ossie': 1, 'davis': 18, 'nicholas': 9, 'sparks': 2, 'jolies': 2, 'changeling': 1, 'mighty': 3, 'true': 16, 'jackie': 10, 'chan': 12, 'chris': 21, 'tucker': 7, 'religion': 4, 'stalker': 1, 'eric': 10, 'idle': 2, 'tyler': 11, 'perry': 8, 'goldfinger': 3, 'zombie': 14, 'hills': 4, 'beverly': 2, 'cop': 9, 'word': 11, 'laser': 2, 'doubt': 1, 'hulk': 5, '70s': 4, '5': 10, 'cousin': 3, 'vinney': 1, 'evil': 23, 'twins': 4, '24': 3, 'dinosaur': 1, 'daze': 1, 'safe': 3, 'noir': 103, 'crime': 115, 'anatomy': 1, 'murder': 26, 'nighy': 1, 'moonrise': 1, 'kingdom': 2, 'plays': 31, '500': 1, 'days': 13, 'spies': 2, 'noire': 1, '1986': 1, 'mystery': 121, 'third': 10, 'chim': 1, 'chimeney': 1, 'bette': 13, 'midler': 6, 'charlies': 1, 'angeles': 5, 'nicole': 14, 'kidman': 10, 'ashton': 4, 'kutcher': 4, 'napoleon': 2, 'dynamite': 1, 'anything': 5, 'leprechaun': 2, 'positive': 1, 'header': 1, 'directing': 9, 'alex': 7, 'haleys': 1, 'x': 11, 'academy': 12, 'polot': 1, 'tokyo': 1, 'france': 2, '60s': 4, 'classic': 18, 'stayin': 1, 'alive': 13, 'lethal': 2, 'weapon': 3, 'voiced': 4, 'donkey': 4, 'snowmen': 1, 'lastest': 2, 'gosling': 5, 'wash': 2, 'theaters': 10, 'currently': 5, 'travel': 3, 'isnt': 2, 'superheroes': 4, 'incredibles': 1, 'shakespearean': 1, 'branagh': 3, 'emma': 7, 'thompson': 13, 'andromeda': 1, 'strain': 1, 'josh': 8, 'duhamel': 2, 'fans': 2, 'christina': 3, 'applegate': 2, 'evening': 2, 'news': 2, 'dead': 39, '1980': 168, 'gone': 10, 'wind': 10, 'else': 4, 'blackhawk': 1, 'down': 8, 'happens': 4, 'he': 22, 'nightmare': 8, 'christmas': 17, 'hellow': 1, 'goodfellas': 10, 'celine': 1, 'dions': 1, 'really': 125, 'classical': 4, 'musicians': 6, 'boxers': 1, 'run': 38, 'super': 5, 'hero': 8, 'review': 26, 'alvin': 3, 'chipmunks': 3, 'die': 8, 'hitler': 4, 'post': 7, 'apocalyptic': 4, 'genesis': 1, 'rodriguez': 2, 'ferrell': 12, 'kirsten': 4, 'dunst': 3, 'hard': 10, 'also': 21, 'shoot': 2, 'em': 2, '21': 4, 'jump': 5, 'critically': 57, 'acclaimed': 56, 'divorce': 3, 'margaret': 1, 'thatcher': 3, 'babe': 3, 'ruth': 1, 'nutrition': 1, 'healthy': 1, 'eating': 4, 'red': 6, 'tails': 2, 'scenes': 5, 'viewers': 34, 'terry': 9, 'gilliam': 4, 'detachment': 1, 'flash': 1, '1967': 2, 'more': 26, '3': 22, 'haunted': 14, 'shark': 4, 'norman': 4, 'reedus': 1, 'mr': 17, 'mrs': 2, 'harrison': 22, 'ford': 52, 'shelby': 1, 'steel': 6, 'magnolias': 3, 'panther': 6, 'glenn': 5, 'close': 5, 'inspirational': 1, 'football': 5, 'principle': 2, 'shooting': 1, 'locations': 1, 'fistful': 1, 'dollars': 1, 'feature': 19, 'mafia': 4, 'goes': 4, 'audiences': 9, 'jane': 15, 'eyre': 2, 'carpenter': 3, 'urban': 4, 'cowboy': 14, 'pauly': 2, 'shore': 2, 'panda': 2, '1995': 10, 'puss': 2, 'n': 3, 'boots': 2, 'seven': 198, 'samurai': 5, 'amish': 1, 'beals': 1, 'dancer': 1, 'american': 23, 'reunion': 1, 'matrix': 15, 'hartnett': 1, '200s': 1, 'joseph': 12, 'gordon': 6, 'levitt': 3, 'hammer': 2, 'nemo': 5, 'finding': 5, 'spider': 9, 'max': 1, 'candace': 1, 'bergen': 3, 'minnie': 1, 'driver': 8, 'trilogy': 4, 'mirrens': 1, 'biographies': 2, 'crazy': 4, 'horse': 9, 'liz': 3, 'wayans': 3, 'castle': 3, 'sky': 4, 'vow': 4, 'macgoohan': 2, 'weird': 4, 'motion': 3, 'guns': 3, 'witches': 6, 'destroy': 1, 'cowboys': 2, 'polticial': 1, 'oliver': 17, 'stone': 19, 'voices': 4, 'gingerbread': 1, 'legal': 2, '80s': 15, 'barry': 8, 'lyndon': 2, 'metro': 1, 'goldwyn': 1, 'mayer': 2, 'become': 2, 'studio': 3, 'pie': 4, 'sparrow': 5, '1930': 3, 'average': 219, 'stare': 2, 'goats': 1, 'zombies': 9, 'lorax': 2, 'doom': 5, 'booth': 2, 'am': 75, 'maya': 2, 'rudolph': 2, 'beowulf': 1, 'drive': 11, 'killer': 21, 'birds': 5, 'casino': 3, 'royale': 1, 'aaliyah': 1, 'soundtracks': 8, 'switchfoot': 1, 'kate': 30, 'beckinsale': 4, 'rush': 5, 'hour': 4, 'eilera': 1, 'ray': 4, 'liota': 2, 'wiz': 1, 'gillian': 3, 'teenage': 5, 'wasteland': 1, 'ash': 1, 'angry': 2, 'seeking': 9, 'justice': 4, 'kirstie': 2, 'alley': 2, 'travolta': 20, 'spanish': 3, 'pattinson': 3, 'spirited': 1, 'away': 4, 'video': 1, 'denzel': 13, 'washington': 13, 'truman': 1, 'blackstreets': 1, 'times': 3, 'hamm': 1, 'groundhog': 1, 'valor': 1, 'playing': 8, 'neo': 3, 'midlers': 1, 'sta': 1, 'dash': 2, 'theremin': 1, 'docudrama': 7, 'iraq': 1, 'telepathic': 1, 'relationship': 8, 'human': 3, 'meshell': 1, 'dialog': 8, 'espiranto': 1, 'under': 2, 'sea': 5, 'kenan': 1, 'fantasy': 207, 'europe': 1, 'diesal': 1, 'road': 9, 'trips': 1, 'licensed': 1, 'bridesmaids': 2, 'ball': 3, 'apart': 1, 'surfer': 7, 'lestat': 1, 'planet': 10, 'space': 30, 'jerry': 12, 'seinfeld': 2, 'boys': 8, 'mcg': 1, 'lithgow': 4, 'detective': 5, 'casa': 1, 'de': 27, 'mi': 1, 'padre': 1, 'runaway': 7, 'jury': 2, 'hood': 6, 'nurses': 2, 'baker': 1, 'special': 2, 'effects': 2, 'lees': 1, 'lois': 1, 'maxwells': 1, 'vantage': 2, 'point': 6, 'japan': 3, 'walken': 8, 'white': 17, 'graduate': 1, 'scanners': 1, 'dee': 6, 'werewolves': 3, 'fight': 7, 'loving': 2, 'screenplay': 3, 'dark': 31, 'shadows': 2, 'walking': 1, 'tall': 1, 'heres': 1, 'penguin': 2, 'returns': 2, 'gia': 1, 'hunting': 2, 'elephants': 2, 'horrific': 1, 'thieves': 4, 'teachers': 1, 'facing': 1, 'adversity': 2, 'warwick': 2, 'heath': 8, 'ledger': 7, 'themed': 5, 'beast': 3, 'bob': 21, 'hoskins': 1, 'cooley': 1, 'riddler': 1, 'riding': 3, 'determinate': 1, 'book': 18, 'benjamin': 2, 'bratt': 1, 'campbell': 10, 'credited': 1, 'writer': 4, 'male': 4, 'abraham': 1, 'lincoln': 1, 'santa': 6, 'clause': 4, 'waht': 2, 'shaq': 2, 'shutter': 1, 'island': 7, 'nosferatu': 2, 'lion': 9, 'mona': 1, 'carvell': 1, 'maggie': 4, 'meet': 5, 'halfway': 1, 'politicians': 1, 'gus': 4, 'zant': 1, 'genres': 2, 'besides': 4, 'salt': 1, 'legolas': 1, 'franco': 4, 'theres': 3, 'no': 4, 'place': 20, 'lex': 2, 'boreanaz': 1, 'renee': 11, 'zellweger': 4, 'flicks': 8, 'forest': 3, 'fires': 1, 'said': 19, 'kings': 2, 'speech': 1, 'molly': 3, 'ringwold': 1, 'viewer': 7, '1972': 1, 'after': 12, 'skaters': 1, 'reese': 5, 'witherspoon': 4, 'large': 1, 'scale': 2, 'natural': 3, 'disasters': 1, 'casablanca': 3, 'snl': 1, 'madden': 1, 'sergio': 11, 'leone': 11, 'philip': 4, 'seymour': 8, 'hoffman': 13, 'transvestite': 1, 'mulholland': 1, 'justin': 8, 'timberlake': 6, 'jodie': 11, 'foster': 12, 'rio': 2, 'actresses': 4, 'miss': 5, 'moneypenny': 1, 'terminator': 9, 'bio': 4, 'pic': 3, 'e': 65, 'happened': 1, 'eleven': 3, 'cher': 3, 'fp': 1, 'greenaway': 1, 'aerosmith': 1, '1947': 1, 'dany': 1, 'boyles': 1, 'burton': 26, 'underworld': 6, 'streetcare': 1, 'desire': 4, 'stalag': 1, 'age': 10, 'corrupt': 1, 'policeman': 2, 'celebrity': 1, 'inc': 1, 'industry': 5, 'faye': 4, 'dunaway': 4, 'appears': 1, 'blues': 2, 'm': 12, 'quantum': 1, 'solace': 1, 'harrelson': 3, 'werewolf': 2, 'london': 12, 'during': 35, 'transformation': 1, 'scene': 8, 'channing': 3, 'tatum': 3, 'fred': 11, 'keaunu': 1, 'inner': 1, 'celebration': 1, 'richard': 27, 'gere': 6, 'amount': 2, 'sequels': 7, 'gena': 2, 'district': 1, '9': 3, 'thousand': 2, 'words': 1, '1920s': 2, 'torch': 1, 'singer': 9, 'jim': 32, 'carrey': 22, 'devil': 5, 'inside': 2, 'bottle': 1, 'shock': 1, 'farley': 2, 'critical': 2, 'future': 19, 'whoopi': 3, 'goldberg': 3, 'ip': 1, 'cube': 3, 'shadow': 2, 'fourth': 1, '1945': 1, 'transformers': 5, 'william': 22, 'shatner': 4, 'strada': 2, 'acted': 16, 'psychiatrist': 1, 'enters': 2, 'vincent': 8, 'price': 4, 'boris': 1, 'karloff': 1, 'winslet': 8, 'thats': 23, 'train': 13, 'woman': 22, 'mgm': 3, 'kaye': 1, 'betty': 3, 'yellow': 2, 'labrador': 1, 'control': 3, 'selection': 1, 'premiere': 2, 'jj': 4, 'abrams': 4, 'worth': 3, 'watching': 4, 'claire': 2, 'god': 6, 'bernard': 5, 'herman': 7, 'mewes': 1, 'identity': 15, 'chevy': 8, 'golden': 4, 'pond': 2, 'knight': 9, 'rises': 3, 'beneath': 2, 'wings': 1, 'alan': 16, 'rickman': 4, 'silence': 9, 'lambs': 7, '1988': 5, 'cherry': 1, 'pepsie': 1, 'queen': 8, 'victoria': 4, 'practical': 1, 'magic': 8, 'chasing': 1, 'amy': 8, 'starrted': 1, 'harakiri': 1, 'crown': 1, 'affair': 5, 'prison': 7, '50s': 2, 'quinton': 1, 'tarrantino': 1, 'mia': 1, 'farrow': 1, 'gun': 5, 'elmer': 4, 'bernstein': 3, 'hogan': 1, 'rope': 1, 'speilberg': 2, 'produced': 21, 'han': 1, 'solo': 2, 'co': 10, 'hula': 1, 'they': 13, 'sandlot': 2, 'tina': 5, 'fey': 4, 'screenplays': 1, 'greed': 4, 'wallstreet': 1, '1989': 3, 'cabiria': 1, 'shawshank': 11, 'redemption': 16, 'ana': 1, 'karenina': 1, 'performed': 2, 'emelio': 1, 'estevez': 3, 'hockey': 5, 'mel': 40, 'gibson': 20, 'least': 25, 'box': 8, 'office': 7, 'upon': 4, 'bacon': 5, 'off': 11, 'novels': 2, 'inspector': 4, 'clouseau': 1, 'mars': 4, 'needs': 1, 'moms': 1, 'grossing': 12, 'prequel': 3, 'roots': 1, 'pandas': 1, 'type': 17, 'rosemarys': 1, 'renoylds': 1, 'critic': 4, 'ebert': 1, 'shortbus': 2, 'adams': 8, 'enchanted': 1, '2007': 1, 'replacements': 2, 'sigourney': 2, 'weaver': 4, 'franchise': 2, 'skellington': 1, '2006': 3, 'directs': 1, 'notebook': 2, 'tourist': 1, 'linkin': 2, 'park': 6, 'leading': 6, 'actoractress': 1, 'wwii': 7, 'prince': 7, 'hell': 7, 'anymore': 2, 'mcgoohan': 1, 'diane': 3, 'pltf': 1, 'nemesis': 1, 'connery': 17, 'hope': 4, 'cuttlefish': 1, 'benny': 3, 'period': 8, 'pieces': 3, 'medieval': 2, 'superbad': 2, '1941': 2, 'rosebud': 3, 'blazing': 4, 'saddles': 5, 'supporting': 6, 'kermit': 2, 'muppets': 2, 'gael': 1, 'garcia': 4, 'bernal': 1, 'balloon': 1, 'giants': 1, 'francis': 18, 'hostel': 2, 'whose': 2, 'mouse': 4, 'staring': 32, 'j': 19, 'edgar': 4, 'hackman': 12, 'without': 6, 'gardening': 1, 'patterson': 1, 'handle': 2, 'truth': 3, 'humphrey': 13, 'bogart': 14, 'airplane': 11, 'austin': 9, 'powers': 10, 'feel': 4, 'buster': 4, 'higest': 1, 'know': 62, 'loves': 1, 'audry': 1, 'hepburn': 25, 'thr': 3, 'sixth': 2, 'zed': 1, 'lockhart': 1, 'pre': 1, '911': 3, 'trade': 1, 'center': 1, 'covered': 1, 'web': 1, 'waynes': 2, 'grit': 5, 'include': 4, 'incredible': 1, 'cry': 1, 'sister': 2, 'dillon': 4, 'drug': 4, 'use': 3, 'merican': 1, 'jessica': 10, 'alba': 3, 'marty': 4, 'mcfly': 2, '1977': 1, 'cast': 17, 'ivory': 3, 'everyone': 3, 'malcovich': 3, 'larenz': 2, 'tates': 1, 'chracter': 1, 'menace': 1, 'society': 1, 'pacimp': 1, 'houston': 6, 'luck': 7, 'lady': 5, 'comes': 8, 'bullit': 1, 'whay': 2, 'protocol': 1, 'droid': 1, 'tornadoes': 1, 'jeremy': 11, 'renner': 2, 'columbiana': 1, 'oscars': 8, 'extremely': 1, 'loud': 1, 'incredibly': 1, 'blow': 7, 'gaslight': 1, 'old': 12, '40': 1, 'ago': 5, 'league': 6, 'extraordinary': 6, 'gentlemen': 4, 'congeniality': 1, 'second': 6, 'carribean': 5, 'hudsons': 1, 'ranking': 1, 'd': 10, 'w': 1, 'griffith': 3, 'epic': 6, 'him': 4, 'offer': 3, 'refuse': 3, 'doc': 6, 'professor': 2, 'snapes': 1, 'executive': 1, 'producer': 2, 'numbers': 1, 'mohicans': 1, 'meg': 8, 'strange': 4, 'land': 3, 'eleased': 1, 'ingmar': 19, 'bergman': 23, 'abu': 1, 'calire': 1, 'danes': 2, 'blockbuster': 6, 'positively': 1, 'pitts': 2, 'barrymores': 2, 'monkees': 1, 'left': 4, 'wright': 3, 'western': 173, 'popularize': 1, 'knights': 2, 'important': 4, 'figure': 2, 'history': 61, 'india': 2, 'fighting': 6, '20s': 1, 'spain': 1, 'battlestar': 1, 'galactica': 1, 'secret': 12, 'ooze': 1, 'motorcycle': 1, 'showed': 3, 'pactick': 1, 'drag': 3, 'dave': 7, 'chappelle': 2, 'weed': 1, 'youll': 2, 'lorre': 3, 'fallon': 1, 'serious': 2, 'hits': 7, 'wonderful': 5, 'thing': 5, 'tigger': 1, 'cartoon': 5, 'scored': 5, 'tupac': 1, 'janet': 2, 'tackles': 1, 'idea': 1, 'discrimination': 1, 'b': 7, 'goonies': 8, 'major': 9, 'serves': 1, 'rate': 15, 'kissing': 2, 'deseree': 1, 'cannibal': 1, 'holocaust': 4, 'pixars': 2, 'stanley': 23, 'techno': 2, 'gets': 6, 'week': 1, 'claus': 1, 'windtalkers': 1, 'humphry': 1, 'zemeckis': 9, 'chocolates': 2, 'computer': 7, 'graphics': 1, 'encounters': 1, '1981': 4, 'jonas': 1, 'fidelity': 1, 'lebowski': 1, 'casinos': 1, 'kyle': 5, 'charlize': 12, 'therons': 1, 'gangs': 3, 'clips': 7, 'winnie': 1, 'pooh': 1, 'tarantino': 12, 'arthur': 10, 'chocolat': 1, 'giant': 7, 'cock': 1, 'blocking': 1, 'jam': 7, 'composer': 3, 'amelie': 1, 'friends': 16, 'buzz': 12, 'lightyear': 12, 'charley': 1, 'chaplin': 18, 'women': 7, 'german': 3, 'burt': 7, 'goobers': 1, 'titles': 36, 'nation': 2, 'lampoon': 1, 'depicts': 1, 'earth': 13, 'being': 24, 'destroyed': 1, 'asteroid': 1, 'officers': 1, 'capitalism': 1, 'communism': 1, 'ww': 2, 'helped': 1, 'guote': 1, 'odyssey': 8, 'lean': 16, 'agent': 7, 'crackers': 1, 'soup': 2, 'change': 4, 'habit': 1, 'johhny': 2, 'alice': 3, 'cooper': 8, 'viewed': 1, 'frankie': 2, 'avalon': 2, 'beach': 7, 'scrooged': 1, 'extramarital': 1, 'affairs': 2, 'sound': 4, 'nothing': 4, 'tv': 6, 'prostitution': 3, 'bodyguard': 2, 'johanssons': 1, 'roles': 8, 'presleymovie': 1, 'las': 4, 'vegas': 5, 'charlotte': 1, 'hall': 5, 'wallis': 1, 'madison': 4, 'dance': 6, 'composed': 2, 'arabia': 2, 'ariel': 1, 'fonda': 17, 'tonight': 6, 'matthew': 31, 'broderick': 8, 'romances': 2, 'saving': 4, 'private': 6, 'ronald': 2, 'regan': 1, 'omar': 2, 'gooding': 3, 'ratings': 190, 'above': 54, 'bays': 1, 'gaston': 1, 'nimoy': 5, 'gran': 1, 'torino': 1, 'once': 2, 'december': 2, 'biblical': 2, 'charleton': 1, 'heston': 10, 'colors': 1, 'feed': 1, 'search': 17, 'fireworks': 2, 'ricky': 5, 'gervais': 1, 'pair': 1, 'patriot': 1, 'pow': 2, 'camps': 2, 'americans': 2, 'spacey': 8, 'half': 8, 'baked': 5, 'et': 7, 'cat': 7, 'hit': 9, 'mission': 7, 'impossible': 6, 'gantry': 1, 'inception': 72, 'much': 7, 'money': 9, 'earn': 1, 'sow': 1, 'rainbow': 8, 'blacksploitation': 3, 'voodoo': 1, 'saw': 7, 'jingle': 1, 'way': 3, 'singlteon': 1, 'parody': 8, 'frankenweenie': 1, 'raja': 1, 'court': 1, 'vinny': 2, 'charlton': 10, 'hestons': 1, 'final': 4, 'died': 1, 'fancy': 1, 'gown': 1, 'blitz': 1, 'japanese': 4, 'mushrooms': 1, 'mcgregor': 2, 'six': 171, 'goodbar': 1, 'skyscraper': 1, 'dad': 2, 'ahead': 4, 'marmalade': 1, 'synopsis': 2, 'comic': 6, 'africa': 5, 'soundrack': 1, 'ferris': 4, 'buelers': 4, 'mivie': 1, 'antagonist': 1, 'invisible': 2, 'cate': 5, 'blanchett': 5, 'les': 3, 'starts': 6, 'titans': 1, 'romero': 1, 'happening': 1, 'reading': 1, 'theron': 11, 'moranis': 2, 'balls': 4, 'read': 1, 'eastbound': 1, 'threme': 1, 'futureovies': 1, 'march': 2, 'roddy': 5, 'mcdowel': 4, 'isaaks': 1, 'grisham': 1, 'spencer': 13, 'tracy': 16, '1965': 2, 'creature': 4, 'gollum': 1, 'metaphor': 1, 'environmental': 1, 'causes': 1, 'commander': 2, 'data': 1, 'folk': 1, 'bands': 2, 'extra': 4, 'terrestrial': 4, 'previews': 2, 'sandlers': 1, 'zoolander': 1, 'heavyweights': 1, 'pan': 2, 'episode': 3, 'paul': 42, 'newman': 15, 'jazz': 2, 'viwers': 1, 'why': 4, 'married': 4, 'spy': 8, 'obsessed': 1, 'sings': 4, 'blowers': 1, 'daughter': 5, 'field': 5, 'sharon': 3, 'meat': 2, 'gross': 7, 'ultimatum': 2, 'cranky': 1, 'teething': 1, 'wich': 1, 'forever': 1, 'salsa': 1, 'weekend': 3, 'bernies': 1, 'pop': 2, 'culture': 1, 'rutger': 1, 'hauer': 1, 'darrel': 2, 'hanna': 3, 'brooks': 21, 'o': 2, 'brother': 2, 'thou': 1, 'sub': 1, 'titled': 24, 'samuel': 12, 'l': 16, 'frost': 2, 'witness': 1, 'precognitive': 1, 'along': 2, 'wha': 1, 'karate': 1, 'pipi': 2, 'longstocking': 2, 'mean': 3, 'girls': 7, 'blade': 3, 'runner': 2, 'flaming': 1, 'computers': 2, 'sant': 1, 'val': 6, 'kilmer': 6, 'romeros': 1, 'direcorial': 1, 'debut': 2, 'person': 5, 'few': 2, 'disneypixar': 1, '2009': 8, 'realesed': 1, 'katy': 3, 'liquid': 1, 'terminatior': 1, 'ai': 1, 'artificial': 3, 'intelligence': 6, 'rebel': 6, 'against': 1, 'golf': 3, '1985': 4, 'watched': 3, 'benefits': 4, 'razzie': 2, 'depps': 2, 'piece': 4, 'frankly': 2, 'dear': 2, 'dont': 9, 'damn': 3, 'acedemy': 1, 'nominees': 1, 'often': 2, 'shown': 5, 'wa': 1, 'legend': 5, 'katherine': 7, '2pac': 1, 'shakur': 1, 'whar': 1, 'doing': 1, 'appearance': 4, 'commando': 1, 'valkyrie': 1, 'meyers': 2, 'fincher': 14, '2008': 7, 'rotten': 2, 'tomatoes': 2, 'ant': 1, 'bully': 3, 'yul': 2, 'brenner': 2, 'death': 29, 'forrester': 2, 'boorman': 4, 'unforgiven': 2, 'biker': 2, 'marlboro': 1, 'rod': 3, 'serling': 1, 'leslie': 5, 'gore': 7, 'sing': 5, 'sunshine': 1, 'lolipops': 1, 'rainbows': 1, '1958': 1, 're': 2, 'written': 2, 'wiig': 1, 'fletcher': 1, 'conspiracy': 1, 'theory': 1, 'rhett': 1, 'butler': 4, 'boyle': 5, 'twist': 3, 'shout': 1, 'version': 8, 'popeye': 2, 'schindlers': 1, 'colin': 6, 'higgins': 1, 'dolly': 2, 'parton': 2, 'hal': 1, 'plant': 1, 'tirantino': 1, 'supernatural': 4, 'hakuna': 1, 'matata': 1, 'destination': 2, 'olivia': 4, 'newton': 5, 'titantic': 1, 'players': 2, 'eisenhower': 1, 'power': 12, 'outer': 3, 'limits': 2, 'episodes': 1, 'included': 5, 'kidmans': 2, 'neilson': 1, 'godfellas': 1, 'howards': 1, 'production': 4, 'partner': 1, 'micky': 1, 'rooney': 1, 'own': 2, 'arc': 2, 'daniel': 22, 'jordan': 4, 'p': 4, 'tv13': 1, 'darby': 1, 'shaw': 1, 'annette': 5, 'blanket': 1, 'bingo': 1, 'sabrina': 3, 'blood': 11, 'honey': 2, 'psycho': 2, 'babysitter': 3, 'wife': 7, 'south': 7, 'african': 5, 'cage': 11, 'hollywood': 4, 'redford': 9, 'mcqueen': 13, 'poor': 1, 'west': 7, 'side': 8, 'bieber': 1, 'native': 2, 'acorns': 1, 'trainer': 1, 'working': 5, 'girl': 23, '1939': 1, 'cancer': 3, 'washingtons': 1, 'trouble': 1, 'considering': 2, 'frightening': 1, 'golfing': 1, 'tubthumping': 1, 'chased': 1, 'bear': 5, 'wood': 10, 'sheriff': 1, 'brody': 2, 'nolan': 15, 'rival': 1, 'magician': 1, 'broke': 3, 'records': 2, '1000': 2, 'corpes': 1, 'doris': 2, 'sharp': 1, 'shooter': 1, 'aragorn': 1, 'queens': 1, 'stared': 6, 'kiddman': 1, 'alices': 1, 'restaurant': 1, 'flying': 3, 'repeated': 1, 'scarface': 4, 'gangstas': 1, 'paradise': 1, 'wit': 1, 'ustinov': 3, 'pat': 1, 'benetar': 1, 'q': 1, 'rap': 1, 'mike': 12, 'tyson': 4, 'tiger': 3, 'gallo': 5, 'elton': 1, 'chocolate': 7, 'factory': 9, 'gilmore': 1, 'shakespeare': 4, 'denmark': 1, 'lamp': 1, 'nc17': 1, 'wname': 1, 'collaborates': 1, 'iv': 2, 'gods': 1, 'framed': 2, 'rabbit': 3, 'met': 4, 'sally': 2, 'crystals': 1, 'young': 11, 'frankenstein': 6, 'mental': 5, 'institution': 3, 'pacifist': 1, 'historic': 3, 'raiders': 5, 'ark': 3, 'statham': 1, 'hired': 3, 'balboa': 1, 'actual': 1, 'cult': 3, 'classics': 2, 'cloris': 1, 'leachman': 1, 'oif': 1, 'segel': 2, 'growing': 4, 'seth': 8, 'rogens': 1, 'pineapple': 1, 'express': 1, 'banned': 1, 'ship': 6, 'wwi': 1, 'middle': 4, 'east': 2, 'kelly': 15, 'debbie': 3, 'mall': 1, 'butterfinger': 1, 'nautical': 3, 'huston': 12, 'heavy': 3, 'metal': 3, 'tribute': 3, 'achilles': 1, 'endless': 1, '1940': 127, '1970': 162, 'ellen': 2, 'roark': 1, 'bars': 1, 'oldman': 6, 'rachel': 3, 'mcadams': 2, 'her': 9, 'allie': 1, 'hamilton': 8, 'britanny': 1, 'thin': 2, 'hasta': 1, 'vista': 1, 'start': 3, 'today': 3, 'prostitute': 3, 'godzilla': 3, 'pool': 1, 'monty': 5, 'pythons': 1, 'circus': 3, 'korea': 1, 'luther': 2, 'voight': 1, 'geres': 1, 'spoken': 1, 'orson': 22, 'welles': 22, 'sidney': 1, 'poitier': 1, 'fox': 13, 'v': 2, 'ringwald': 2, 'owning': 1, 'ponies': 1, 'record': 1, 'label': 1, 'rim': 1, 'english': 2, 'speaking': 1, 'salma': 2, 'hayek': 2, 'gleason': 2, 'bale': 15, 'troopers': 1, 'racing': 6, 'col': 1, 'oneill': 2, 'stargate': 1, 'continuum': 1, 'forgetting': 2, 'sarah': 8, 'britan': 1, 'myers': 3, 'moldy': 1, 'peaches': 1, 'parents': 1, 'blind': 5, 'dracula': 5, 'apolo': 1, 'multiplicity': 1, 'dealing': 11, 'disease': 1, 'roll': 1, 'scout': 1, 'diamonds': 1, 'cats': 3, 'showcase': 1, 'acting': 6, 'talent': 1, 'pierce': 7, 'brosnan': 2, 'bedazzled': 1, 'watson': 6, 'circle': 1, 'perrys': 1, 'screen': 1, 'isaac': 3, 'biopic': 2, 'blading': 1, 'moves': 4, 'kong': 2, 'gorrilla': 1, 'h': 5, 'macy': 3, 'earned': 4, 'firth': 3, 'katniss': 2, 'success': 2, 'independent': 31, 'neil': 1, 'harris': 4, 'travelling': 2, 'astronaut': 1, 'peace': 1, 'driving': 2, 'daisy': 1, 'hercules': 2, 'york': 10, 'undertaker': 2, 'buried': 3, 'similiar': 1, 'youve': 1, 'mail': 1, 'air': 5, 'force': 2, 'pulp': 1, 'twelve': 1, 'mainstream': 2, 'sick': 2, 'killers': 5, 'saying': 8, 'shot': 7, 'stardom': 1, 'containing': 3, 'koreas': 1, 'musketeers': 2, 'theplot': 1, 'while': 1, 'sleeping': 1, 'lassie': 1, 'lucille': 2, 'provide': 15, 'cleese': 5, 'somg': 1, 'notting': 2, 'farmer': 1, 'corn': 1, 'boy': 11, 'saves': 2, 'gabor': 2, 'piper': 1, 'arent': 2, 'violent': 7, 'liberace': 1, 'titular': 1, 'track': 3, 'focus': 7, 'concept': 2, 'seeing': 4, 'ground': 1, 'leaves': 2, 'then': 3, 'animation': 70, 'braveheart': 4, 'stein': 1, 'wish': 5, 'dogfight': 1, 'dragon': 3, 'tatoo': 1, 'mob': 8, 'live': 7, 'banjo': 1, 'belong': 1, 'airport': 1, 'cloning': 1, 'morgan': 12, 'freeman': 11, 'mila': 2, 'kunis': 2, 'johnson': 7, 'dicaprios': 2, 'pregnant': 1, 'teenager': 2, '007': 1, 'moviies': 1, 'perfrmance': 1, 'stuart': 7, 'apes': 2, 'mary': 7, 'reilly': 1, 'mustangs': 1, 'portmans': 1, 'robbie': 1, 'skywalker': 1, 'waterfront': 1, 'sometime': 2, '1994': 3, 'simba': 1, 'potte': 1, 'mrovies': 1, 'gonna': 2, 'fly': 4, 'zimmer': 2, 'dinasour': 1, 'fron': 1, 'cushing': 2, 'eddie': 9, 'murphys': 2, 'paris': 3, 'right': 48, 'caddyshack': 2, 'acts': 1, 'sword': 2, 'sorcery': 2, 'madonna': 4, 'went': 3, 'undercover': 5, 'beauty': 10, 'pageant': 1, 'emmy': 2, 'rossum': 2, 'bulma': 1, 'navy': 3, 'pilot': 2, 'warden': 1, 'lola': 1, 'swahili': 1, 'gangster': 22, 'wonka': 8, 'choclate': 1, 'witch': 14, 'wizard': 7, 'oz': 4, 'jean': 11, 'dujardin': 1, 'translation': 1, 'rihanna': 1, 'current': 1, 'ends': 2, 'ester': 1, 'seabiscuit': 1, 'possible': 3, 'fun': 4, 'foxs': 1, 'berrys': 1, 'hungry': 1, 'again': 3, 'originate': 1, 'christo': 1, 'son': 3, 'kills': 2, 'mother': 5, 'leigh': 21, 'anne': 5, 'tuohy': 1, 'extreme': 5, 'violence': 17, 'work': 5, 'hughes': 3, '1984': 2, 'cg': 1, 'presleyovie': 1, 'kubricks': 1, 'johannsen': 1, 'gloria': 4, 'grahame': 2, '1960': 169, 'dolphins': 2, 'frog': 2, 'shirley': 8, 'earliest': 1, 'johnathan': 1, 'tayler': 1, '28': 2, 'wonderland': 3, 'denver': 2, 'don': 9, 'knotts': 3, 'rollerball': 1, 'pony': 1, 'entirely': 2, 'furry': 2, 'monster': 8, 'worked': 5, 'green': 7, 'eyed': 2, 'gang': 5, 'warriors': 2, 'avildsen': 1, 'karat': 1, 'night': 29, 'voldemort': 1, 'bythe': 2, 'danner': 2, '1979': 4, 'forrest': 2, 'gump': 1, 'maids': 1, 'greek': 1, 'mythology': 2, '1950': 162, 'guys': 2, 'escaped': 1, 'chain': 2, 'supermans': 1, 'bakeds': 1, 'if': 17, 'so': 7, 'ones': 1, 'redd': 2, 'babysitters': 2, 'issaiah': 1, 'kurt': 5, 'russell': 15, 'ritchies': 1, 'pets': 2, 'glen': 2, 'russia': 1, 'socery': 1, 'bugs': 2, 'regarded': 1, '1943': 1, 'spaghetti': 41, 'galahad': 1, 'videodrome': 1, 'pictures': 4, 'bollywood': 3, 'prominent': 3, 'panama': 1, 'halen': 1, 'airheads': 1, 'serial': 4, 'hughs': 1, 'mannequin': 1, 'experiments': 5, 'barber': 1, 'writtin': 1, 'rogan': 1, 'ringo': 1, 'starr': 1, 'prometheus': 2, 'aside': 2, 'student': 1, 'winfield': 1, 'sottish': 1, 'deep': 4, 'lauren': 5, 'bacall': 3, 'shogun': 1, 'sherlock': 4, 'holmes': 7, 'joins': 1, 'carnival': 1, 'sam': 14, 'worthington': 1, 'penguins': 1, 'ted': 5, 'lawyer': 1, 'hidden': 2, 'psychic': 3, 'abducts': 1, 'child': 25, 'rko': 1, 'ginger': 4, 'rogers': 6, 'ralph': 9, 'latifah': 3, 'italy': 3, 'veteran': 1, 'dinosaurs': 5, 'empire': 2, 'craig': 6, 'kindergarten': 1, 'diver': 2, 'saga': 2, 'eclipse': 2, 'doug': 3, 'glatt': 1, 'schooner': 1, 'sailboat': 1, 'unexpected': 1, 'journey': 8, '14th': 1, 'tony': 9, 'randal': 2, 'jetsons': 1, 'jet': 5, 'li': 4, 'opposite': 5, 'tights': 1, 'goldie': 4, 'hawn': 6, 'surfing': 1, 'laura': 6, 'linney': 1, 'kicking': 1, 'screaming': 2, 'hot': 3, 'gorgo': 1, '300': 4, 'rodney': 6, 'dangerfield': 3, '1968': 1, 'solved': 1, 'thier': 1, 'issues': 5, 'season': 3, 'polynesian': 1, 'bettany': 1, 'charles': 22, 'darwin': 1, 'gogh': 1, 'silberling': 1, '1974': 1, 'sid': 1, 'krofft': 1, 'lmovie': 1, 'mason': 2, 'boxer': 4, 'bamed': 1, 'flynn': 3, 'geoffrey': 2, 'marquis': 1, 'sade': 1, 'punk': 1, 'eddy': 6, 'kathy': 4, 'bates': 3, 'youth': 2, 'farce': 3, 'alli': 1, 'magraw': 1, 'addiction': 2, 'claude': 3, 'damme': 2, 'almighty': 1, 'location': 1, 'cyd': 1, 'charisse': 1, 'danced': 1, 'deborah': 1, 'foreman': 1, 'bulge': 1, 'fresno': 1, 'california': 1, 'webb': 2, 'october': 2, 'diazs': 2, 'civil': 9, 'supremacy': 1, 'freida': 1, 'mock': 1, 'jericho': 2, 'wanna': 1, 'cannonball': 1, 'thelma': 1, 'louise': 1, 'opera': 6, 'bernie': 3, 'mac': 3, 'bath': 1, 'tub': 2, 'nims': 1, 'smiths': 2, 'con': 3, 'becomes': 1, 'commodities': 1, 'trader': 1, 'parsons': 1, 'plots': 5, 'doctor': 4, 'nyc': 1, 'kill': 10, 'members': 2, 'katie': 2, 'mistaken': 2, 'scorcese': 1, 'blanc': 2, 'kennedy': 2, 'marooned': 2, 'committed': 1, 'tommy': 10, 'lights': 3, 'call': 3, '1900s': 1, 'stood': 1, 'still': 3, 'treasure': 5, 'pioneers': 2, 'fisher': 2, 'soldiers': 7, 'jolson': 1, 'littlest': 1, 'workers': 1, 'lily': 1, 'tomlin': 1, 'gambling': 1, 'marvin': 2, 'throws': 1, 'boiling': 1, 'coffee': 1, 'grahames': 1, 'face': 2, 'soul': 3, 'gandalf': 2, 'niro': 4, 'verbinski': 3, 'viggo': 5, 'mortensen': 5, 'fasten': 1, 'seatbelts': 1, 'ides': 1, 'crawford': 5, 'governor': 1, 'louisiana': 1, 'costners': 1, 'contacts': 1, 'trailerfor': 1, 'ewoks': 1, 'hays': 1, 'escape': 5, 'turkish': 1, 'kis': 1, 'meryly': 1, 'lawyers': 2, 'each': 1, 'pageants': 1, 'merchant': 1, 'antony': 2, 'zelweger': 1, 'lindsey': 2, 'feels': 1, 'gangsta': 1, 'cagney': 12, 'quaids': 1, 'keaches': 1, 'carradines': 1, 'jesse': 5, 'racial': 2, 'differences': 1, 'soon': 9, 'penelope': 2, 'cruzs': 1, 'language': 2, 'filmography': 1, 'akira': 8, 'kurusawas': 1, 'fools': 1, 'magicians': 1, '571': 1, 'mexican': 3, 'powered': 2, 'individual': 2, 'million': 3, 'dollar': 2, 'corpse': 1, 'kubrik': 1, 'barbara': 8, 'stanwyck': 4, 'mcmurray': 1, 'snider': 1, 'strangland': 1, 'belushi': 5, 'cameos': 1, 'freddie': 1, 'sonny': 1, 'steele': 2, 'citizen': 3, 'kane': 3, 'chinese': 2, 'crenna': 1, 'salesman': 2, 'swayzes': 1, 'apocalypse': 4, 'olaf': 1, 'asian': 1, 'soccer': 5, 'pryor': 6, 'anti': 1, 'erasing': 2, 'memory': 7, 'catherine': 8, 'zeta': 5, 'andersons': 1, 'window': 3, 'monkeys': 2, 'richie': 2, 'rich': 2, 'return': 4, 'jedi': 2, 'categories': 8, 'tarot': 1, 'card': 1, 'reader': 1, 'apocalpse': 1, 'swordsman': 1, 'party': 10, 'werner': 2, 'herzog': 2, 'threatened': 1, 'dimples': 1, 'provided': 1, 'himself': 1, 'kruger': 1, 'snape': 1, 'francos': 1, 'dumb': 1, 'dumber': 1, 'lightning': 2, 'found': 6, 'believe': 2, '1944': 1, 'born': 2, 'connick': 1, 'jr': 22, 'doll': 4, 'chuckie': 1, 'artist': 1, 'colonel': 1, 'fosse': 1, 'dowop': 1, 'veiwers': 1, 'ozarks': 1, 'emil': 1, 'renny': 1, 'overboard': 1, 'winter': 4, '68th': 1, 'rose': 4, 'korean': 1, 'punks': 1, 'hugh': 8, 'project': 3, 'better': 7, 'hangover': 2, 'upcoming': 5, 'lizard': 2, 'elephant': 1, 'kline': 3, 'helena': 5, 'bonham': 4, 'carter': 11, 'ancient': 1, 'indians': 4, 'mexico': 5, 'latin': 1, 'revolutions': 2, 'goldgerg': 1, 'hitchcocks': 1, 'french': 6, 'captain': 6, 'picard': 1, 'cheerleaders': 2, 'dan': 12, 'akyrod': 1, 'marlon': 12, 'brando': 8, 'we': 5, 'consider': 2, 'relied': 1, 'kindness': 1, 'strangers': 1, 'herbie': 2, 'bananas': 1, 'duvall': 9, 'country': 3, 'wrestling': 3, 'lightening': 1, 'rainmaker': 1, 'bongo': 1, 'mark': 22, 'wahlberg': 2, 'mccall': 1, 'different': 2, 'voorhees': 1, 'ex': 1, 'nba': 1, 'martins': 2, 'madam': 1, 'cullen': 4, 'moss': 1, 'dalai': 1, 'lama': 1, 'sloth': 1, 'brooke': 2, 'shields': 2, 'orleans': 1, 'prostitutes': 1, 'ladies': 2, 'oprah': 1, 'winfrey': 1, 'henson': 2, 'recommend': 30, 'australian': 1, 'pearce': 1, 'bears': 2, 'strips': 1, 'nude': 2, 'sjywalker': 1, 'lemmon': 14, 'avanti': 1, 'bag': 2, 'lions': 2, 'audrey': 9, 'nun': 1, 'belgian': 1, 'congo': 1, 'complete': 2, 'told': 2, 'kidnapped': 3, 'caine': 10, 'ireland': 1, 'atleast': 1, 'dreaming': 1, 'living': 7, 'kenny': 2, 'loggins': 1, 'akyroyd': 1, 'englends': 1, 'elm': 1, 'st': 2, 'college': 5, 'spiral': 2, 'staircase': 1, 'dorothy': 2, 'mcguire': 1, 'amnesia': 1, 'sufferer': 1, 'babylon': 2, 'adwhat': 1, 'tattooed': 1, 'diesels': 1, 'hand': 1, 'inceptioncategories': 1, 'england': 2, 'those': 1, 'brasher': 1, 'doubloon': 1, 'merlin': 2, 'kim': 5, 'basinger': 2, 'speilburg': 1, 'fortress': 2, 'ginnifer': 1, 'goodwin': 1, 'purple': 9, 'rain': 12, 'joyful': 2, 'holden': 2, 'joan': 12, 'bennett': 1, 'angela': 4, 'lansbury': 1, 'alaska': 2, 'gwyenth': 1, 'paltrow': 2, 'radio': 2, 'lounge': 1, 'annie': 1, 'racer': 1, 'whisperer': 1, 'louis': 3, 'malle': 2, 'lana': 1, 'european': 1, 'thornton': 2, 'alyson': 3, 'kidnapping': 1, 'fist': 2, 'nesbitt': 1, 'leave': 1, 'law': 2, 'tamilyn': 1, 'tomita': 1, 'heralded': 1, 'adapted': 1, 'wild': 12, 'burgess': 1, 'rabbits': 2, 'glow': 1, 'condom': 1, 'short': 49, 'megan': 3, 'actionsci': 2, 'infamous': 2, 'thorton': 2, 'ferrel': 2, 'c': 9, 'riely': 1, 'fat': 1, 'liotta': 2, 'amp': 4, 'hayak': 2, 'thunder': 1, 'superheros': 1, 'between': 6, 'battlefield': 3, 'help': 23, 'fellow': 1, 'tragedy': 4, 'sinking': 1, 'calamity': 1, 'boyer': 1, 'tries': 1, 'convince': 1, 'ingrid': 4, 'wyatt': 2, 'earp': 2, 'holliday': 2, 'fraser': 2, 'navarone': 2, 'whole': 6, 'rodrick': 1, 'estelle': 1, 'getty': 1, 'newly': 1, 'murderer': 3, 'united': 2, 'states': 1, 'fluffy': 1, 'deanna': 1, 'durbin': 1, 'francois': 7, 'truffaut': 7, 'scientist': 2, 'klingons': 1, 'fights': 2, 'oddjob': 1, 'seen': 19, 'stapler': 1, 'case': 3, 'mondays': 1, 'snakes': 3, 'welcome': 2, 'carmen': 4, 'someone': 18, 'dears': 1, 'neve': 1, 'trey': 1, 'sleepless': 1, 'seattle': 1, 'holds': 1, 'takes': 8, 'hostages': 1, 'bud': 4, 'receiver': 1, 'lupus': 1, 'skin': 2, 'revolutionary': 2, 'flight': 1, 'navigator': 1, 'bane': 1, 'cruses': 1, 'sisters': 1, 'lina': 1, 'wertmuller': 1, 'candice': 2, 'pig': 2, 'demons': 6, 'flesh': 5, 'fishes': 1, 'dori': 1, 'deneuve': 2, 'penny': 2, 'sees': 3, 'fountain': 1, 'suceessful': 1, 'unlaterally': 1, 'deliverance': 1, 'holly': 1, 'hunter': 9, 'kidnap': 1, 'uma': 2, 'thurman': 2, 'hayes': 2, 'regina': 2, 'speedway': 2, 'malcom': 3, 'radcliffe': 1, 'dances': 2, 'snake': 1, 'hudson': 5, 'race': 4, 'cash': 1, 'gwyneth': 1, 'receives': 1, 'former': 1, 'beetlejuice': 5, 'stop': 1, 'believin': 1, 'deneuvre': 1, 'mom': 2, 'wjat': 1, 'isla': 1, 'lang': 9, 'fishers': 1, 'soap': 1, 'elisabeth': 4, 'shue': 4, 'mclaine': 1, 'marriage': 5, 'jude': 1, 'kelley': 1, 'submarines': 1, 'buddy': 3, 'zanzibar': 1, 'snatch': 2, 'gennie': 1, 'pigs': 1, 'sequelsprequels': 1, 'marathon': 1, 'deal': 2, 'childhood': 2, 'memories': 1, 'jungle': 5, 'soldier': 2, 'vietnam': 3, 'montana': 1, 'pacinos': 1, 'newer': 1, 'backward': 1, 'aging': 1, 'father': 5, 'emily': 1, 'blunt': 1, 'wears': 1, 'prada': 1, 'gotta': 1, 'friend': 7, 'received': 211, 'scotts': 1, 'elliott': 3, 'gould': 2, 'teller': 1, 'raging': 2, 'bull': 2, 'shyamalan': 2, 'order': 6, 'gerald': 1, 'step': 1, 'roller': 2, 'derby': 1, 'aids': 2, 'hiv': 1, 'rosario': 1, 'dawson': 3, 'pearl': 2, 'bailey': 1, '1987': 1, 'chess': 1, 'match': 1, 'joy': 2, 'schwarzeneggers': 1, 'katharine': 10, 'cape': 1, 'fear': 11, 'iron': 2, 'catchphrase': 1, 'enjoy': 2, 'later': 4, 'elijah': 3, 'turns': 1, 'into': 7, 'chariots': 3, 'fire': 6, 'christie': 2, 'michelle': 6, 'trilogies': 2, 'angie': 1, 'dickinson': 1, 'speak': 1, 'inmvented': 1, 'depression': 3, 'listall': 1, 'ducks': 2, 'ledgers': 1, 'paranorman': 1, 'jenny': 3, 'mccarthy': 3, 'swank': 1, 'methods': 1, 'game': 3, 'python': 2, 'holy': 5, 'grail': 3, 'kiss': 10, 'ruin': 1, 'heigel': 1, 'clark': 4, 'gable': 3, 'reporter': 1, 'claudette': 1, 'colbert': 1, 'heiress': 1, 'jiminy': 1, 'cricket': 1, 'lancaster': 3, 'espianoge': 1, 'husband': 3, 'michigan': 1, 'hereafter': 1, 'orwells': 1, 'liam': 11, 'neeson': 10, 'slap': 1, 'mimi': 2, 'duchovney': 1, 'lynch': 14, 'vivien': 14, 'raiser': 2, 'cosby': 4, 'roadhouse': 1, 'ride': 2, 'surf': 1, 'jonze': 1, 'beyond': 3, 'unfortunate': 1, 'events': 1, 'crab': 2, 'grodin': 1, 'whom': 1, 'nigel': 1, 'suspicion': 1, 'highschool': 1, 'dennis': 4, 'hopper': 2, 'canadian': 1, 'leo': 1, 'gregory': 14, 'brian': 23, 'broadway': 4, 'shows': 1, 'dwayne': 1, 'nazis': 2, 'cornfield': 1, 'owl': 1, 'rocket': 1, 'waiting': 1, 'tables': 1, 'june': 2, 'laughton': 1, 'tyrone': 2, 'trial': 2, 'panic': 2, 'room': 3, 'past': 430, 'turtle': 1, 'stripes': 1, 'perlman': 2, 'assasins': 1, 'nerds': 1, 'ken': 3, 'pioneered': 1, 'cgi': 3, 'techniques': 1, 'secretariat': 1, 'cure': 1, 'ballet': 1, 'costner': 5, 'dystopian': 1, 'bomb': 2, 'longest': 1, 'yard': 1, 'decaprio': 1, 'controlling': 1, 'franki': 1, 'lane': 3, 'pet': 2, 'versions': 1, 'carol': 3, 'miracle': 1, 'juno': 1, 'ghostbusters': 2, 'angels': 5, 'duel': 1, 'burns': 3, 'danson': 1, 'sammy': 1, 'rhythm': 1, 'lifes': 1, 'worker': 1, 'coppolas': 1, 'pendleton': 1, 'heaven': 4, 'butterfly': 2, 'effect': 1, 'ws': 1, 'noyce': 1, 'moneyball': 1, 'nudity': 1, 'andy': 8, 'perfect': 3, 'storm': 2, 'bite': 1, 'racehorse': 1, 'bruckheimer': 1, 'houses': 1, 'avatars': 1, 'differ': 1, 'hasselhoff': 1, 'mutiny': 2, 'footloose': 2, 'coaches': 1, 'championship': 1, 'cross': 1, 'dressing': 1, 'masked': 1, 'minute': 1, 'lucy': 3, 'liu': 1, 'linda': 5, 'ronstadt': 1, 'here': 4, 'reynold': 1, 'plimpton': 1, 'alda': 3, 'collette': 1, 'greg': 8, 'kinnear': 1, 'contest': 1, 'scorseses': 2, 'arkin': 1, 'terrorizes': 1, 'gigi': 1, 'scottish': 1, 'schoolteacher': 1, 'veronica': 2, 'lake': 5, 'nicolas': 6, 'smashing': 1, 'pumpkins': 1, 'jake': 2, 'gyllanhall': 1, 'airlines': 1, 'plane': 4, 'brue': 1, 'walter': 5, 'mathau': 1, '3000': 1, 'kaiser': 1, 'soze': 1, 'marijuana': 1, 'ghandi': 1, 'animates': 1, 'rights': 3, 'crossdressing': 2, 'diggstown': 1, 'scanner': 1, 'darkly': 1, 'sir': 1, 'eye': 2, 'trolley': 1, 'loaf': 1, 'ovie': 1, 'baskervilles': 1, 'gutenburg': 1, 'eve': 5, 'rib': 1, 'pokemon': 1, 'brians': 1, 'caan': 1, 'omedy': 1, 'toby': 3, 'keith': 8, 'pup': 1, 'leader': 1, 'm2m': 1, 'adaptation': 2, 'jk': 1, 'rowling': 1, 'number': 2, 'rule': 1, 'aviator': 1, 'publisher': 1, 'jameson': 1, 'bird': 5, 'crimson': 1, 'tide': 2, 'caddy': 1, 'crippled': 1, 'hoodwinked': 1, 'utopia': 3, 'spiders': 1, 'biopics': 1, 'rudd': 1, 'cuba': 2, 'stealing': 1, 'befriending': 1, 'truck': 1, 'dean': 7, 'willem': 1, 'dafoe': 1, 'chopper': 1, 'frodo': 2, 'bunny': 2, 'descendants': 1, 'tin': 2, 'cup': 1, 'barbra': 1, 'streisand': 1, 'rogen': 1, 'banks': 1, 'bobby': 1, 'cruz': 3, 'era': 2, 'alcoholism': 2, 'everybody': 2, 'wants': 2, 'thanksgiving': 1, 'cooking': 1, 'debra': 8, 'messing': 5, 'mmpa': 1, 'basketnall': 1, 'diaries': 3, 'sequal': 1, 'wildcats': 1, 'hello': 1, 'nielson': 3, 'drum': 1, 'wheedon': 1, 'guess': 1, 'dinner': 1, 'kept': 1, 'fromthe': 1, 'dawn': 4, 'kiefer': 3, 'versus': 4, 'martians': 1, 'contain': 2, 'cruises': 2, 'fantacy': 1, 'cruse': 1, 'apocolypse': 1, 'lambada': 1, 'spinal': 1, 'tap': 1, 'kardashian': 1, 'contribute': 1, 'hop': 1, 'springtime': 1, 'taxi': 5, 'narnia': 1, 'adaption': 1, 'weithorn': 1, 'valet': 1, 'sells': 1, 'secrets': 5, 'eat': 1, 'worlds': 2, 'hookey': 1, 'phantom': 1, 'criticism': 1, 'robbins': 3, 'prisoner': 5, 'gelenhal': 1, 'actually': 2, 'changed': 1, 'bringing': 2, 'huge': 1, 'worms': 1, 'combine': 1, 'rookie': 1, 'liv': 2, 'took': 3, 'faces': 3, 'earnest': 1, 'borgnine': 1, 'landau': 1, 'older': 1, 'bela': 3, 'lugosi': 3, 'jodi': 1, 'forster': 1, 'running': 1, 'brave': 1, 'shower': 1, 'wanderin': 1, 'brown': 4, 'sugar': 1, 'tinman': 1, 'teresa': 1, 'opening': 2, 'bennie': 1, 'joone': 1, 'listen': 1, 'actin': 1, 'seagal': 2, 'mcdowall': 2, 'gay': 1, 'themes': 1, 'subcultures': 1, 'keenan': 1, 'bought': 1, 'wachowski': 2, 'fame': 1, 'bowie': 2, 'childs': 1, 'dakota': 1, 'fanning': 1, 'lots': 1, 'singers': 2, 'f': 4, 'gray': 4, 'chameleon': 1, 'shreck': 1, 'christine': 5, 'captian': 1, 'locked': 1, 'basement': 1, 'dana': 2, 'jfks': 1, 'funeral': 4, 'quazi': 1, 'motto': 1, 'loser': 1, 'photographer': 1, 'zealand': 1, 'mcclain': 2, 'fair': 2, 'weather': 2, 'dailey': 1, 'jaime': 2, 'clarkson': 1, 'beginning': 2, 'brokeback': 1, 'mountain': 3, 'because': 1, 'seasons': 2, 'gossip': 1, 'poehler': 1, 'rates': 1, 'hubbard': 1, 'accused': 1, 'low': 2, 'brick': 2, 'leavitts': 1, 'knowing': 1, 'whistle': 1, 'full': 5, 'yoda': 6, 'anger': 1, 'hate': 1, 'sting': 3, 'mirror': 2, 'bus': 1, 'amber': 4, 'waves': 1, 'cleef': 1, 'fully': 1, 'loaded': 1, 'fir': 1, 'reeve': 1, 'mcclure': 1, 'breath': 1, 'airplanes': 1, 'victor': 3, 'mature': 1, 'nick': 5, 'norah': 1, 'town': 8, 'station': 2, 'zebra': 1, 'crush': 1, 'farrah': 1, 'faucett': 1, 'according': 1, 'coach': 3, 'ozzy': 1, 'osbourn': 1, 'rag': 2, 'square': 1, 'wide': 1, 'shut': 1, 'gi': 1, 'prodigy': 3, 'pianist': 1, 'atlanta': 1, 'soft': 1, 'core': 1, 'porn': 1, 'hotel': 2, 'hurricane': 1, 'forced': 1, 'hide': 1, 'credits': 1, '60': 1, 'seconds': 1, 'za': 2, 'spent': 1, 'long': 3, 'centers': 36, 'zellweiger': 2, '1933': 1, 'naomi': 1, 'watts': 1, 'adrien': 1, 'hes': 3, 'tramp': 1, 'rango': 1, 'sleepy': 1, 'hollow': 1, 'curse': 4, 'having': 6, 'perkins': 3, 'sundance': 1, 'our': 3, 'macaulay': 1, 'culkin': 4, 'audi': 1, 'carel': 1, 'locally': 1, 'theater': 2, 'break': 1, 'departed': 2, 'cocoon': 1, 'group': 3, 'including': 3, 'thor': 2, 'warren': 6, 'beatty': 1, 'rosalind': 1, 'nurse': 3, 'helps': 1, 'polio': 1, 'patients': 1, 'ashore': 1, 'steiger': 1, 'attacking': 1, 'zardoz': 1, '20000': 1, 'leagues': 1, 'stooges': 5, 'courtroom': 1, 'florence': 1, 'machine': 3, 'sountracks': 1, 'zone': 2, 'kristin': 1, 'novak': 1, 'electra': 3, 'shamolan': 1, 'mockingbird': 1, 'peck': 11, 'radar': 1, 'oreilley': 1, 'kathryn': 1, 'bigelow': 2, '82nd': 1, 'll': 1, 'cool': 2, 'hilton': 1, 'victim': 1, 'sissy': 2, 'spacek': 2, 'unnatural': 1, 'brandon': 7, 'belle': 1, 'smart': 1, 'tootsie': 1, 'general': 5, 'furby': 1, 'benji': 1, 'altman': 3, 'fashion': 1, 'scarefest': 1, 'sweden': 1, 'chester': 1, 'teacher': 1, 'teaser': 1, 'prometheis': 1, 'blofeld': 1, 'enough': 5, 'attack': 3, 'ricci': 1, 'winona': 2, 'ryder': 2, 'chers': 1, 'daughters': 1, 'album': 1, 'angelica': 2, 'benning': 1, 'artists': 2, 'aladdin': 1, 'tattoo': 1, 'patricia': 3, 'arquette': 8, 'raspberry': 1, 'corman': 1, 'surreal': 1, 'ending': 2, 'sense': 1, 'futuristic': 1, 'everything': 1, 'frozen': 2, 'total': 1, 'degaul': 1, 'let': 5, 'hoosiers': 1, 'octopus': 1, 'devon': 4, 'sawa': 2, 'orlando': 1, 'bloom': 1, 'berenger': 2, 'russian': 6, 'defector': 1, 'maclachlan': 1, 'desert': 2, 'fritz': 10, 'fahrenheit': 1, 'itunes': 1, 'transgender': 1, 'allyson': 1, 'chart': 1, 'topping': 1, 'landis': 1, 'pitch': 1, 'closer': 1, 'clive': 1, 'suggest': 9, 'cort': 1, 'thinks': 1, 'whalberg': 1, 'miley': 1, 'cyrus': 2, 'erich': 1, 'von': 3, 'zipper': 1, 'martial': 1, 'arts': 1, 'chuck': 1, 'norris': 1, 'marilyn': 5, 'monroe': 6, 'farrell': 1, 'temple': 1, 'elite': 1, 'controversial': 1, 'bruno': 2, 'suspenseful': 1, 'tate': 4, 'reno': 2, 'ratinggenre': 1, 'ricki': 1, 'peper': 1, 'slater': 3, 'pheiffer': 1, 'catharine': 1, 'athletes': 1, 'dreyfuss': 1, 'ebsen': 1, 'january': 1, 'showboat': 1, 'venice': 1, 'assasin': 1, 'mathematics': 1, 'bean': 3, 'stage': 1, 'door': 4, 'bullitt': 1, 'whoopie': 1, 'waters': 2, 'modine': 5, 'ocean': 2, 'hands': 1, 'scissors': 1, 'nascar': 1, 'schneider': 5, 's': 752, 'heigl': 1, 'excellent': 57, 'held': 3, 'captive': 1, 'denise': 1, 'richards': 2, 'brent': 2, 'carver': 1, 'roy': 1, 'scheider': 1, 'lena': 3, 'horne': 1, 'juliana': 2, 'margulies': 2, 'biographical': 28, 'wildlife': 4, 'domineering': 2, 'riches': 3, 'glickenhaus': 1, 'decent': 49, 'spade': 2, 'liked': 134, 'lot': 49, 'miguel': 2, 'arteta': 1, 'fate': 2, 'tia': 4, 'mowry': 2, 'garber': 1, 'safehouse': 1, 'shemp': 2, 'demon': 2, 'body': 6, 'laughable': 3, 'nine': 172, 'decades': 335, 'jonathon': 3, 'silverman': 2, 'cadillac': 2, 'dysfunctional': 2, 'renfro': 4, 'sibling': 6, 'rivalry': 6, 'very': 104, 'harts': 1, 'rebellion': 2, 'bei': 1, 'po': 2, 'clippety': 1, 'clobbered': 1, 'cotto': 1, 'pacquiao': 1, '7': 2, '04': 1, 'chad': 4, 'fees': 4, 'abortion': 2, 'janeane': 1, 'garofolo': 1, 'petrarca': 1, 'portrait': 11, 'within': 91, 'eight': 172, 'watchable': 39, 'finnes': 2, 'fake': 2, 'research': 2, 'decade': 106, 'bridget': 4, 'wagner': 3, 'sport': 36, 'hack': 1, 'unrated': 149, 'experimentation': 1, 'wiebe': 1, 'obach': 2, 'mcneil': 3, 'haunting': 3, 'portraying': 1, 'combat': 3, 'matthews': 1, 'counter': 3, 'terrorism': 1, 'hewitt': 1, 'anchors': 1, 'aweigh': 1, 'ti': 2, 'involved': 19, 'entertaining': 2, 'schiller': 1, 'birthday': 1, 'imaginary': 2, 'interesting': 5, 'samantha': 2, 'becke': 2, 'dianne': 1, 'vassey': 2, 'ok': 70, 'using': 1, 'nathan': 2, 'frankowski': 1, 'ethan': 2, 'embry': 1, 'rakoff': 2, 'abandonment': 2, 'lessin': 1, 'struggle': 5, 'donovan': 1, 'tamera': 1, 'prowse': 2, 'fairies': 2, 'asher': 2, 'brough': 2, 'laurent': 1, 'bouzereau': 1, 'carchietta': 1, 'goldy': 3, 'ira': 2, 'sachs': 1, 'passfield': 1, 'shannen': 2, 'doherty': 2, 'thumbs': 52, 'near': 7, 'experiences': 2, 'capra': 18, 'avant': 30, 'garde': 30, 'mockumentaries': 3, 'theatre': 2, 'audience': 3, 'joachim': 1, 'schroeder': 1, 'tanks': 3, 'rejection': 3, 'younglove': 1, 'occult': 4, 'callum': 3, 'rennie': 3, 'gracen': 3, 'finance': 4, 'discovery': 5, 'missing': 5, 'rent': 3, 'satire': 2, 'melodrama': 23, 'religious': 2, 'cults': 1, 'multiple': 1, 'murders': 4, 'emotional': 14, 'smallwood': 2, 'levitation': 9, 'sheree': 1, 'mediocre': 57, 'bridges': 2, 'novocaine': 1, 'kohnen': 1, 'sorcerer': 3, 'killing': 1, 'bin': 1, 'laden': 1, 'burnette': 1, 'dutcher': 1, 'estes': 5, 'jorma': 1, 'taccone': 1, 'centered': 24, 'tokar': 1, 'available': 18, 'thought': 8, 'cindy': 2, 'lau': 2, 'development': 2, 'vincente': 1, 'minnelli': 3, 'gunfight': 5, 'appignanesi': 1, 'excited': 1, 'stratosphere': 1, '01': 1, 'horner': 1, 'bergin': 3, 'malkovich': 2, 'kopp': 1, 'bring': 1, 'listing': 2, 'farewell': 2, 'callan': 3, 'mulvey': 3, 'pelican': 1, 'hobbs': 1, 'vacation': 3, 'wands': 2, 'wilderness': 7, 'anyone': 4, 'gendreau': 3, 'generally': 2, 'insanity': 3, 'quattrochi': 1, 'timothy': 2, 'dalton': 1, 'lay': 1, 'cia': 4, 'cain': 2, 'sounds': 1, 'sundays': 1, 'tiffanys': 1, 'polanski': 14, 'recall': 2, 'whether': 3, 'derek': 4, 'code': 3, 'tragic': 1, 'form': 7, 'eugene': 1, 'might': 17, 'revolves': 2, 'warrior': 3, 'vulgarity': 2, 'vivica': 2, 'elves': 4, 'given': 25, 'repos': 2, 'paget': 1, 'erased': 3, 'examples': 2, 'manipulation': 2, 'lines': 2, 'megaton': 1, 'earlier': 1, 'delpy': 3, 'adler': 2, 'masterson': 4, 'larue': 1, 'speedman': 3, 'mahiro': 1, 'maeda': 1, 'dating': 2, 'jaeckel': 1, 'kuntz': 1, 'cobb': 3, 'devils': 2, 'moment': 1, 'repossessed': 1, 'officer': 1, 'homeless': 1, 'palmer': 1, 'sorvino': 10, 'tommi': 1, 'lepola': 1, 'juan': 2, 'delancer': 1, 'chaos': 3, 'direction': 2, 'carolina': 1, 'moon': 6, 'able': 9, 'picardo': 4, 'damiano': 1, 'damiani': 1, 'smuggling': 4, 'hyde': 2, 'sidoni': 2, 'describe': 5, 'terror': 6, 'tournament': 1, 'audition': 9, 'kurosawa': 7, 'autry': 5, 'pakula': 1, 'chapple': 2, 'marine': 1, 'averaged': 6, 'andrei': 5, 'tarkovsky': 5, 'andrew': 10, 'cymek': 1, 'jacobs': 1, 'leman': 1, 'bening': 2, 'haywood': 1, 'antoine': 3, 'fuqua': 1, 'marks': 1, 'genius': 1, 'gags': 4, 'benenson': 1, 'peyton': 1, 'contact': 2, 'palma': 9, 'smrz': 2, 'mysterious': 2, 'talkington': 2, 'carlos': 4, 'leon': 5, 'wedge': 1, 'falling': 2, 'category': 2, 'palko': 1, 'mosbacher': 1, 'deidre': 1, 'denis': 1, 'leary': 1, 'jacobi': 3, 'dermot': 3, 'mulroney': 3, 'dolph': 1, 'lundgren': 1, 'duncan': 3, 'mass': 3, 'eli': 1, 'roth': 2, 'federico': 14, 'fellini': 13, 'downey': 10, 'darabont': 10, 'astaire': 4, 'savage': 2, 'homicide': 1, 'hayao': 12, 'miyazaki': 12, 'narrate': 1, 'experience': 5, 'guillermo': 14, 'del': 15, 'toro': 14, 'shumlin': 1, 'opinion': 3, 'weeks': 1, 'couffer': 1, 'government': 4, 'assassin': 3, 'gold': 12, 'garner': 1, 'horan': 1, 'michelangelo': 11, 'antonioni': 11, 'patric': 1, 'lando': 1, 'slowsky': 1, 'wells': 1, 'whitesell': 1, 'lidstrom': 1, 'kenn': 1, 'navarro': 1, 'kirt': 2, 'gunn': 3, 'mario': 2, 'azzopardi': 1, 'investigations': 2, 'marlee': 2, 'matlin': 2, 'suitable': 2, 'rosen': 1, 'melanie': 2, 'biehn': 4, 'caretaker': 1, 'business': 2, 'marnos': 1, 'ziller': 1, 'painter': 4, 'mcgennis': 1, 'phil': 4, 'hartman': 3, 'siegel': 1, 'mcgowan': 2, 'phillipe': 4, 'neill': 4, 'bernhard': 3, 'sara': 1, 'sugarman': 1, 'lately': 9, 'wincer': 1, 'fugitive': 2, 'dorff': 1, 'simmons': 3, 'terrence': 10, 'malick': 9, 'entertain': 1, 'vittoria': 5, 'sica': 5, 'kramer': 1, 'marcy': 1, 'walker': 2, 'open': 4, 'bonnell': 1, 'loane': 1, 'griffiths': 1, 'laugh': 6, 'rodger': 2, 'los': 3, 'piano': 1, 'agree': 1, 'unrequited': 2, 'eden': 2, 'harvey': 3, 'concerning': 1, 'tara': 1, 'judelle': 1, 'behind': 1, 'enemy': 2, 'remote': 1, 'viewings': 1, 'pee': 2, 'wee': 2, 'hironobu': 1, 'sakaguchi': 1, 'dina': 4, 'meyer': 4, 'averaging': 1, 'grauman': 1, 'explorer': 1, 'jonathan': 8, 'glazer': 2, 'agents': 3, 'lili': 5, 'carry': 9, 'megacorporation': 3, 'happen': 8, 'memoirs': 1, 'wags': 1, 'selfishness': 1, 'across': 2, 'davidson': 4, 'assassination': 5, 'yancovic': 2, 'poltergeist': 1, 'prom': 1, 'sonja': 1, 'truckers': 2, 'giraldi': 1, 'underground': 3, 'resistance': 3, 'kris': 2, 'sherwood': 1, 'andre': 1, 'braugher': 1, 'rags': 1, 'meugniot': 1, 'politics': 2, 'nic': 2, 'izzi': 2, 'fischa': 1, 'skeet': 1, 'ulrich': 1, 'albert': 2, 'finney': 2, 'heidi': 1, 'hurt': 2, 'recommendations': 2, 'vaughan': 2, 'vengeance': 4, 'rapaport': 1, 'giancarlo': 1, 'esposito': 1, 'eaten': 6, 'intrusion': 3, 'odonell': 2, 'loretta': 1, 'alper': 1, 'elliot': 1, 'lebovitz': 1, 'doyle': 3, 'emilio': 2, 'blindness': 2, 'notorious': 1, 'scarecrow': 2, 'singles': 1, 'th': 4, 'voyage': 2, 'sinbad': 1, 'trotsky': 1, 'chosen': 2, 'dantes': 2, 'inferno': 2, 'hooker': 1, 'trunk': 1, 't': 10, 'nell': 2, 'shopgirl': 2, 'spotswood': 1, 'hustle': 1, 'toe': 2, 'raymond': 3, 'massey': 2, 'goal': 1, 'prayers': 1, 'staininger': 1, 'kenneth': 3, 'degeneres': 1, 'kunert': 2, 'salisbury': 1, 'cohn': 1, 'randolph': 2, 'mantooth': 2, 'bloody': 2, 'imposter': 2, 'heartache': 1, 'whaling': 1, 'erskine': 1, 'buscemi': 2, 'lisa': 9, 'prinze': 3, 'mistress': 1, 'clements': 1, 'bauchau': 1, 'ninja': 3, 'hopewell': 1, 'sleepwalking': 1, 'brooklyn': 4, 'gabbert': 2, 'shepard': 1, 'gabriel': 5, 'byrne': 2, 'false': 1, 'accusation': 1, 'dougray': 3, 'sight': 10, 'pumpkin': 1, 'karver': 1, 'chamitoff': 1, 'exist': 1, 'akroyd': 3, 'schizophrenia': 2, 'forget': 1, 'dolans': 1, 'eriq': 1, 'lesalle': 1, 'usually': 2, 'excuse': 1, 'connolly': 3, 'inventory': 1, 'famke': 2, 'janssen': 2, 'dealt': 4, 'deportation': 1, 'master': 3, 'kathleen': 2, 'information': 31, 'easy': 3, 'junichi': 1, 'fujisaku': 1, 'malkovic': 1, 'uth': 1, 'satterfield': 1, 'warlock': 2, 'guerrilla': 3, 'warfare': 3, 'atencio': 1, 'spirit': 3, 'dreyfus': 2, 'afghanistan': 1, 'brain': 2, 'smasher': 1, 'duck': 1, 'le': 1, 'samoura': 1, 'shaft': 1, 'static': 1, 'vanessa': 6, 'angel': 3, 'amadeus': 2, 'delirious': 1, 'insanitarium': 1, 'stand': 1, 'bulletproof': 1, 'warm': 1, 'slow': 1, 'moving': 1, 'gregg': 1, 'champion': 1, 'lapica': 1, 'cecil': 1, 'demented': 1, 'selleck': 2, 'reba': 1, 'mcentire': 1, 'moretti': 1, 'rancher': 4, 'amos': 2, 'kollek': 2, 'saulnier': 1, 'helicopter': 2, 'raid': 3, 'hiroyuki': 2, 'kitakubo': 2, 'guiness': 4, 'colm': 3, 'meaney': 2, 'joaquin': 3, 'phoenix': 3, 'theodore': 2, 'witcher': 2, 'pankov': 4, 'mitzi': 1, 'kapture': 1, 'balaban': 2, 'starvation': 1, 'takahiro': 1, 'tanaka': 1, 'mellodrama': 1, 'andrea': 3, 'destruction': 3, 'jaclyn': 1, 'norbit': 1, 'peanuts': 1, 'state': 1, 'property': 1, 'forbidden': 5, 'ocallaghan': 1, 'drake': 3, 'details': 2, 'maniac': 2, 'phoebe': 1, 'eng': 7, 'subs': 7, 'guessing': 1, 'stonerville': 1, 'getting': 4, 'stoned': 1, 'harmony': 1, 'korine': 1, 'elfont': 1, 'kwok': 2, 'alexander': 3, 'rapp': 2, 'britt': 1, 'allcroft': 1, 'cheryl': 1, 'ladd': 1, 'chrispopher': 1, 'corey': 4, 'feldman': 4, 'deforest': 2, 'erik': 2, 'macarthur': 1, 'erika': 3, 'eleniak': 3, 'small': 4, 'hayley': 2, 'mills': 2, 'jackman': 4, 'ian': 7, 'mckellen': 3, 'gedrick': 2, 'kiristine': 2, 'bravery': 5, 'kristy': 1, 'swanson': 1, 'majors': 3, 'lynn': 4, 'shelton': 4, 'hoge': 2, 'melissa': 3, 'gilbert': 2, 'feifer': 1, 'pare': 1, 'phillip': 5, 'recovering': 1, 'alcoholic': 1, 'avary': 1, 'raimi': 3, 'gellar': 2, 'teri': 2, 'polo': 2, 'heroine': 3, 'jokes': 6, 'pearson': 1, 'quality': 1, 'glover': 2, 'anybody': 1, 'yet': 5, 'sobieski': 1, 'parole': 1, 'hearing': 1, 'talisa': 1, 'soto': 1, 'cavalline': 1, 'charisma': 1, 'dacascos': 3, 'criminals': 4, 'breaking': 2, 'rescue': 5, 'nuclear': 2, 'coen': 8, 'abandoned': 4, 'infotainment': 1, 'subject': 2, 'religulous': 1, 'wallace': 1, 'heard': 6, 'meatballs': 2, 'vibes': 1, 'mm': 1, 'upwards': 1, 'swamp': 1, 'flags': 1, 'fathers': 1, 'pete': 1, 'travis': 2, 'costas': 1, 'imprisonment': 4, 'san': 3, 'giacomo': 2, 'vincie': 1, 'heartless': 1, 'barkin': 1, 'karen': 5, 'ridings': 1, 'hurtz': 1, 'due': 3, 'double': 5, 'indemnity': 1, 'documentarys': 3, 'langella': 2, 'gotten': 1, 'revolved': 6, 'campaign': 1, 'may': 10, 'revolver': 1, 'clouds': 1, 'collars': 1, 'burlesque': 1, 'compared': 1, 'shes': 4, 'duty': 1, 'perceived': 1, 'maximum': 1, 'risk': 1, 'hows': 1, 'saech': 1, 'kinear': 1, 'miranda': 1, 'nancy': 4, 'grahn': 2, 'murakami': 1, 'joel': 4, 'lloyd': 6, 'greta': 2, 'scacchi': 1, 'sedative': 2, 'kellie': 2, 'prisoners': 2, 'mine': 1, 'greenwood': 3, 'allan': 1, 'goldstein': 1, 'sands': 1, 'bosses': 1, 'edwards': 6, 'jensen': 2, 'ackles': 2, 'lovers': 2, 'raquel': 1, 'welch': 1, 'making': 1, 'dagenham': 1, 'requesting': 1, 'pam': 2, 'grier': 2, 'seem': 1, 'fishburne': 3, 'outlanders': 1, 'suggestions': 1, 'alienation': 2, 'basket': 1, 'info': 6, 'wicksboro': 1, 'incident': 1, 'moody': 1, 'bummer': 1, 'cavalry': 2, 'kuo': 1, 'ren': 1, 'wu': 2, 'public': 3, 'meins': 1, 'august': 2, 'undergrounds': 2, 'mordum': 1, 'armitage': 2, 'lange': 3, 'loneliness': 2, 'cusak': 2, 'shane': 5, 'hammond': 2, 'niall': 1, 'maccormick': 1, 'georgina': 3, 'riedel': 2, 'dimitry': 1, 'elyashkevich': 1, 'irwin': 1, 'mattson': 1, 'marcil': 2, 'addis': 1, 'wimpenny': 1, 'triumphant': 1, 'fugue': 1, 'key': 2, 'village': 1, 'mccabe': 1, 'miller': 3, 'northern': 1, 'loose': 1, 'spirits': 3, '2081': 1, 'fantastic': 1, 'thirteenth': 1, 'floor': 2, 'brittany': 2, 'snatchers': 1, 'joey': 3, 'mutilation': 2, 'driven': 1, 'masahiko': 1, 'maesawa': 1, 'definitely': 2, 'porretta': 3, 'rie': 1, 'rasmussen': 1, 'wolotzky': 2, 'franklin': 1, 'guerrero': 1, 'avery': 2, 'asner': 1, 'wonderment': 1, 'rowland': 3, 'involves': 7, 'confession': 2, 'forgiven': 1, 'lea': 8, 'salonga': 1, 'phillips': 5, 'galactic': 1, 'kendall': 1, 'prairie': 1, 'companion': 1, 'nature': 3, 'wicked': 1, 'vince': 3, 'vieluf': 1, 'leick': 2, 'masciantonio': 1, 'berkley': 1, 'larry': 6, 'peerce': 1, 'lyde': 1, 'dementia': 1, 'tribe': 1, 'levin': 1, 'kenya': 4, 'montgomery': 5, 'clift': 4, 'ann': 4, 'unconventional': 4, 'appreciate': 1, 'theresa': 4, 'randall': 3, 'gangsters': 2, 'beth': 1, 'evans': 4, 'mcdowell': 3, 'donna': 2, 'derrico': 2, 'karl': 1, 'hirsch': 2, 'sherman': 1, 'curly': 2, 'miner': 5, 'warsaw': 1, 'ghetto': 1, 'dudikoff': 2, 'whiteman': 1, 'mandy': 1, 'patinkin': 1, 'foxes': 1, 'omalley': 1, 'trejo': 1, 'morse': 2, 'dunn': 4, 'malevolence': 1, 'dwells': 1, 'wing': 1, 'chaser': 1, 'supposedly': 1, 'unicorn': 1, 'asking': 2, 'inquiring': 1, 'peterson': 3, 'interested': 3, 'immortality': 1, 'vaughn': 3, 'parallel': 2, 'universes': 1, 'balloons': 3, 'rommel': 1, 'lustig': 2, 'imagination': 2, 'pritts': 1, 'danielle': 2, 'fishel': 2, 'mulgrew': 1, 'catholicism': 1, 'osamu': 1, 'dezaki': 1, 'quaid': 1, 'ubaldo': 1, 'ragona': 1, 'basannavar': 1, 'mackenzie': 2, 'alicia': 2, 'silverstone': 2, 'rel': 1, 'filming': 1, 'style': 2, 'passion': 1, 'career': 1, 'mcclellan': 1, 'garett': 2, 'maggart': 2, 'suffering': 3, 'buzzell': 1, 'notarile': 1, 'setting': 2, 'survival': 4, 'mathew': 3, 'lillard': 3, 'julianna': 2, 'lavin': 2, 'specific': 2, 'treachery': 1, 'skelding': 2, 'wise': 4, 'vino': 1, 'salame': 1, 'joshua': 2, 'carroll': 4, 'oconner': 2, 'priestley': 2, 'ziering': 3, 'friendship': 2, 'alvarez': 1, 'peggy': 1, 'whaley': 3, 'durante': 1, 'mastrantonio': 1, 'moira': 1, 'lander': 1, 'marra': 1, 'casper': 1, 'dien': 1, 'naim': 1, 'sematary': 1, 'galaxy': 2, 'kelber': 1, 'warthog': 2, 'kazuo': 1, 'terada': 1, 'sure': 1, 'wen': 1, 'jiang': 1, 'searching': 12, 'buchanan': 1, 'akhurst': 1, 'bounty': 6, 'hunters': 1, 'mann': 2, 'xavier': 2, 'puslowski': 2, 'dome': 1, 'dogville': 1, 'robinson': 2, 'crusoe': 1, 'sharpes': 1, 'stabbing': 2, 'infantolino': 1, 'colman': 1, 'kalin': 2, 'paradox': 1, 'pilgrim': 1, 'englishman': 1, 'brandis': 1, 'disturbed': 1, 'zacharias': 1, 'proyas': 1, 'survivors': 1, 'paolo': 3, 'montalban': 3, 'wondering': 1, 'karaoke': 2, 'isitt': 1, 'entitled': 6, 'electric': 1, 'mist': 1, 'replace': 1, 'clinton': 3, 'choose': 1, 'abbess': 1, 'wizards': 2, 'moran': 1, 'forbes': 1, 'cook': 2, 'yancy': 1, '1800': 1, 'bologna': 1, 'schwimmer': 3, 'summit': 1, 'dustin': 5, 'diamond': 2, 'devitos': 1, 'downtown': 2, 'adrian': 2, 'vitoria': 1, 'steiman': 1, 'pena': 1, 'barbie': 2, 'fabrice': 2, 'du': 2, 'welz': 1, 'schrader': 1, 'leonetti': 1, 'stevan': 1, 'mena': 1, 'deals': 4, 'infantry': 1, 'vohrer': 1, 'deception': 1, 'zaphiratos': 1, 'karbelnikoff': 1, 'brett': 2, 'ricardo': 2, 'montalbon': 2, 'beghe': 3, 'gavin': 1, 'budd': 1, 'emmett': 2, 'alston': 2, 'collins': 2, 'lynda': 2, 'jail': 4, 'betrayal': 2, 'ernest': 1, 'shari': 1, 'lies': 2, 'marciano': 1, 'jada': 3, 'pinkett': 3, 'sollett': 1, 'nazi': 7, 'occupation': 1, 'activism': 4, 'isabella': 3, 'rosselini': 3, 'bryan': 1, 'sivertson': 1, 'hindman': 2, 'spacecrafts': 2, 'marcus': 3, 'raboy': 1, 'memorial': 1, 'ransom': 3, 'kearsley': 1, 'steyermark': 1, 'yasmine': 2, 'bleeth': 2, 'bayne': 1, '18': 2, 'century': 2, 'wilkins': 1, 'caruso': 1, 'tatnya': 1, 'ali': 1, 'walters': 2, 'optimism': 1, 'anarchy': 2, 'brickman': 1, 'chong': 3, 'paddy': 1, 'breathnach': 1, 'parasites': 4, 'awarded': 1, 'tori': 1, 'spelling': 1, 'cory': 1, 'brenden': 1, 'sexton': 1, 'pantiliano': 1, 'stephenson': 1, 'lara': 3, 'fabian': 1, 'joven': 1, 'tan': 1, 'cynthia': 1, 'occupied': 2, 'poland': 2, 'lederman': 1, 'diana': 2, 'riggs': 1, 'illness': 1, 'zane': 2, 'planes': 1, 'water': 2, 'cert': 1, 'mayor': 1, 'these': 2, 'drop': 1, 'gorgeous': 1, 'earthlings': 1, 'edwin': 1, 'mccain': 1, 'grandmas': 1, 'jealousy': 5, 'hoodrats': 1, 'hoodrat': 1, 'slated': 1, 'platoon': 5, 'sorority': 3, 'row': 2, 'nicholsons': 1, 'anywhere': 2, 'rosalie': 1, 'bunch': 2, 'fastest': 1, 'indian': 1, 'obvious': 1, 'marsters': 2, 'morey': 2, 'darjeeling': 1, 'limited': 1, 'necropolis': 1, 'mckenzie': 2, 'astin': 3, 'rydell': 1, 'michell': 2, 'entertainment': 3, 'thir': 1, 'en': 1, 'adults': 1, 'rip': 1, 'circuit': 1, 'remaking': 1, 'gathering': 1, 'eagles': 1, 'check': 1, 'altitude': 1, 'donnell': 1, 'rawlings': 1, 'ashy': 1, 'classy': 1, 'toons': 1, 'hells': 1, 'kitchen': 1, 'television': 1, 'ring': 2, 'shaun': 1, 'sheep': 1, 'leap': 1, 'lambkind': 1, 'siren': 1, 'slaughtered': 1, 'vomit': 1, 'dolls': 2, 'museum': 1, 'even': 1, 'alden': 1, 'redgrave': 1, 'tracie': 4, 'lords': 4, 'somehow': 2, 'ginsburg': 2, 'hooks': 2, 'widen': 1, 'mccaulay': 3, 'tales': 1, 'execution': 6, 'jordanna': 3, 'brewster': 3, 'foley': 1, 'dylan': 2, 'neal': 2, 'flintstones': 1, 'honor': 4, 'mayfield': 1, 'gina': 2, 'gershon': 2, 'heather': 2, 'graham': 1, 'ruzickova': 1, 'bonamy': 1, 'jaye': 2, 'erick': 1, 'dowdle': 1, 'taking': 2, 'stevenson': 2, 'morio': 1, 'asaka': 1, 'invisibility': 3, 'sebastian': 2, 'panneck': 2, 'amitri': 1, 'espejo': 1, 'hartle': 1, 'ned': 1, 'farr': 1, 'market': 3, 'gluck': 1, 'daniela': 2, 'pestova': 2, 'jag': 1, 'mundhra': 1, 'nikki': 2, 'cox': 3, 'vonda': 3, 'shepherd': 4, 'rosson': 1, 'scoggins': 3, 'soderbergh': 6, 'sage': 1, 'date': 3, 'duchovny': 2, 'illegal': 1, 'activity': 1, 'drenner': 1, 'patient': 1, 'disguised': 1, 'mira': 7, 'furlan': 1, 'quod': 2, 'drugs': 2, 'sexuality': 4, 'dandridge': 1, 'contend': 1, 'revolving': 7, 'stan': 4, 'kirsch': 4, 'beltran': 1, 'sticks': 1, 'starrcade': 1, 'ultimate': 1, 'mirrormask': 1, 'preston': 1, 'cruelty': 1, 'fine': 2, 'decision': 2, 'gunpoint': 1, 'focuses': 2, 'torture': 6, 'detectives': 3, 'explores': 1, 'focusing': 2, 'neel': 1, 'tetsuro': 1, 'amino': 1, 'teens': 3, 'societal': 2, 'girly': 1, 'hardship': 1, 'everlasting': 3, 'base': 2, 'bandits': 3, 'hilarious': 1, 'couple': 1, 'twisters': 1, 'snipers': 3, 'neretva': 1, 'shack': 1, 'joneses': 1, '976': 1, 'bangkok': 1, 'dr': 10, 'caparulo': 1, 'cap': 1, 'malibu': 1, 'motocrossed': 1, 'paper': 1, 'woodstock': 1, 'tenten': 1, 'diemens': 1, 'cannabis': 1, 'cured': 1, 'irving': 3, 'lerner': 1, 'revenge': 5, 'tx': 1, 'ado': 2, 'kalangis': 1, 'church': 1, 'gorney': 3, 'baron': 2, 'munchausen': 1, 'sunset': 2, 'briefcase': 1, 'oregon': 1, 'jennie': 1, 'garth': 2, 'amateurs': 1, 'silver': 1, 'haim': 3, 'gero': 1, 'bancroft': 2, 'investigation': 3, 'gosselar': 2, 'roxann': 2, 'browne': 1, 'unhappy': 2, 'sexual': 2, 'fischer': 1, 'content': 1, 'kentis': 1, 'harvest': 3, 'waitt': 1, 'mathilde': 1, 'bittner': 1, 'self': 3, 'saffa': 1, 'fever': 3, 'head': 9, 'plan': 1, 'anticipated': 1, 'jeri': 3, 'sweetheart': 1, 'gasaway': 1, 'cattle': 1, 'marc': 2, 'rocco': 2, 'biel': 2, 'townsend': 3, 'criminal': 1, 'mastermind': 1, 'sherry': 1, 'stringfield': 1, 'dual': 2, 'identities': 1, 'boaz': 1, 'yakin': 1, 'appeal': 1, 'jeffery': 1, 'friedman': 1, 'fickman': 1, 'maddocks': 1, 'freiburger': 1, 'surgery': 1, 'shalmar': 1, 'joke': 3, 'kimberly': 1, 'mcculough': 1, 'merit': 1, 'badge': 1, 'ari': 1, 'taub': 1, 'robins': 1, 'shakes': 1, 'jennings': 1, 'shoehorned': 1, 'subplot': 1, 'hitchhikers': 1, 'guide': 1, 'toys': 3, 'chance': 1, 'dissociative': 1, 'disorder': 1, 'jfk': 2, 'gomez': 1, 'bleckner': 1, 'fighter': 1, 'jarmusch': 1, 'krista': 2, 'morrit': 2, 'fuest': 1, 'julian': 3, 'kemp': 1, 'files': 1, 'hoover': 1, 'levar': 3, 'blackmail': 1, 'donuts': 2, 'carrot': 1, 'keri': 1, 'maid': 1, 'cole': 1, 'scar': 2, 'alison': 1, 'raimund': 1, 'huber': 1, 'shunji': 1, 'iwai': 1, 'transylvania': 1, 'micheal': 1, 'lester': 2, 'calista': 1, 'flockhart': 1, 'mccrudden': 1, 'haggerty': 4, 'lucci': 1, 'turteltaub': 1, 'lozano': 1, 'powell': 2, 'whitworth': 2, 'jan': 4, 'rachman': 1, 'crimes': 4, 'velcrow': 1, 'ripper': 1, 'garrett': 2, 'wang': 2, 'johan': 1, 'grimonprez': 1, 'peopled': 1, 'laughed': 1, 'gian': 2, 'keth': 1, 'szarabajka': 1, 'wolochatiuk': 1, 'huntington': 2, 'seiji': 1, 'chiba': 1, 'stolz': 3, 'comical': 1, 'nibbelink': 1, 'technology': 1, 'delaney': 2, 'manger': 1, 'aidan': 4, 'quinn': 4, 'maurice': 4, 'chair': 1, 'mellisa': 3, 'hart': 3, 'carrere': 2, 'polson': 1, 'josetxo': 1, 'mateo': 1, 'bell': 3, 'ferland': 1, 'bassett': 3, 'dorn': 2, 'storage': 1, 'felicias': 1, 'fusion': 1, 'trip': 2, 'fro': 1, 'tanya': 4, 'cyborg': 1, 'ileana': 2, 'winkler': 2, 'tiffani': 3, 'thiessen': 3, 'byington': 1, 'loss': 1, 'mitchum': 1, 'takashi': 1, 'ishii': 1, 'reconnaissance': 1, 'cafiero': 1, 'broken': 2, 'engagement': 1, 'hagman': 1, 'sugg': 1, 'rowlands': 1, 'herbert': 2, 'coleman': 1, 'gerard': 2, 'depardieu': 2, 'pushed': 1, 'lew': 1, 'attempted': 2, 'milestone': 1, 'relating': 1, 'gino': 1, 'nichele': 1, 'natascha': 1, 'mcelhone': 1, 'jeffrey': 1, 'katzenberg': 1, 'kathie': 1, 'gifford': 1, 'kriv': 1, 'stenders': 1, 'burgi': 1, 'midlife': 1, 'crisis': 1, 'geno': 1, 'mcgahee': 1, 'zach': 1, 'hofmeyr': 1, 'brock': 3, 'basil': 1, 'cloke': 1, 'minutes': 1, 'graffiti': 1, 'tennis': 1, 'pursuit': 2, 'sortie': 1, 'des': 1, 'ateliers': 1, 'vibert': 1, 'bride': 3, 'toms': 1, 'midnight': 1, 'garden': 1, 'valmont': 1, 'melman': 1, 'shanley': 1, 'strike': 2, 'schlesinger': 1, 'lambert': 4, 'lon': 1, 'chaney': 1, 'attempt': 3, 'orphans': 1, 'wrye': 2, 'fernando': 1, 'colunga': 1, 'sher': 1, 'anna': 2, 'galvin': 1, 'einstein': 2, 'incorporate': 1, 'shapeshifting': 3, 'ryans': 1, 'rupert': 3, 'everet': 3, 'poet': 2, 'irish': 2, 'friedkin': 1, 'hamburg': 1, 'winger': 2, 'lebrock': 2, 'margret': 1, 'places': 3, 'basic': 1, 'instinct': 1, 'strong': 1, 'gornick': 2, 'melodramas': 2, 'several': 12, 'jeffs': 1, 'kieran': 1, 'carney': 2, 'twenty': 1, 'reprieve': 1, 'pale': 1, 'gates': 3, 'mcfadden': 2, 'central': 2, 'believers': 1, 'dangerous': 2, 'saturn': 1, 'territories': 1, 'cooler': 1, 'rimshop': 1, 'delbert': 1, 'lawless': 1, 'listed': 1, 'dishonor': 2, 'receiving': 2, 'kazuaki': 1, 'kiriya': 1, 'skill': 1, 'stevens': 2, 'infidelity': 1, 'fraunces': 1, 'premise': 2, 'exciting': 1, 'maniacts': 1, 'chuang': 1, 'mellinda': 2, 'mention': 1, 'haussman': 1, 'michel': 2, 'orion': 1, 'suede': 1, 'ide': 1, 'shootout': 1, 'brokedown': 1, 'palace': 1, 'fiddler': 1, 'roof': 1, 'sarahs': 1, 'choice': 1, 'maria': 3, 'bello': 4, 'favorites': 1, 'pinson': 3, 'cukor': 1, 'woodward': 1, 'debuted': 1, 'frankel': 1, 'osteen': 1, 'lachman': 1, 'heroes': 2, 'crash': 2, 'polito': 1, 'addresses': 1, 'consequences': 2, 'deforestation': 1, 'skirmishes': 1, 'border': 2, 'reiser': 2, 'meehl': 1, 'womens': 1, 'der': 1, 'beek': 1, 'ultimately': 1, 'ikea': 1, 'rise': 1, 'byrd': 1, 'mcdonald': 1, 'archer': 1, 'luner': 2, 'disfigured': 1, 'supper': 1, 'doomsday': 1, 'device': 1, 'directory': 1, 'bam': 1, 'margera': 1, 'global': 3, 'climate': 1, 'issue': 1, 'boase': 1, 'madsen': 2, 'gains': 1, 'guilietta': 1, 'masina': 1, 'haley': 2, 'flatland': 1, 'okay': 1, 'cky': 1, 'k': 1, 'enjoyed': 1, 'retirement': 3, 'sullivan': 1, 'kirby': 1, 'dick': 1, 'example': 1, 'security': 2, 'kell': 1, 'evan': 1, 'detten': 1, 'willard': 2, 'ivan': 2, 'mitov': 1, 'melinda': 1, 'clarke': 1, 'pregnancy': 2, 'jo': 2, 'chandler': 2, 'wilcox': 1, 'ric': 1, 'monte': 1, 'snatching': 1, 'salley': 1, 'dennehy': 3, 'kari': 2, 'wuhrer': 2, 'mccormack': 1, 'marcos': 1, 'efron': 1, 'fina': 1, 'torres': 1, 'duhame': 1, 'jac': 1, 'schaeffer': 1, 'kinney': 1, 'loverboy': 1, 'mnemonic': 1, 'soapdish': 1, 'bradley': 1, 'stockholm': 1, 'syndrome': 1, 'summarize': 2, 'serpent': 1, 'keegan': 1, 'raising': 1, 'arizona': 1, 'panorama': 1, 'hardware': 1, 'cover': 1, 'zerophilia': 1, 'acceptance': 1, 'exit': 1, 'milla': 1, 'jovovich': 1, 'rafelson': 1, 'oh': 1, 'thunderbolt': 1, 'lightfoot': 1, 'sweatshop': 1, 'elektra': 1, 'actions': 1, 'phedon': 1, 'papamichael': 1, 'cess': 1, 'silvera': 1, 'terrorist': 3, 'travelers': 1, 'boondock': 1, 'saints': 1, 'corky': 1, 'romano': 1, 'bunraku': 1, 'jay': 3, 'duplass': 1, 'calvert': 1, 'rigg': 1, 'hybrid': 1, 'unsettling': 1, 'alaimo': 1, 'sackheim': 2, 'cheetah': 1, 'gonzalez': 1, 'paxton': 1, 'tatopoulos': 1, 'killed': 5, 'laughs': 1, 'fugitives': 2, 'lamont': 1, 'starman': 1, 'offered': 7, 'such': 3, 'damato': 2, 'promises': 2, 'assayas': 1, 'shou': 1, 'wanted': 2, 'wanting': 4, 'lasted': 1, 'martyn': 1, 'pick': 1, 'alberta': 1, 'company': 2, 'cassandra': 2, 'alyssa': 2, 'moy': 1, 'job': 5, 'claudia': 3, 'debrah': 1, 'farentino': 1, 'donnie': 1, 'walberg': 1, 'ridgemont': 1, 'geena': 1, 'geri': 2, 'halliwell': 2, 'poole': 1, 'chen': 1, 'madeline': 1, 'stowe': 1, 'follows': 1, 'shanks': 3, 'valen': 1, 'boer': 1, 'instrumental': 1, 'appearing': 1, 'stacey': 1, 'ugly': 1, 'guard': 1, 'fridel': 1, 'oconnor': 4, 'sitch': 2, 'cambodia': 2, 'blaine': 1, 'patton': 2, 'budreau': 1, 'pistol': 1, 'whip': 1, 'leah': 1, 'sturgis': 1, 'lou': 2, 'sims': 1, 'friz': 1, 'freleng': 1, 'outcast': 1, 'food': 2, 'approved': 1, 'meerkat': 1, 'ri': 1, 'chard': 1, 'though': 1, 'kinji': 1, 'fukasaku': 1, 'morrison': 1, 'dominique': 1, 'milano': 2, 'parise': 1, 'corporate': 2, 'ernst': 1, 'gossner': 1, 'doran': 1, 'dale': 3, 'steffanino': 1, 'barnz': 1, 'doucette': 1, 'largely': 1, 'ulmer': 1, 'florian': 1, 'henckel': 1, 'donnersmarck': 1, 'averages': 1, 'murdered': 1, 'marek': 1, 'losey': 1, 'solitary': 2, 'confinement': 2, 'clancy': 1, 'roos': 1, 'germany': 1, 'alphonso': 2, 'tze': 1, 'chun': 1, 'deemed': 3, 'larabe': 2, 'kolton': 1, 'focused': 3, 'kilner': 1, 'norda': 1, 'aronoff': 1, 'seller': 1, 'nia': 1, 'vardalos': 1, 'technically': 1, 'prendergast': 1, 'mazin': 1, 'holechek': 1, 'anjelica': 1, 'russ': 1, 'todd': 4, 'verow': 1, 'alcohol': 4, 'fairuza': 1, 'balk': 1, 'kristine': 1, 'mendes': 1, 'huck': 1, 'botko': 1, 'jalmari': 1, 'helander': 1, 'kevan': 1, 'rage': 2, 'kostas': 1, 'karagiannis': 1, 'genocide': 1, 'nair': 1, 'gurland': 1, 'mohr': 1, 'lance': 3, 'weiler': 1, 'pepper': 2, 'argott': 1, 'standoff': 3, 'toni': 1, 'harman': 1, 'liza': 2, 'int': 2, 'decerchio': 1, 'anita': 1, 'laselva': 1, 'vehicle': 1, 'demme': 1, 'psychopath': 2, 'northam': 1, 'isacsson': 1, 'pays': 1, 'melski': 1, 'accidental': 1, 'alberto': 1, 'cavalcanti': 1, 'embezzlement': 2, 'menell': 1, 'hypocrisy': 1, 'deblois': 1, 'reflection': 1, 'levien': 1, 'pascal': 1, 'franchot': 1, 'mcconaughey': 2, 'fiennes': 1, 'smits': 1, 'patty': 2, 'jenkins': 3, 'scribner': 1, 'revolution': 1, 'bangalter': 1, 'rowell': 1, 'warner': 1, 'spiner': 1, 'till': 2, 'barker': 1, 'axelgaard': 1, 'teddy': 4, 'tylo': 1, 'christensen': 1, 'oliveira': 1, 'ruben': 1, 'preuss': 1, 'brownrigg': 1, 'ritter': 1, 'betz': 1, 'solunga': 2, 'despair': 1, 'sangiuliano': 1, 'natasha': 3, 'pavlovich': 2, 'benard': 3, 'annakin': 1, 'aldas': 1, 'reiners': 1, 'myles': 2, 'fergusons': 2, 'mccarthys': 1, 'biographys': 1, 'thora': 1, 'birch': 2, 'choices': 1, 'hank': 1, 'braxtan': 1, 'jerami': 1, 'asquith': 1, 'gariazzo': 1, 'shyu': 1, 'greenfield': 1, 'whale': 1, 'facts': 1, 'malloy': 1, 'lissa': 1, 'rinna': 1, 'thief': 2, 'henstridge': 1, 'wainwright': 1, 'gabby': 1, 'peters': 1, 'tuukka': 1, 'tiensuu': 1, 'matthau': 1, 'chu': 1, 'amusing': 1, 'annoying': 1, 'roommate': 1, 'simple': 1, 'platt': 1, 'cybill': 1, 'tea': 2, 'leoni': 2, 'afterlife': 1, 'portillo': 2, 'speers': 1, 'binder': 1, 'kent': 1, 'jared': 2, 'leto': 2, 'stern': 2, 'wolfinger': 1, 'sadofsky': 1, 'axel': 1, 'rebecca': 4, 'cammisa': 1, 'sophia': 1, 'loren': 1, 'river': 4, 'doremus': 2, 'rosanna': 2, 'munro': 1, 'blutman': 1, 'desi': 1, 'arnaz': 1, 'cheech': 1, 'marin': 1, 'engstrom': 1, 'tomas': 1, 'sandquist': 1, 'baird': 1, 'wiley': 1, 'fishman': 1, 'classified': 1, 'dover': 1, 'koshashvili': 1, 'henriksen': 1, 'kiersch': 1, 'humber': 1, 'balderstone': 1, 'placid': 1, 'monkey': 1, 'presidents': 1, 'sleepover': 1, 'freebie': 1, 'capote': 1, 'elf': 1, 'hustler': 1, 'pasdar': 1, 'denk': 1, 'motor': 1, 'hurley': 1, 'rentzel': 1, 'thueson': 1, 'squeakquel': 1, 'boogeyman': 1, 'turin': 1, 'webber': 1, 'orgy': 1, 'becoming': 1, 'darklands': 1, 'def': 1, 'diggers': 1, 'dysfunktional': 2, 'electroma': 1, 'gung': 1, 'ho': 1, 'carlsons': 1, 'makin': 1, 'corpses': 1, 'confusing': 1, 'internal': 1, 'jakes': 1, 'corner': 1, 'luster': 1, 'brandos': 1, 'bones': 2, 'blueberry': 1, 'lai': 2, 'prefontaine': 1, 'niros': 1, 'satans': 2, 'helper': 2, 'simpatico': 1, 'figures': 1, 'slingshot': 1, 'someones': 1, 'knocking': 1, 'splintered': 1, 'hawks': 1, 'hounds': 1, 'adele': 1, 'sec': 1, 'grey': 1, 'hire': 1, 'ambush': 1, 'others': 2, 'signal': 1, 'trading': 1, 'trick': 1, 'twin': 1, 'peaks': 1, 'unrivaled': 1, 'wisegal': 1, 'zenon': 1, 'zequel': 1, 'molina': 1, 'maggi': 1, 'sykes': 1, 'tiffany': 1, 'kilbourne': 1, 'raged': 1, 'katsuhiro': 1, 'ohtomo': 1, 'gibney': 1, 'duffy': 1, 'freudenthal': 1, 'concerns': 1, 'barnick': 1, 'ving': 1, 'rhames': 1, 'monika': 1, 'treut': 1, 'injustice': 1, 'crowder': 1, 'february': 1, 'christophe': 1, 'gans': 1, 'conway': 1, 'jessy': 1, 'terrero': 1, 'organized': 2, 'troupe': 1, 'description': 1, 'carvey': 1, 'scolari': 1, 'abilities': 1, 'lien': 1, 'bobbie': 1, 'ellis': 1, 'brutality': 1, 'gayheart': 2, 'boogie': 1, 'woogie': 1, 'laughing': 1, 'twisted': 2, 'tasers': 1, 'isaacs': 1, 'lorentzon': 1, 'darlene': 1, 'vogel': 1, 'rene': 3, 'russo': 4, 'vaungh': 1, 'heady': 2, 'anat': 1, 'seftel': 1, 'polish': 1, 'williamson': 1, 'abuse': 1, 'felitta': 1, 'callahan': 3, 'mining': 3, 'minkoff': 1, 'wincat': 1, 'alcala': 1, 'bellware': 1, 'lieutenant': 1, 'chelsom': 1, 'dom': 1, 'rotheroe': 1, 'champions': 1, 'gale': 1, 'harding': 1, 'regarding': 2, 'hacker': 1, 'mikey': 1, 'hilb': 1, 'strock': 1, 'posey': 1, 'ferguson': 1, 'altieri': 1, 'langer': 1, 'burny': 1, 'mattinson': 1, 'delta': 2, 'burke': 3, 'rowan': 1, 'bednarski': 1, 'stephens': 1, 'cassidy': 3, 'jewish': 2, 'paudge': 1, 'behan': 1, 'yudis': 2, 'jorge': 1, 'solis': 1, 'sweeney': 1, 'sorbo': 1, 'foot': 1, 'wayons': 2, 'stellan': 2, 'olsson': 2, 'cerasolis': 1, 'damian': 1, 'roland': 1, 'emmerich': 1, 'bo': 1, 'zenga': 1, 'dogma': 1, 'headless': 1, 'horseman': 1, 'darwell': 1, 'longis': 1, 'homo': 1, 'erectus': 1, 'rollercoaster': 1, 'mouseketeers': 1, 'breakin': 1, 'everett': 1, 'harmon': 1, 'overton': 1, 'courtship': 1, 'eddies': 1, 'basis': 1, 'believer': 1, 'adventurer': 1, 'narrated': 1, 'monkeybone': 1, 'edge': 1, 'annapolis': 1, 'blackboard': 1, 'teds': 1, 'bullet': 1, 'cherrybomb': 1, 'americas': 1, 'cyber': 1, 'threat': 1, 'divine': 1, 'ya': 2, 'sisterhood': 1, 'dumping': 2, 'gekijouban': 1, 'stay': 1, 'unlimited': 1, 'works': 1, 'gymkata': 1, 'normal': 1, 'newsmakers': 1, 'majestys': 1, 'service': 1, 'silicon': 1, 'valley': 3, 'rancid': 1, 'rockaway': 1, 'sanctimony': 1, 'taboo': 1, 'botany': 1, 'confidant': 1, 'tremors': 1, 'triangle': 2, 'waking': 1, 'wilde': 1, 'marina': 1, 'gavrilova': 1, 'exotic': 1, 'mask': 1, 'trikonis': 1, 'cream': 1, 'coscarellis': 1, 'tapping': 2, 'bairstow': 1, 'garfield': 1, 'gabrielle': 1, 'anwar': 1, 'barely': 1, 'preparatory': 1, 'goldblum': 2, 'rater': 1, 'loop': 1, 'ally': 1, 'beverley': 1, 'mitchell': 2, 'schatzberg': 1, 'ingo': 1, 'rademacher': 1, 'darrell': 1, 'mapson': 1, 'jumbo': 1, 'roadkill': 1, 'bonifacio': 1, 'frakes': 1, 'cocktail': 1, 'assemblage': 1, 'sphere': 1, 'koehler': 1, 'zuccon': 1, 'quadrophenia': 1, 'tides': 4, 'fit': 1, 'uninvited': 1, 'unloved': 1, 'camping': 1, 'dusk': 1, 'destiny': 1, 'hedge': 1, 'spinout': 1, 'roscoe': 1, 'penance': 1, 'metamorphosis': 1, 'tigerland': 1, 'bongwater': 1, 'cow': 1, 'target': 2, 'patti': 1, 'labelle': 1, 'sargent': 1, 'kopple': 1, 'segal': 2, 'resnikoff': 1, 'abell': 1, 'storke': 1, 'tells': 2, 'locklear': 1, 'fairy': 1, 'far': 1, 'elliotte': 1, 'bankruptcy': 1, 'hamid': 1, 'gurkha': 1, 'grief': 1, 'wim': 1, 'wenders': 1, 'dannelly': 1, 'rory': 1, 'mchenry': 1, 'bangs': 1, 'grossman': 1, 'heisler': 1, 'egleson': 1, 'mandelbaum': 1, 'fields': 1, 'zieff': 1, 'demoniacs': 1, 'tile': 1, 'simcha': 1, 'jacobovici': 1, 'create': 2, 'fellinis': 1, 'zeitgeist': 1, 'griffin': 1, 'reflections': 1, 'groomsmen': 1, 'radiation': 1, 'fated': 2, 'hurst': 1, 'judging': 1, 'digital': 1, 'evolution': 2, 'houchins': 1, 'witchcraft': 1, 'renoire': 1, 'settings': 1, 'spader': 1, 'naked': 2, 'lighter': 1, 'lower': 2, 'fahey': 1, 'salli': 1, 'reputations': 1, 'viewing': 1, 'jann': 1, 'butch': 1, 'saul': 1, 'dibb': 1, 'lockheart': 1, 'switzer': 1, 'mccallum': 1, 'gondry': 1, 'hartfield': 1, 'qualifies': 7, 'pitof': 1, 'fairchild': 1, 'stouffer': 1, 'dani': 1, 'menkin': 1, 'colleges': 1, 'bouzaglo': 1, 'sheridan': 1, 'hollis': 1, 'chamberlain': 1, 'lacey': 1, 'chabert': 1, 'thaw': 1, 'brady': 1, 'barrett': 1, 'iren': 1, 'koster': 1, 'pytka': 1, 'ciarn': 1, 'hinds': 1, 'shorts': 1, 'ipson': 1, 'donal': 1, 'mosher': 1, 'batchelor': 1, 'courtois': 1, 'galland': 1, 'brings': 1, 'forth': 1, 'totally': 1, 'paint': 1, 'wagon': 1, 'leitch': 1, 'colleen': 1, 'dewhurst': 1, 'rosenberg': 1, 'kosick': 1, 'tillman': 1, 'valentine': 1, 'pelka': 1, 'coneybeare': 1, 'hillcoat': 1, 'carl': 1, '964': 2, 'pinocchio': 2, 'neighbours': 1, 'cages': 1, 'rudo': 1, 'y': 1, 'cursi': 1, 'shrink': 2, 'musician': 1, 'unwed': 1, 'gunslinger': 1, 'alderton': 1, 'chih': 1, 'leong': 1, 'hallucination': 1, 'signorelli': 1, 'lowe': 1, 'cypher': 1, 'cipolla': 1, 'cicely': 1, 'dammes': 1, 'atwell': 1, 'passenger': 1, '57': 1, 'wad': 1, 'coal': 1, 'rufus': 1, 'sewell': 1, 'rosas': 1, 'terence': 1, 'daw': 1, 'crasher': 1, 'bend': 1, 'cheyenne': 1, 'cruel': 1, 'necessary': 1, 'creation': 1, 'dressed': 1, 'straight': 1, 'magoo': 1, 'magnificent': 2, 'sleep': 2, 'drunks': 1, 'boxleitner': 1, 'whatever': 1, 'wire': 1, 'brotherhood': 1, 'tapes': 1, 'grimble': 1, 'bender': 1, 'enjoyable': 1, 'trandem': 1, 'mate': 1, 'gullivers': 2, 'travels': 2, 'weights': 1, 'kinsey': 1, 'joffe': 1, 'necromentia': 1, 'overnight': 1, 'delivery': 1, 'pounds': 1, 'subtle': 1, 'seduction': 1, 'saved': 1, 'pain': 1, 'proudly': 1, 'hail': 2, 'whaledreamers': 1, 'chucky': 1, 'fruit': 1, 'metropia': 1, 'judgment': 2, 'walls': 1, 'cinema': 2, 'afraid': 1, 'senior': 1, 'gamers': 1, 'dorkness': 1, 'rising': 2, 'billion': 1, 'scooby': 2, 'doo': 3, 'abracadabra': 1, 'heredity': 1, 'valdemar': 1, 'bittersweet': 1, 'backyards': 1, 'fierlinger': 1, 'vadim': 1, 'sheng': 1, 'ding': 1, 'gummo': 1, 'grizzly': 1, 'loch': 1, 'ness': 1, 'lawn': 1, 'principal': 1, 'purchase': 5, 'online': 2, 'sandor': 1, 'rank': 2, 'among': 2, 'nicola': 1, 'bounce': 1, 'showing': 1, 'sunday': 1, 'dagger': 1, 'okuribito': 1, 'ran': 1, 'rear': 2, 'graveyard': 1, 'shift': 1, 'hours': 1, 'pepin': 1, 'markowitz': 1, 'storytelling': 1, 'lovelace': 1, 'conrad': 1, 'rooks': 1, 'centering': 1, 'hutchison': 1, 'palance': 1, 'cutler': 1, 'stacy': 1, 'zinn': 1, 'bakshi': 1, 'dealers': 1, 'fleming': 1, 'aquarium': 1, 'sabato': 1, 'berkelys': 1, 'deniros': 1, 'mcnee': 1, 'joness': 1, 'diary': 1, 'wolf': 1, 'assassins': 1, 'merry': 1, 'bridge': 2, 'camouflaged': 1, 'blackie': 1, 'wheel': 1, 'candyman': 1, 'cleo': 1, 'pimpernel': 1, 'cello': 1, 'aykroyd': 1, 'unplugged': 1, 'ufos': 1, 'bone': 2, 'hat': 1, 'chamber': 1, 'hey': 2, 'esther': 1, 'blueburger': 1, 'player': 1, 'abby': 1, 'kick': 1, 'ass': 1, 'laputa': 1, 'lets': 2, 'lornas': 1, 'neshoba': 1, 'northfork': 1, 'orange': 1, 'county': 1, 'perestroika': 1, 'primal': 1, 'cut': 1, 'dirt': 1, 'seraphim': 1, 'falls': 1, 'silkwood': 1, 'slam': 1, 'bang': 1, 'splatter': 1, 'disco': 1, 'blvd': 1, 'sweet': 1, 'sweetbacks': 1, 'baadasssss': 1, 'bells': 1, 'ipcress': 1, 'file': 1, '1900': 1, 'knew': 1, 'bees': 1, 'sweetest': 1, 'telling': 1, 'unknown': 1, 'wanderers': 1, 'wicker': 1, 'thunderpants': 1, 'mini': 1, 'midkiff': 1, 'environmentalism': 1, 'gallipoli': 1, 'nineteen': 1, 'eighty': 1, 'bricktown': 1, 'flypaper': 1, 'longshots': 1, 'wolves': 1, 'buck': 1, 'expedition': 1, 'invaders': 1, 'tortured': 1, 'irishman': 1, 'zoom': 1, 'cu': 1, 'mama': 1, 'rugrats': 1, 'sitting': 1, 'waterfall': 1, 'almost': 1, 'meth': 1, 'legion': 1, 'macbeth': 1, 'masters': 1, 'universe': 1, 'terrifying': 1, 'girlfriends': 1, 'lunch': 1, 'babes': 1, 'thon': 1, 'tenshi': 1, 'tamago': 1, 'lovely': 2, 'winters': 1, 'curfew': 1, 'disclosure': 1, 'salvation': 1, 'chips': 1, 'lifetime': 1, 'cosmos': 1, 'ace': 1, 'hole': 1, 'candle': 1, 'wine': 1, 'cemetery': 1, 'junction': 1, 'chains': 1, 'worry': 1, 'italian': 1, 'fubar': 1, 'gargoyles': 1, 'conquering': 1, 'changi': 1, 'keeping': 1, 'mum': 1, 'alphabet': 1, 'charge': 1, 'feather': 1, 'souled': 1, 'watermelon': 1, 'chicken': 1, 'paranormal': 1, 'penomena': 1, 'relic': 1, 'matilda': 1, 'miles': 1, 'profondo': 1, 'rosso': 1, 'smash': 1, 'camera': 1, 'grinch': 1, 'batteries': 1, 'wives': 1, 'lefay': 1, 'crossover': 1, 'wax': 1, 'passage': 1, 'marseille': 1, 'jawbreaker': 1, 'highway': 1, 'runnery': 1, 'eagle': 1, 'easel': 1, 'robocop': 1, 'gear': 1, 'solid': 1, 'orchid': 1, 'vicious': 1, 'priest': 1, 'cedric': 2, 'gibbons': 1, 'apartment': 1, 'sorted': 1, 'hutton': 1, 'hardwicke': 1, 'traci': 1, 'bingham': 1, 'movei': 1, 'movement': 2, 'fictional': 1, 'espionage': 1, 'fuminori': 1, 'kizaki': 1, 'trevor': 1, 'saloon': 1, 'thrown': 1, 'situations': 1, 'crossed': 1, 'axe': 1, 'horde': 1, 'withing': 1, 'geraint': 1, 'wyne': 1, 'davies': 1, 'website': 5, 'detailed': 2, 'opinions': 1, 'common': 1, 'already': 1, 'users': 1, 'evita': 5, 'comments': 1, 'boring': 1, 'sentiment': 1, 'site': 1, 'bit': 1, 'awesome': 1, 'glimpse': 1, 'clip': 1, 'extended': 1, 'highlights': 1, 'spoilers': 2, 'snippets': 1, 'cinemas': 1, '97': 2, 'uhhh': 1, 'darth': 6, 'vader': 6, 'chapter': 1, 'muppet': 1, 'isn': 1, 'potato': 8, 'favourite': 1, 'probably': 1, 'quiet': 1, 'front': 1, 'sunrise': 1, 'intolerance': 1, 'schindler': 1, 'north': 1, 'northwest': 1, 'hur': 1, 'mystic': 1, 'paradisio': 1, 'sierra': 1, 'madre': 1, 'heat': 3, 'connection': 1, 'breathless': 1, 'avventura': 1, 'saturday': 1, 'battleship': 1, 'potemkin': 1, 'un': 1, 'chien': 1, 'andalou': 1, 'strawberries': 1, 'ugestu': 1, 'monogatari': 1, 'ambersons': 1, 'zhivago': 1, 'exorcist': 1, 'towers': 1, 'dictator': 1, 'deadly': 1, 'darling': 1, 'clementine': 1, 'yojimbo': 1, 'cabinet': 1, 'caligari': 1, 'discreet': 1, 'charm': 1, 'bourgeoisie': 1, 'peeping': 1, 'marienbad': 1, 'noon': 1, 'wages': 1, 'splendor': 1, 'grass': 1, 'fargo': 1, 'ikiru': 1, 'singin': 2, 'aguirre': 1, ',': 1, 'wrath': 2, 'aparajito': 1, 'philadelphia': 1, 'manchurian': 1, 'candidate': 1, 'afternoon': 1, 'diabolique': 1, 'traffic': 1, 'cries': 1, 'whispers': 1, 'closely': 1, 'trains': 1, 'strangelove': 1, 'maltese': 1, 'falcon': 1, 'chinatown': 1, 'rules': 1, 'scrooge': 1, 'eraserhead': 1, 'apu': 1, 'grapes': 1, 'rosemary': 1, 'das': 1, 'boot': 1, 'stagecoach': 1, 'yankee': 1, 'doodle': 1, 'dandy': 1, 'kwai': 1, 'asphalt': 1, 'stranger': 3, 'alicein': 1, 'hara': 1}\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "d7pIJmZCgRMA" + }, + "source": [ + "### 1.2.4 Cumulative token frequency" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 529 + }, + "id": "1mi4Vw5ABFXg", + "outputId": "0bb72afe-75e2-4f9c-d64c-cc8694710a71" + }, + "source": [ + "# Plot the cumulative distribution of token frequency\n", + "def cumulative_token_frequency(series, limit=20):\n", + " '''\n", + " Input:\n", + " series - pd.Series of words\n", + " Output:\n", + " [plot] - cumulative distribution of token frequency\n", + " '''\n", + " corpus=[word for word in series]\n", + " counter=Counter(corpus)\n", + " tokens_count = dict(counter).items()\n", + "\n", + " prop_list = []\n", + " print(\"Vocabulary Size: \", len(tokens_count))\n", + " for i in range(limit):\n", + " tokens_filtered = len(list(filter(lambda x: x[1]<=i, tokens_count)))\n", + " prop_list.append(round(tokens_filtered*100/len(tokens_count),2))\n", + " a4_dims = (11.7, 8.27)\n", + " fig, ax = plt.subplots(figsize=a4_dims)\n", + " plt.plot(prop_list)\n", + " plt.grid()\n", + " plt.xlabel(\"Counts\")\n", + " plt.ylabel(\"Proportion of Vocabulary (%)\")\n", + " # print(\"Proportion of unique words less than\",limit,\": \", round(tokens_filtered*100/len(tokens_dict),2),\"%\")\n", + "\n", + "cumulative_token_frequency(df[\"Word\"])" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Vocabulary Size: 6710\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "KE4VV7X4fUwj" + }, + "source": [ + "### 1.2.5 Entity Frequency" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 546 + }, + "id": "dB_9MvjOAoqq", + "outputId": "d29bc77c-a1a2-490c-c03f-154021e36148" + }, + "source": [ + "tag_counter = plot_top_non_stopwords_barchart(df[\"Tag\"], top=25, word=False)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "There are 25 distinct tags in dataset\n", + "{'O': 61008, 'B-ACTOR': 3220, 'I-ACTOR': 3474, 'B-YEAR': 2858, 'B-TITLE': 2376, 'B-GENRE': 4354, 'I-GENRE': 786, 'B-DIRECTOR': 1720, 'I-DIRECTOR': 1850, 'B-SONG': 245, 'I-SONG': 446, 'B-PLOT': 1927, 'I-PLOT': 1687, 'B-REVIEW': 221, 'B-CHARACTER': 385, 'I-CHARACTER': 342, 'B-RATING': 2007, 'B-RATINGS_AVERAGE': 1869, 'I-RATINGS_AVERAGE': 1673, 'I-TITLE': 3495, 'I-RATING': 840, 'B-TRAILER': 113, 'I-TRAILER': 7, 'I-REVIEW': 132, 'I-YEAR': 2456}\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAxkAAAHuCAYAAADz68F3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzda1xV1b7/8S9qASoQipl/L0jeBS9TpLR0l3jZdsrDzgu67IJluPfJTMlLaextGUmmbqxdnVctL6mnUMutaW0zj5fSklrgBLlkuBQ1y0KJjLaYIvwfeJi1AlxeFuDl83691oM5xphj/OaqB/wc4zeXl6QyAQAAAICH1KntAAAAAABcXUgyAAAAAHgUSQYAAAAAjyLJAAAAAOBRJBkAAAAAPKpebQeAy09+fr4OHjxY22EAAADgMhYcHKwbb7yx0j6SDFRw8OBBRURE1HYYAAAAuIw5HI4q+0gyUEG3ViHKf215bYcBAAAAN2589IHaDqFS1GQAAAAA8CiSDAAAAAAeRZIBAAAAwKNIMq5yzZs319q1a5Wbmyun06kFCxbouuuuq+2wAAAAcBUjybjK/fOf/9TatWvVvn17tW/fXg0bNtTzzz9f22EBAADgKkaScRWLjIzUyZMn9eabb0qSSktLFRcXp4cffli+vr61GxwAAACuWiQZV7HQ0FClpaW5tBUVFenQoUNq27atS3tsbKwcDoccDofq+HrXZJgAAAC4ypBkQJJkt9sVERGhiIgIlRb/UtvhAAAA4ApGknEVy8nJUXh4uEubn5+fWrVqJafTWUtRAQAA4GpHknEV27x5s+rXr68HHjj7S5B16tTR/Pnz9eabb6q4uLiWowMAAMDViiTjKnfvvfdqxIgRys3NVW5urk6ePKkZM2bUdlgAAAC4itWr7QBQvQ4fPqz//M//rO0wAAAAcA1hJwMAAACAR7GTgQoyDuUp4tEHajsMAAAAXKHYyQAAAADgUSQZAAAAADyKJAMAAACAR1GTgQq6tWqh716bV9thXJCbHp1S2yEAAADg/7CTAQAAAMCjSDIAAAAAeBRJBgAAAACPIsnwkJKSEpmmqfT0dKWlpal3795Vjm3btq3Wr18vp9Op1NRUbdmyRX379pUkxcTEKD8/X6ZpWp9OnTopODhYZWVleuyxx6x5/vGPfygmJkaStGTJEu3fv9+KITIy0hq3detW7dmzx5rvnXfeqaZvAQAAAKDw22OKi4tlGIYkadCgQUpMTNSdd95ZYZy3t7c++OADTZkyRevXr5ckhYaGqmfPntq+fbskaeXKlZowYYLLfcHBwfr+++81ceJEvf766zp9+nSFuadOnarVq1frzjvv1BtvvKH27dtbfffdd5/S0tI89bgAAABAldjJqAb+/v4qLCystO++++7Tzp07rQRDkrKzs7V06VK38x49elSbN2+2di+qsnPnTjVv3vzCggYAAAA8hJ0MD/H19ZVpmvLx8VGzZs1cjiv9VmhoqHbt2nXOuUaOHKk+ffpY1789ejVnzhxt2LBBixcvrvL+wYMHa+3atS5tb731loqLiyVJmzZt0rRp01z6Y2NjNW7cOElSHV/fc8YHAAAAnAtJhof89rhUr169tGzZMoWFhbm975///KfatWun3NxcDRs2TFLlx6XK5eXl6fPPP9fo0aMr9M2dO1ezZ89WixYtKtSEuDsuZbfbZbfbJUmnvj/iNm4AAACgKhyXqgYpKSkKCgpSkyZNlJCQYBVcS2ePRvXo0cMaO3ToUI0ZM0aNGjU67/lnz56tJ598Ul5eXi7tU6dOVYcOHfTkk0+ec6cDAAAAqE4kGdWgQ4cOqlu3rgoKChQfHy/DMKxdjrffflu33367hgwZYo2vX7/+Bc3/1VdfKScnx2WO33rllVdUp04dDRo06OIfAgAAALhIHJfykPKaDEny8vJSTEyMSktLK4w7efKk7rnnHv3973/XggUL9P3336uoqEgJCQnWmN/XZDz66KP69ttvXeZ5/vnnrfUqk5CQoGnTpumjjz6S5FqTcezYMQ0cOPDiHxYAAAA4By9JZbUdBC4vp74/oh9Wv1XbYVyQmx6dUtshAAAAXFMcDociIiIq7eO4FAAAAACP4rgUKsg4dFgR7AwAAADgIrGTAQAAAMCjSDIAAAAAeBRJBgAAAACPoiYDFXRt2UiHXq74i+K1pdXjb9d2CAAAALgA7GQAAAAA8CiSDAAAAAAeRZIBAAAAwKOoyahFRUVF8vPzs67HjBmjiRMnSpI6d+6sr776SmfOnNGHH36oPXv2qGfPnjpy5IhGjBghSerSpYsyMzMlSYsXL1ajRo30888/a/78+S7rlJSUWOMkacWKFZozZ051Px4AAACuUSQZl5E333xTb775piQpLy9P/fr1U0FBgSQpJiZGkjR79mzNnj1b0tkkxTAM6/6ZM2dWOm9xcbHLOAAAAKA6cVwKAAAAgEeRZFwDfH19ZZqm9YmOjq4wJjY2Vg6HQw6HQ3V8/SqZBQAAADg/HJe6BpzPcSm73S673S5J+uW7fTURFgAAAK5S7GRcBhISEqxdBgAAAOBKR5JxGYiPj5dhGBRnAwAA4KpAknGViY+P19dff219pIo1GYmJibUcJQAAAK5mXpLKajsIXF5++W6fvl/119oOw9Lq8bdrOwQAAAD8jsPhUERERKV97GQAAAAA8CjeLoUKdn/9gyLYPQAAAMBFYicDAAAAgEeRZAAAAADwKJIMAAAAAB5FTQYq6NQqQGn/fU9th2EJ/6/3azsEAAAAXAB2MgAAAAB4FEkGAAAAAI8iyQAAAADgUSQZ1aioqKjKvsaNG+vUqVP685//7NLetGlTJScny+l0KjU1VR988IG6du0q0zRlmqYKCgq0f/9+maapTZs2SZI6d+6szZs3a8+ePcrNzVV8fLw1X0xMjPLz82Wapr788ktNmjSpeh4WAAAA+D8kGbVkxIgRSklJkc1mc2lfs2aNtm3bprZt26pnz56aPn26/P39ZRiGDMPQunXrNHXqVBmGoYEDB8rHx0fr1q3TCy+8oI4dO6pbt2667bbb9Oijj1pzrly5UoZh6Pbbb9fTTz+tFi1a1PTjAgAA4BpCklFLbDabJk+erObNm6t58+aSpH79+un06dN6/fXXrXG7d+/Wjh07qpxn9OjR+vTTT61djeLiYj322GN66qmnKoz94Ycf5HQ61axZMw8/DQAAAPArkoxa0KJFCzVr1kwOh0OrVq3SyJEjJUlhYWFKS0u7oLlCQ0Mr3LN//341bNhQfn5+Lu0tW7aUj4+Pdu/eXWGe2NhYORwOORwO1fMNuMAnAgAAAH5FklELRo4cqVWrVkmSVqxYUeHIVHWsl5GRIafTqddee02//PJLhTF2u10RERGKiIhQSfHxao0HAAAAVzeSjBqQkJBgFW5LZ49KjRkzRnl5eVq3bp26du2qtm3bKjs7W+Hh4Rc0d05OToV7QkJC9PPPP1uF5ytXrrRqNV544QU1bdrUMw8GAAAAVIIkowbEx8dbhdvt2rVTw4YN1aJFC4WEhCgkJESJiYmy2WzasmWLvL29FRsba93bpUsX9enTp8q533rrLfXp00f9+/eXJPn4+Ojll1/Wiy++WGFsWlqali9frokTJ3r+IQEAAID/Q5JRw2w2m9asWePStnr1auvI1L333qsBAwbI6XQqKytLiYmJ+u6776qc7+TJk4qKilJ8fLz27NmjzMxMORwOvfLKK5WOnzNnjh566CE1bNjQcw8FAAAA/IaXpLLaDgKXl5+/z9VX/3yitsOwhP/X+7UdAgAAAH7H4XAoIiKi0j52MgAAAAB4VL3aDgCXny8PHVcEuwcAAAC4SOxkAAAAAPAokgwAAAAAHkWSAQAAAMCjqMlABe2DA7TJfletxjAwdkOtrg8AAICLx04GAAAAAI8iyQAAAADgUSQZAAAAADyKJMODSkpKZJqm0tPTlZaWpt69e1c5tnHjxjp16pT+/Oc/u7Q3bdpUycnJcjqdSk1N1QcffKCuXbvKNE2ZpqmCggLt379fpmlq06ZNkqTOnTtr8+bN2rNnj3JzcxUfH2/NFxMTo/z8fJmmqS+//FKTJk2qnocHAAAA/g9JhgcVFxfLMAx1795d06dPV2JiYpVjR4wYoZSUFNlsNpf2NWvWaNu2bWrbtq169uyp6dOny9/fX4ZhyDAMrVu3TlOnTpVhGBo4cKB8fHy0bt06vfDCC+rYsaO6deum2267TY8++qg158qVK2UYhm6//XY9/fTTatGiRbV9BwAAAABJRjXx9/dXYWFhlf02m02TJ09W8+bN1bx5c0lSv379dPr0ab3++uvWuN27d2vHjh1VzjN69Gh9+umn1q5GcXGxHnvsMT311FMVxv7www9yOp1q1qzZxT4WAAAA4BavsPUgX19fmaYpHx8fNWvWTJGRkZWOa9GihZo1ayaHw6FVq1Zp5MiR+vvf/66wsDClpaVd0JqhoaEV7tm/f78aNmwoPz8/l/aWLVvKx8dHu3fvrjBPbGysxo0bJ0m6zsf/gmIAAAAAfoudDA8qPy7VqVMnDR48WMuWLat03MiRI7Vq1SpJ0ooVKyocmfK0kSNHKiMjQ06nU6+99pp++eWXCmPsdrsiIiIUERGh0yd/qtZ4AAAAcHUjyagmKSkpCgoKUpMmTZSQkGAVbktnj0qNGTNGeXl5Wrdunbp27aq2bdsqOztb4eHhF7ROTk5OhXtCQkL0888/q6ioSNLZmozyWo0XXnhBTZs29cxDAgAAAJUgyagmHTp0UN26dVVQUKD4+HircLtdu3Zq2LChWrRooZCQEIWEhCgxMVE2m01btmyRt7e3YmNjrXm6dOmiPn36VLnOW2+9pT59+qh///6SJB8fH7388st68cUXK4xNS0vT8uXLNXHiRM8/MAAAAPB/SDI8qLwmwzRNrVy5UjExMSotLXUZY7PZtGbNGpe21atXW0em7r33Xg0YMEBOp1NZWVlKTEzUd999V+WaJ0+eVFRUlOLj47Vnzx5lZmbK4XDolVdeqXT8nDlz9NBDD6lhw4aX+LQAAABA5bwkldV2ELi8HM/P1Rfv1e5ux8DYDbW6PgAAAM7N4XAoIiKi0j52MgAAAAB4FK+wRQW5B4+zkwAAAICLxk4GAAAAAI8iyQAAAADgUSQZAAAAADyKmgxUcHNrf72z+I81uuaIhzfW6HoAAACoPuxkAAAAAPAokgwAAAAAHkWSAQAAAMCjSDJqWElJiUzTVHp6utLS0tS7d2+X/kceeUQrVqywrv38/OR0OhUSEqIlS5Zo//79Mk1Tpmnq008/dbl3zZo12rlzp0vbzJkzdfjwYZmmqezsbI0aNar6Hg4AAAAQSUaNKy4ulmEY6t69u6ZPn67ExESX/oULF6ply5bq37+/JGnWrFlavHix8vLyJElTp06VYRgyDEO33367dV9AQIDCw8MVEBCgkJAQlzmTkpJkGIaioqL0+uuvq1496v0BAABQfUgyapG/v78KCwsrtP/lL3/RggULFB4erv79+2vu3Llu5xo6dKjWr1+vFStWVLlb4XQ6deLECQUGBl5y7AAAAEBV+CftGubr6yvTNOXj46NmzZopMjKywpjMzExt3LhRmzdvVlRUlE6fPm31zZ07V/Hx8ZKk7Oxs3X///ZIkm82mWbNm6fvvv9fq1asr7JBIkmEY2rt3r44ePVqhLzY2VuPGjZMkefsEeORZAQAAcG0iyahh5celJKlXr15atmyZwsLCKox79dVXddddd+njjz92aZ86dapWr17t0nbjjTeqXbt22rFjhyTp9OnTCg0NVXZ2tiQpLi5ODz30kNq3b68hQ4ZUGpfdbpfdbpckFRz96tIeEgAAANc0jkvVopSUFAUFBalJkyZKSEiwCrolqbS0VKWlpec1T3R0tAIDA5WXl6e8vDy1bt1aNpvN6k9KSlJYWJiGDRumRYsWydvbu1qeBwAAAJBIMmpVhw4dVLduXRUUFCg+Pt4q6L5QNptNgwcPVkhIiEJCQhQeHl5pXcb69euVmpqqmJgYT4QPAAAAVIrjUjWsvCZDkry8vBQTE3PeOxaSa02GdHYXIzg4WCkpKVbbgQMHdPz4cd1yyy0V7p81a5befvtt2e12lZWVXcKTAAAAAJXzksRfmnBRcPQrbVn/eI2uOeLhjTW6HgAAAC6Nw+FQREREpX0clwIAAADgURyXQgX7D/zEzgIAAAAuGjsZAAAAADyKJAMAAACAR5FkAAAAAPAoajJQQXBrfy1cOqha13gk5qNqnR8AAAC1h50MAAAAAB5FkgEAAADAo0gyAAAAAHgUSUYtKyoqcrl+5JFHtGLFCuvaz89PTqdTISEhWrJkifbv3y/TNGWapj799FOXe9esWaOdO3e6tM2cOVOHDx+WaZrKzs7WqFGjqu9hAAAAAJFkXHYWLlyoli1bqn///pKkWbNmafHixcrLy5MkTZ06VYZhyDAM3X777dZ9AQEBCg8PV0BAgEJCQlzmTEpKkmEYioqK0uuvv6569aj3BwAAQPUhybgM/eUvf9GCBQsUHh6u/v37a+7cuW7vGTp0qNavX68VK1ZUuVvhdDp14sQJBQYGejpkAAAAwEKScRnKzMzUxo0btXnzZk2YMEGnT5+2+ubOnWsdl/qf//kfq91msyk5OVnJycmy2WyVzmsYhvbu3aujR49W6IuNjZXD4ZDD4ZCPT4DnHwoAAADXDM7NXKZeffVV3XXXXfr4449d2qdOnarVq1e7tN14441q166dduzYIUk6ffq0QkNDlZ2dLUmKi4vTQw89pPbt22vIkCGVrme322W32yVJ+Ue/8vTjAAAA4BrCTsZlIiEhwdqhkKTS0lKVlpae173R0dEKDAxUXl6e8vLy1Lp1a5fdjKSkJIWFhWnYsGFatGiRvL29q+UZAAAAAIkk47IRHx9vFXRfKJvNpsGDByskJEQhISEKDw+vtC5j/fr1Sk1NVUxMjCdCBgAAACpFknGF+W1NhmmaateunYKDg5WSkmKNOXDggI4fP65bbrmlwv2zZs3SE088IS8vr5oMGwAAANcQL0lltR0ELi/5R7/Sun9NqNY1Hon5qFrnBwAAQPVyOByKiIiotI+dDAAAAAAexdulUMHBAz+x0wAAAICLxk4GAAAAAI8iyQAAAADgUSQZAAAAADyKmgxU0CLEX3PfHlSta0wdTc0HAADA1YqdDAAAAAAeRZIBAAAAwKNIMgAAAAB4FDUZNaikpESZmZny8vLSmTNn9Nhjj2nnzp1W/5gxYzRx4kRJUufOnfXVV1/pzJkz+vDDD7Vnzx717NlTR44c0YgRIyRJXbp0UWZmpiRp8eLFatSokX7++WfNnz+/0nXLrVixQnPmzKnuxwUAAMA1iiSjBhUXF8swDEnSoEGDlJiYqDvvvNPqf/PNN/Xmm29KkvLy8tSvXz8VFBRIkmJiYiRJs2fP1uzZsyVJRUVF1nySNHPmTLfrAgAAANWN41K1xN/fX4WFhbUdBgAAAOBx7GTUIF9fX5mmKR8fHzVr1kyRkZE1um65xMRErVq1ymVMbGysxo0bd3a8T0CNxAUAAICrE0lGDfrtsaVevXpp2bJlCgsLq9F1q2K322W32yVJR459Ve0xAQAA4OrFcalakpKSoqCgIDVp0kQJCQkyTdNltwEAAAC4UpFk1JIOHTqobt26KigoUHx8vAzDoDgbAAAAVwWOS9Wg39ZGeHl5KSYmRqWlpR5dIz4+XpMmTbKuW7ZsWaEm48MPP9T06dM9ui4AAABQzktSWW0HgcvLkWNf6X8+mlCta0wd/VG1zg8AAIDq5XA4FBERUWkfx6UAAAAAeBTHpVDB4byf2GkAAADARWMnAwAAAIBHkWQAAAAA8CiSDAAAAAAeRU0GKrixjb8mvvtHj8750vCNHp0PAAAAly92MgAAAAB4FEkGAAAAAI8iyQAAAADgUSQZl6ikpESmaSo9PV1paWnq3bt3peNmzpypw4cPyzRNZWdna9SoUS79UVFRKisrU4cOHSRJKSkpMk1TBw8eVH5+vkzTlGmaCg4OVl5enho3bixJKisr07x586x5Jk+erJkzZ1rX9913nzIyMpSVlaX09HTZ7XYFBAR4+msAAAAALCQZl6i4uFiGYah79+6aPn26EhMTqxyblJQkwzAUFRWl119/XfXq/Vp3b7PZtH37dtlsNklSr169ZBiG/va3v2nlypUyDEOGYejgwYMuc548eVJDhw61ko7f+uMf/6i4uDjdddddCgsLU48ePfTZZ5+padOmHnp6AAAAoCKSDA/y9/dXYWGh23FOp1MnTpxQYGCgJKlBgwbq06ePxo4dW2GHw52SkhK98cYbiouLq9D39NNPa8qUKfr2228lSaWlpVqyZIlyc3MvaA0AAADgQvAK20vk6+sr0zTl4+OjZs2aKTIy0u09hmFo7969Onr0qKSzR6U+/PBD7d27VwUFBerRo4d27dp13jG8+uqr2r17t1588UWX9tDQ0POeJzY2VuPGjZMkNbye41QAAAC4eOxkXKLy41KdOnXS4MGDtWzZsirHxsXFKSsrS59//rmef/55q91ms2nFihWSpBUrVlhHps5XUVGRli1bpscff7zKMWFhYTJNU06nU9HR0RX67Xa7IiIiFBERoZ9PHb+g9QEAAIDfIsnwoJSUFAUFBalJkyZKSEiwirXLJSUlKSwsTMOGDdOiRYvk7e2twMBARUZGauHChcrLy9PUqVMrTQLcWbBggcaOHasGDRpYbdnZ2erRo4ckKSsrS4ZhaMOGDfL19b30hwUAAACqQJLhQR06dFDdunVVUFCg+Ph4q1j799avX6/U1FTFxMRo+PDhWr58uVq3bq2QkBC1atVKeXl56tu37wWtXVhYqFWrVmns2LFWW2JioubNm6fmzZtbbSQYAAAAqG7UZFyi8poMSfLy8lJMTIxKS0vd3jdr1iy9/fbb+vbbbzVnzhyXvtWrV1tvm7oQ8+fP12OPPWZdb9iwQU2aNNGGDRtUt25d/fjjj8rKytLGjRsvaF4AAADgQnhJKqvtIHB5OfjDV/r7lqrrOy7GS8NJbAAAAK4mDodDERERlfZxXAoAAACAR3FcChXk7/uJnQcAAABcNHYyAAAAAHgUSQYAAAAAjyLJAAAAAOBR1GSgAv+2QRq89mGPzffhnxZ7bC4AAABc/tjJAAAAAOBRJBkAAAAAPIokAwAAAIBHkWTUsJKSEpmmqfT0dKWlpal3796Vjps5c6YOHz4s0zSVmZmpIUOGWO2TJ0+uMD4qKkoZGRnKycnR7t27FRUVJUl65ZVXZJqmsrOzdeLECZmmKdM0NWzYsOp7SAAAAFzTKPyuYcXFxTIMQ5I0aNAgJSYm6s4776x0bFJSkubPn6+OHTtq+/btuvHGGysd17VrV82bN08DBw7UgQMH1Lp1a23atEn79+/XY489JkkKDg7W+++/b60NAAAAVBd2MmqRv7+/CgsL3Y7bs2ePSkpKFBQUVGn/lClTNHv2bB04cECSdODAASUmJmrq1KmeDBcAAAA4L+xk1DBfX1+ZpikfHx81a9ZMkZGRbu+55ZZbVFpaqqNHj1baHxoaqnnz5rm0paamavz48ecdV2xsrMaNGydJCri+4XnfBwAAAPweSUYN++1xqV69emnZsmUKCwurdGxcXJzuv/9+FRUVaeTIkdUal91ul91ulyR9VZhXrWsBAADg6sZxqVqUkpKioKAgNWnSRAkJCVZRdrmkpCQZhqE//OEP2rFjR5Xz5OTkKDw83KUtPDxc2dnZ1RY7AAAAUBWSjFrUoUMH1a1bVwUFBYqPj5dhGBdVmD1v3jxNnz5dwcHBks4Wec+YMUPz58/3dMgAAACAWxyXqmHlNRmS5OXlpZiYGJWWll7QHPHx8Zo0aZJ13bJlSz355JNav369rrvuOp0+fVrTpk1TRkaGR2MHAAAAzoeXpLLaDgKXl68K8zTx4+c8Nt+Hf1rssbkAAABweXA4HIqIiKi0z+1xqZtvvlnXX3+9JOmOO+7QhAkTFBAQ4NkIAQAAAFw13O5kmKapnj17qnXr1vrXv/6l9957T6Ghobr77rtrKETUtHNlpQAAAIB0iTsZpaWlOnPmjO6991794x//0LRp09SsWTOPBwkAAADg6uA2yTh9+rRGjRqlmJgYvf/++5Kk6667rtoDAwAAAHBlcptkPPTQQ+rdu7eef/55HThwQK1bt9by5ctrIjYAAAAAV6ALervUDTfcoJYtWyozM7MaQ0Jty/3hG03c5pk3Qm0Y+lePzAMAAIDLyyXVZGzdulV+fn4KDAzUrl27ZLfb+ZE3AAAAAFVym2QEBASoqKhIQ4cO1bJly9SrVy8NGDCgJmIDAAAAcAVym2TUq1dPN910k6Kjo63CbwAAAACoitskY9asWdq4caOcTqdSU1MVEhKivXv31kRsAAAAAK5AbpOMd999V926ddP48eMlSXl5eRo+fLjbiUtKSmSaptLT05WWlqbevXtXOm7mzJk6fPiwTNNUdna2Ro0a5dIfFRWlsrIydejQQZKUkpIi0zR18OBB5efnyzRNmaap4OBg5eXlqXHjxpKksrIyzZs3z5pn8uTJmjlzpnV93333KSMjQ1lZWUpPT5fdbrd+yfzuu+/Wrl27lJ6eruzsbI0bN87t85qmqeTkZEmSr6+vjh07Jj8/P5cxa9asUXR0tGJiYlxiN01TnTp1UnBwsE6cOGF9F0uXLlW9evWs++vWrav8/HwlJia6zNugQQO99tprcjqdSktLU2pqqh555BFJcpmz/PPAAw+4fR4AAADgYtVzN+Cll16q0Hb8+HGlpqZq3bp1Vd5XXFwswzAkSYMGDVJiYqLuvPPOSscmJSVp/vz5atu2rdLS0vTuu++qpKREkmSz2bR9+3bZbDY988wz6tWrlyQpJiZGPXv21IQJEyqd8+TJkxo6dKgSExNVUFDg0vfHP/5RcXFxuuuuu/Ttt9+qTp06iomJUdOmTfXvf/9bb7zxhm655RZ98803uv7669W6detzfkcdO3ZU3bp11bdvX9WvX18nTpzQxo0bde+992rZsmWSJH9/f/Xp00ejR49WdHS0Vq5cWSH24OBg7du3T1EuwrsAACAASURBVIZhqE6dOtq0aZOio6P19ttvS5IGDhyo3NxcjRgxQtOnT7fuW7hwofbv36927dqprKxMQUFBevjhh63+8jkBAACAmuB2J8PHx0fdu3fX3r17tXfvXnXt2lUtWrTQ2LFjlZSUdF6L+Pv7q7Cw0O04p9OpEydOKDAwUNLZf6Hv06ePxo4dW2GHw52SkhK98cYbiouLq9D39NNPa8qUKfr2228lnf1V8yVLlig3N1d+fn6qV6+elZicOnVKubm551zLZrNp+fLl+uijjxQVFSVJSk5Odon53nvv1caNG1VcXHxe8ZeWluqLL75Q8+bNXdZ56aWXdOjQIWtn6Oabb9Ytt9yi+Ph4lZWdfRvxsWPH9OKLL57XOuViY2PlcDjkcDjk713/gu4FAAAAfsvtTkbXrl11++23q7S0VJL03//939q+fbv69Olzzt/L8PX1lWma8vHxUbNmzRQZGek2GMMwtHfvXh09elTS2aNSH374ofbu3auCggL16NFDu3btOt9n06uvvqrdu3dX+IM7NDS0ynkKCwu1bt06HTx4UJs3b9b777+v5ORk6w/4yowcOVIDBw5Ux44dNWHCBCUnJ2vjxo1auHChGjVqpB9++EGjRo3SK6+84nJPnz59rOvfHyfz9vbWrbfeqokTJ1rXAwYM0J///GfdcMMNstls2rlzp0JDQ5WRkXHO+Nq0aSPTNK3rCRMmaMeOHS5j7Ha77Ha7pLO/kwEAAABcLLc7GYGBgWrYsKF13aBBAzVq1EilpaX65Zdfqryv/LhUp06dNHjwYOvYUGXi4uKUlZWlzz//XM8//7zVbrPZtGLFCknSihUrZLPZzuuhyhUVFWnZsmV6/PHHqxwTFhYm0zTldDoVHR0t6ey/6vfv319ffPGFpkyZosWLq/5huvDwcB07dkxff/21Nm/eLMMwFBgYqNOnT2vdunUaPny4GjduLMMwtHHjRuu+lStXyjAM63Py5ElJvyYE33//vY4cOWIlcvfcc4+2bt2qkydPavXq1frTn/6kOnUq/uebMWOGTNPUN9/8miiUH5cq//w+wQAAAAA8yW2S8eKLLyo9PV2LFy/WkiVLZJqm5s6dq/r16+t///d/z2uRlJQUBQUFqUmTJkpISLAKkMslJSUpLCxMw4YN06JFi+Tt7a3AwEBFRkZq4cKFysvL09SpU60k4EIsWLBAY8eOVYMGDay27Oxs9ejRQ5KUlZUlwzC0YcMG+fr6WmOysrK0YMECDRw4UMOGDatyfpvNpo4dOyovL0/79u2Tv7+/Nb78yNTw4cP13nvvWXUm51KeELRp00bh4eEaMmSItc6AAQOUl5entLQ0NW7cWJGRkcrJyVG3bt3k5eUlSZo9e7YMw5C/v/8Ff1cAAACAJ7hNMhYvXqzbbrtNa9eu1Zo1a9SnTx8tWrRIJ06c0LRp085rkQ4dOqhu3boqKChQfHy89S/qv7d+/XqlpqYqJiZGw4cP1/Lly9W6dWuFhISoVatWysvLU9++fS/oAQsLC7Vq1SqNHTvWaktMTNS8efNc6h3KE4wGDRrojjvusNq7d++ugwcPVjq3l5eXoqOj1aVLF4WEhCgkJERRUVHWjsu2bdvUrl07jR8/3nrz1PkqKCjQU089penTp8vPz099+/ZVq1atrHXGjx8vm82mffv2KTU1VQkJCdbOhre3t5V0AAAAADXNbZIhnX1T05EjR1RYWKi2bdue1x/65TUZpmlq5cqViomJseo6zmXWrFl64oknZLPZtGbNGpe+1atXX/CRKUmaP3++goKCrOsNGzbo5Zdf1oYNG5Sdna1PP/1UZ86c0caNG+Xl5aVp06Zpz549Mk1Tzz77rMaMGVPpvH379tU333yjI0eOWG2ffPKJOnfurJtuukllZWV699131bhxY3388ccu944cOdLltbKVveJ37dq1ql+/vuLi4rRlyxadOnXK6nvvvfc0ZMgQXX/99XrkkUfUuHFjOZ1OORwObdq0ySUBLD+CVf6p6o1cAAAAgCd4Saq6YljS2LFjNXHiRLVo0ULp6enq1auXdu7cqf79+9dQiKhpuT98o4nbqq5DuRAbhv7VI/MAAADg8uJwOBQREVFpn9u3S02cOFERERFKSUlRZGSkOnTooNmzZ3s8SFw+ju/7luQAAAAAF81tknHy5EnrLVLXX3+9vvrqK+vXt68lM2bM0IgRI1za3nnnHRIuAAAA4HfcJhmHDx9WQECA1q5dq02bNqmwsLDKQuir2ezZs0koAAAAgPPgNskYOnSoJOnZZ5/V1q1bFRAQoA0bNlR7YAAAAACuTG6TjGXLlunBBx+UdPbNSb9vw9UnoE0r/cfqf3hkrn8N401WAAAA1xq3r7ANDQ11vaFOHYWHh1dbQAAAAACubFUmGU899ZR++uknde3aVcePH9fx48f1008/KT8/X++9915NxggAAADgClJlkvHCCy/I399fc+fOVUBAgAICAuTv76+goCDNmDGjJmMEAAAAcAVxe1zqWk0oioqKKm2fOXOmDh8+LNM0lZubq9WrV6tTp05W/9atW63jZHl5edq9e7cyMjK0bds2tWrVyhpXUlLi8ivcTz75pCSpXr16SkxMVG5urtLS0vTZZ59p8ODBSklJkWmaOnjwoPLz8637goOD5e/vr6VLl2rv3r1yOp1aunSp/P39JUnBwcE6ceKETNNUdna2li5dqnr13JbiAAAAABfNbZKBipKSkmQYhtq3b6+VK1dqy5YtCgoKqnRsv3791K1bN23btk3x8fFWe3FxsQzDsD5z5syRJD333HNq1qyZwsLCFB4erj/96U/y8/NTr169ZBiG/va3v2nlypXWfQcPHtSiRYu0f/9+tWvXTm3btlVeXp4WLlxorbVv3z4ZhqEuXbqoRYsWio6Ort4vCAAAANc0koxLtGrVKn300UcaPXr0Ocft3LlTzZs3P+cYX19fxcbGasKECTp16pQkKT8/X++8806V97Rp00bh4eF67rnnrLZZs2apZ8+euvnmm13GlpaW6osvvnAbBwAAAHApzuvcTJ06ddS0aVOXYzZff/11tQV1pdm1a5c6dux4zjGDBw/W2rVrrWtfX1+ZpmldJyYm6ssvv9ShQ4eqPKpVmc6dOys9PV2lpaVWW2lpqdLT0xUaGqrdu3db7d7e3rr11ls1ceLECvPExsZq3LhxkiT/633Oe30AAADg99wmGY899phmzpyp77//3vpDtqysTN26dav24K4UXl5eVfZt3bpVjRo10s8//6y//vWvVnv5canf6tKlS7XE16ZNG5mmqZCQEH3wwQfKzMysMMZut8tut0uScn/4vlriAAAAwLXB7XGpiRMnqkOHDgoLC1PXrl3VtWvXayrBSEhIsIqsq2IYhr788stK+/r166fg4GClp6fr2WefPedaTqdTrVq1kp+f33nHl5OTo+7du7skOl5eXurevbtycnIk/VqTUX60asiQIec9PwAAAHCh3CYZX3/9tY4fP14TsVyW4uPjrSLrygwdOlSDBg1ScnJylXOcOXNGkyZN0oMPPqjAwMAqxxUXF2vRokV66aWXdN1110mSgoKCNHz48Crv2bdvn0zTdCkqj4+P165du7Rv3z6XsQUFBXrqqac0ffr0KucDAAAALpXbJGP//v3atm2bnnrqKcXFxVmfa1lcXJz1Ctv7779fkZGROnbs2Dnv+e6775ScnKzx48dL+rUmo/yTmJgo6WyCcPToUeXk5CgzM1Pvv/++fvrpp3POPXbsWLVv315Op1NOp1Pt27fX2LFjKx27du1a1a9fX3369LmIJwcAAADc85JUdq4Bf/vb3yptnzVrVnXEg8tA7g/fa9LWVR6Z61/DJnhkHgAAAFxeHA6HIiIiKu1zW/hdnkw0aNBAkvTvf//bg6EBAAAAuNq4TTJCQ0O1fPlyNWrUSJJ07NgxPfjgg1ZRMa4+x/cdYgcCAAAAF81tTcYbb7yhJ554Qq1bt1br1q01efJk61WnAAAAAPB7bpOMBg0aaNu2bdb1xx9/bB2dAgAAAIDfO6+3S8XHxys4OFjBwcF6+umntX///pqIDQAAAMAVyG1NxsMPP6xnn31W//znPyVJ27dv18MPP1ztgaH2BLRprbvfXXLJ83ww/CEPRAMAAIArjdsk48cff9TEiRNrIhYAAAAAVwG3SUZ4eLhmzJih1q1bq169X4d369atWgMDAAAAcGVym2S89dZbmjp1qjIzM1VaWloTMQEAAAC4grkt/D569KjWr1+vAwcO6NChQ9bnalNSUiLTNJWenq60tDT17t270nEzZ87U4cOHZZqmcnNztXr1anXq1Mnq37p1q8LDwyVJeXl52r17tzIyMrRt2za1atWqwnrlnyeffFKSVK9ePSUmJio3N1dpaWn67LPPNHjwYKWkpMg0TR08eFD5+fnWfcHBwfL399fSpUu1d+9eOZ1OLV26VP7+/pKk4OBgnThxQqZpKjs7W0uXLnXZkQIAAAA8ze1fmzNnzpTdbtfmzZv1yy+/WO1r1qyp1sBqWnFxsQzDkCQNGjRIiYmJuvPOOysdm5SUpPnz50uSoqOjtWXLFnXp0kXHjh2rMLZfv34qKCjQM888o/j4eI0bN67Cer/13HPPqVmzZgoLC9OpU6d044036o477lCvXr0kSTExMerZs6cmTPj1x/LeeecdZWVlKSYmRpL0zDPPaOHChYqOjpYk7du3T4ZhqE6dOtq0aZOio6P19ttvX+Q3BQAAAJyb2yTjoYceUseOHXXddddZx6XKysquuiTjt/z9/VVYWHheY1etWqW7775bo0eP1ssvv1zluJ07d+rxxx8/51y+vr6KjY1VSEiITp06JUnKz8/XO++8U+U9bdq0UXh4uEaOHGm1zZo1S06nUzfffLPOnDljtZeWluqLL75Q8+bNz+vZAAAAgIvhNsmIiIhQx44dayKWWuXr6yvTNOXj46NmzZopMjLyvO/dtWuX2+9o8ODBWrt2bYX1yiUmJurLL7/UoUOHVFRUdN5rd+7cWenp6S71MqWlpUpPT1doaKh2795ttXt7e+vWW2+t9G1hsbGx1i6L//U+570+AAAA8Htuk4zPPvtMnTp10pdfflkT8dSa3x5f6tWrl5YtW6awsLDzutfLy6vKvq1bt6pRo0b6+eef9de//rXS9cp16dLlIiJ3r02bNjJNUyEhIfrggw+UmZlZYYzdbpfdbpck5f5wtFriAAAAwLXBbeF3r169lJ6erj179igjI8MqZL6apaSkKCgoSE2aNFFCQoJVZF0VwzCqTML69eun4OBgpaen69lnnz3nuk6nU61atZKfn995x5qTk6Pu3bu7JDpeXl7q3r27cnJyJP1ak1F+tGrIkCHnPT8AAABwodwmGYMHD1a7du00aNAgDRkyRPfcc89V/0dqhw4dVLduXRUUFCg+Pl6GYVRapC1JQ4cO1aBBg5ScnFzlfGfOnNGkSZP04IMPKjAwsMpxxcXFWrRokV566SVdd911kqSgoCANHz68ynv27dsn0zQVHx9vtcXHx2vXrl3at2+fy9iCggI99dRTmj59epXzAQAAAJfKbZJx6NAh/fTTTwoICFDjxo2tz9WmvEbCNE2tXLlSMTExVf4uSFxcnPUK2/vvv1+RkZGVvlnqt7777jslJydr/PjxFdYzTVOJiYmSziYIR48eVU5OjjIzM/X+++/rp59+OufcY8eOVfv27eV0OuV0OtW+fXuNHTu20rFr165V/fr11adPH3dfCQAAAHBRvCSVnWvArFmzNGbMGO3bt09lZWeHlpWVqX///jURH2pB7g9HFbfl/Uue54PhD3kgGgAAAFyOHA6HIiIiKu1zW/gdHR2tNm3a6PTp0x4PDAAAAMDVx22SkZWVpRtuuEFHj/LGoWvF8X0H2IUAAADARXObZCQmJso0TWVlZbn84ndUVFS1BgYAAADgyuQ2yVi6dKnmzJmjzMzMKguhAQAAAKCc2yTjxIkT+sc//lETsQAAAAC4CrhNMrZv367Zs2dr3bp1LselzvXjdLiyBdx8s+5+Z+Ulz/PBiJEeiAYAAABXGrdJRvmP0PXq1ctq4xW2AAAAAKriNsmIjIysiTgAAAAAXCXc/uK3v7+/5s+fL4fDIYfDoXnz5snf378mYgMAAABwBXKbZCxevFhFRUWKjo5WdHS0fvrpJy1ZsqQmYrsmFBUVVdo+c+ZMHT58WKZpKjMzU0OGDLHaJ0+eXGF8VFSUMjIylJOTo927d1uvGH7llVdkmqays7N14sQJmaYp0zQ1bNiw6nsoAAAAXNPcHpdq06aNhg8fbl3PmjWLou8akpSUpPnz56tjx47avn27brzxxkrHde3aVfPmzdPAgQN14MABtW7dWps2bdL+/fv12GOPSZKCg4P1/vvvWzU2AAAAQHVxu5NRXFys22+/3bq+7bbbVFxcXK1BwdWePXtUUlKioKCgSvunTJmi2bNn68CBA5KkAwcOKDExUVOnTq3BKAEAAICz3O5k/Nd//ZeWLl2qgIAASVJhYaHGjBlT3XHhN2655RaVlpbq6NGjlfaHhoZq3rx5Lm2pqakaP378ea8RGxurcePGSZL8vb0vPlgAAABc89wmGRkZGerevbv8/PwkVV1DAM+Li4vT/fffr6KiIo0cWb2/OWG322W32yVJuQUF1boWAAAArm5uj0s9//zzCggIUFFRkYqKinTDDTfoueeeq4nYrikJCQlWUXa5pKQkGYahP/zhD9qxY0eV9+bk5Cg8PNylLTw8XNnZ2dUWLwAAAFAVt0nGXXfdpePHj1vXP/74o/7jP/6jWoO6FsXHx8swjIsqzJ43b56mT5+u4OBgSWeLvGfMmKH58+d7OkwAAADALbfHperWravrr79ep06dkiT5+PjImzP7tSo+Pl6TJk2yrlu2bKknn3xS69ev13XXXafTp09r2rRpysjIqMUoAQAAcK3yklR2rgHTpk3TkCFDrN/GeOihh7Ru3TrNnTu3JuJDLcgtKFDclv+95Hk+GFG9dSQAAACoPQ6HQxEREZX2ud3JePHFF5WRkaEBAwZIkp577jl99NFHno0QAAAAwFXD7U4Grj3nykoBAAAA6dx/M7ot/AYAAACAC0GSAQAAAMCj3CYZjz/++Hm1AQAAAIB0HjUZaWlpFX7obdeuXerRo0d1xoVatLfgR03e8uklzbFuxN0eigYAAACXo4t6u9SoUaM0evRohYSE6L333rPa/f399cMPP3g+SgAAAABXhSqTjM8++0xHjhxRUFCQyy9HFxUVaffu3TUSHAAAAIArT5VJxqFDh3To0CENGDBAxcXFKisrU7t27dSxY0dlZmbWZIwAAAAAriBuC78/+eQT+fj46P/9v/+njz76SA888IDefPPNGggNAAAAwJXIbZLh5eWl4uJiDR06VK+99pqio6MVGhp6wQsVFRVV2j5z5kwdPnxYpmkqOztbo0aNcumPiopSWVmZOnToIElKSUmRaZo6ePCg8vPzZZqmTNNUcHCw8vLy1LhxY0lSWVmZ5s2bZ80zefJkzZw507q+7777lJGRoaysLKWnp8tutysgIECSdPfdd2vXrl1KT09Xdna2xo0b5/b5TNNUcnKyJMnX11fHjh2Tn5+fy5g1a9YoOjpaMTExLrGbpqlOnTopODhYJ06csL6LpUuXql69Xzeb6tatq/z8fCUmJrrM26BBA7322mtyOp1KS0tTamqqHnnkEUlymbP888ADD7h9HgAAAOBinVeS0atXL91333364IMPJJ39Y9eTkpKSZBiGoqKi9Prrr7v8YW2z2bR9+3bZbDZJUq9evWQYhv72t79p5cqVMgxDhmHo4MGDLnOePHlSQ4cOtZKO3/rjH/+ouLg43XXXXQoLC1OPHj302WefqWnTpqpXr57eeOMNDRkyRN27d5dhGNq2bds54+/YsaPq1q2rvn37qn79+iouLtbGjRt17733WmP8/f3Vp08frV+/XpJcYjcMQ19++aUkad++fTIMQ126dFGLFi0UHR1tzTFw4EDl5uZqxIgRLusvXLhQhYWFateuncLDwzV48GA1atTI6i+fs/yzfPnycz4PAAAAcCncJhmTJk3S9OnTtWbNGuXk5CgkJERbt26tlmCcTqdOnDihwMBASWf/hb5Pnz4aO3ZshR0Od0pKSvTGG28oLi6uQt/TTz+tKVOm6Ntvv5UklZaWasmSJcrNzZWfn5/q1aungoICSdKpU6eUm5t7zrVsNpuWL1+ujz76SFFRUZKk5ORkl5jvvfdebdy4UcXFxecVf2lpqb744gs1b97cZZ2XXnpJhw4dUu/evSVJN998s2655RbFx8errOzs24iPHTumF1988bzWKRcbGyuHwyGHwyF/7+su6F4AAADgt86rJiMqKkqvvvqqGjRooLy8PE2cOLFagjEMQ3v37tXRo0clnT0q9eGHH2rv3r0qKCi44N/mePXVV3XffffJ39/fpT00NFS7du2q9J7CwkKtW7dOBw8e1Ntvv63Ro0fLy8vrnOuMHDlSK1asUHJysrXjsnHjRvXo0cPaURg1apR1nKr8nt8eYfLx8XGZ09vbW7feeqs+/PBD63rAgAFav369yzqhoaHKyMiwEozKtGnTxmWtPn36VBhjt9sVERGhiIgI/fTL6XM+LwAAAHAubpOMsLAw7dq1S9nZ2crJyVFqaqo6d+7s0SDi4uKUlZWlzz//XM8//7zVbrPZtGLFCknSihUrrD+sz1dRUZGWLVt2zl8oDwsLk2macjqd1tGk2NhY9e/fX1988YWmTJmixYsXV3l/eHi4jh07pq+//lqbN2+WYRgKDAzU6dOntW7dOg0fPlyNGzeWYRjauHGjdd/vj0udPHlS0q8Jwffff68jR45Yb/K65557tHXrVp08eVKrV6/Wn/70J9WpU/E/34wZM2Sapr755hur7ffHpXbs2HFB3yMAAABwIdwmGa+//rqeeOIJtW7dWsHBwZo8ebLsdvtFL5iQkGD9i3q5pKQkhYWFadiwYVq0aJG8vb0VGBioyMhILVy4UHl5eZo6dapLfcL5WrBggcaOHasGDRpYbdnZ2dauSFZWlgzD0IYNG+Tr62uNycrK0oIFCzRw4EANGzasyvltNps6duyovLw87du3T/7+/tb48iNTw4cP13vvvaeSkhK38ZYnBG3atFF4eLiGDBlirTNgwADl5eUpLS1NjRs3VmRkpHJyctStWzdrt2X27NkyDKPC7g0AAABQU9wmGQ0aNHApfP74449d/mC/UPHx8da/qP/e+vXrlZqaqpiYGA0fPlzLly9X69atFRISolatWikvL099+/a9oPUKCwu1atUqjR071mpLTEzUvHnzXOodyhOMBg0a6I477rDau3fvXqGovJyXl5eio6PVpUsXhYSEKCQkRFFRUdaOy7Zt29SuXTuNHz/e5ajU+SgoKNBTTz2l6dOny8/PT3379lWrVq2sdcaPHy+bzaZ9+/YpNTVVCQkJ1s6Gt7e32yNeAAAAQHVxm2Ts379f8fHxCg4OVnBwsJ5++mnt37+/2gKaNWuWnnjiCdlsNq1Zs8alb/Xq1Rd8ZEqS5s+fr6CgIOt6w4YNevnll7VhwwZlZ2fr008/1ZkzZ7Rx40Z5eXlp2rRp2rNnj0zT1LPPPqsxY8ZUOm/fvn31zTff6MiRI1bbJ598os6dO+umm25SWVmZ3n33XTVu3Fgff/yxy72/r8koL+T+rbVr16p+/fqKi4vTli1bdOrUKavvvffe05AhQ3T99dfrkUceUePGjeV0OuVwOLRp0yZNmzbNGvv7mowJEyZc8HcIAAAAnC8vSVVXDEu64YYb9Oyzz1rFwtu3b9czzzyjH3/8sSbiQy3YW/CjJm/59JLmWDfibg9FAwAAgMuRw+FQREREpX1ukwxce871PwwAAAAgnftvxnqVturscZxzKf89iGvFjBkzKvwI3jvvvKPZs2fXUkQAAADA5anKnYz8/Hx9/fXXSk5O1ueff16hkPiTTz6pifhQC9jJAAAAgDsXtZNx0003aeDAgbLZbBo9erT+f3v3HhZVtf8P/D3cvXERFQl0IC8cg6TRgylqppbXU5poQFZYRtrRUkPUzuGEFgVGhaWeUlTsoqIexFJTo5QvpoIDs7mjcplUTAHRPOQNxfX7g5/7OM4MeBkcjPfreT7Pw15r7bXWnkUTH9feMzt27MCGDRtQWFjYZBMlIiIiIqIHn9Ek4/r169i9ezd2794NGxsbBAcHIzU1FYsWLcLy5cvv5xzpPnPs5oXx/9nXYJutE+/so4SJiIiIqOUwmmQAgI2NDcaOHYvg4GB4eHjg888/1/tYWSIiIiIiopsZTTK++uor+Pj44IcffsCiRYtQUFBwP+dFREREREQPKKMPftfV1eHChQsAACH+10ShUEAIAQcHh/syQbr/is/+F+F7chpsw9uliIiIiFq2hh78NvqN35aWlrC3t4e9vT0cHBzkuHFMhtXU1Bgsj4yMRHl5OSRJQkFBAYKCgnTqx40bByEEvLy8AADp6emQJAnHjh1DZWWl/G3dSqUSWq0Wzs7OAOoTwI8//ljuJywsDJGRkfLx5MmTkZOTg/z8fGRnZyM+Pp7rR0RERERNymiSQaYXFxcHlUqFcePGYcWKFbCy+t/dasHBwdi3bx+Cg4MBAP3794dKpcK7776LjRs3QqVSQaVS4dixYzp9Xr58GRMmTJCTjpuNHDkSc+bMwejRo+Hj44M+ffrgwIEDcHFxadoLJSIiIqIWjUmGGZSUlODixYtwcnICALRp0waDBg3C1KlT9XY4GnPt2jWsXLkSc+bM0av75z//iblz5+K3334DUP+JYQkJCTh69Oi9XwQRERERkRFMMsxApVKhuLgYVVVVAOpvldq1axeKi4tRXV2NPn363FF/y5cvx+TJk2Fvb69T7u3tDY1Gc1t9hIaGQq1W/2uaVAAAIABJREFUQ61Ww97G+o7GJyIiIiK6GZOM+2jOnDnIz89HRkYGPvjgA7k8ODgYiYmJAIDExET5lqnbVVNTg6+//hpvvfWW0TY+Pj6QJAklJSV4/vnn9erj4+Ph5+cHPz8//Lf26h2NT0RERER0MyYZTSQqKkp+WPuGuLg4+Pj4ICAgAKtXr4atrS2cnJwwbNgwrFq1ClqtFuHh4QaTgMYsWbIEU6dORZs2beSygoICeVckPz8fKpUKO3fuRKtWre79AomIiIiIjGCS0UQiIiLkh7VvtW3bNmRmZiIkJAQTJ07EN998Aw8PD3h6eqJr167QarUYPPjOPiL23Llz2LRpE6ZOnSqXRUdH4+OPP4abm5tcxgSDiIiIiJpag9/4TU3nvffew/r16/Hbb79h8eLFOnVJSUnyp03diU8++QQzZ86Uj3fu3ImOHTti586dsLS0xO+//478/Hzs3r3bJNdARERERGSI0S/jo5aLX8ZHRERERI25qy/jIyIiIiIiuhu8XYr0/F56hDsVRERERHTXuJNBREREREQmxSSDiIiIiIhMikkGERERERGZFJ/JID1O3bwx6T8FRus3T/S+j7MhIiIiogcNdzKIiIiIiMikmGQQEREREZFJMckgIiIiIiKTYpLRRGpqaozWde/eHdu2bUNJSQkyMzOxZ88eDB5c/70UISEhqKyshCRJcvTq1QtKpRJCCMycOVPuZ+nSpQgJCQEAJCQkoKysDJIkITs7G8OGDZPb7d27F4cPH5b727x5cxNdNRERERERH/y+72xtbbFjxw7MnTsX27ZtAwB4e3vjr3/9K/bt2wcA2LhxI958802d85RKJSoqKjBr1iysWLECV69e1es7PDwcSUlJePLJJ7Fy5Ur07NlTrps8eTKysrKa8MqIiIiIiOpxJ+M+mzx5Mg4ePCgnGABQUFCAr776qtFzq6qq8PPPP8u7F8YcPHgQbm5u9zxXIiIiIqK7wZ2M+8zb2xsajabBNoGBgRg0aJB8PGDAAPnnxYsXY+fOnVizZo3R80eNGoWtW7fqlK1btw6XLl0CAKSkpGDevHk69aGhoXj99dcBAPY2lrd3MUREREREBjDJMLMtW7agR48eOHr0KAICAgAYvl3qBq1Wi4yMDLzwwgt6dbGxsfjwww/h7u6uk5gAjd8uFR8fj/j4eABAydmLd3s5RERERES8XaqpRUVFyQ9cA/W3RvXp00eunzBhAqZMmYL27dvfdp8ffvgh5s+fD4VCoVMeHh4OLy8vzJ8/v8GdDiIiIiKipsQko4lFRERApVJBpVIBANavX4+BAwfimWeekdu0bt36jvo8cuQICgsLdfq42bJly2BhYYERI0bc/cSJiIiIiO4Sb5e6zy5fvoy//e1v+PTTT7FkyRJUVFSgpqYGUVFRcptbn8n4+9//jt9++02nnw8++EDeHTEkKioK8+bNw48//ghA95mMM2fO4OmnnzblZRERERERyRQAhLknQc1LydmLeGeP1mj95one93E2RERERNQcqdVq+Pn5Gazj7VJERERERGRSvF2K9JwrLcDmiYazUiIiIiKixnAng4iIiIiITIpJBhERERERmRSTDCIiIiIiMik+k0F6HurWG5FbTuqULZrgZqbZEBEREdGDhjsZRERERERkUkwyiIiIiIjIpJhkEBERERGRSTHJuM9qamoMlvfs2RN79+6FJEkoLCzEihUr5LqBAwciIyMDRUVFKCoqQmhoqFwXGRmJCxcuoGPHjgbH6NSpE9atW4fS0lJkZmbiwIEDGD9+fBNcGRERERFRPSYZzcTnn3+OuLg4qFQqPPLII1i6dCkAwMXFBevXr8f06dPRq1cvDBo0CNOmTcOYMWPkc8+cOYOwsDCD/W7duhVpaWno1q0b/vrXvyIoKAju7u735ZqIiIiIqGViktFMuLq6ory8XD7Oz88HAMyYMQNr166FJEkAgOrqasybNw8LFiyQ265ZswaBgYFwcnLS6XPYsGGora3V2RU5fvw4li1b1pSXQkREREQtHJOMZiIuLg579uzBDz/8gNmzZ8PBwQEA4O3tjaysLJ22mZmZ8Pb2lo//+OMPrFmzBrNmzdJp5+3tDY1Gc1vjh4aGQq1WQ61Wo7Utfy2IiIiI6O7xr8lmYu3atejVqxc2b96MJ598Eunp6bCxsbnt8z///HOEhISgbdu2RtssW7YM2dnZOHTokF5dfHw8/Pz84Ofnh4tXrt/VNRARERERAUwyzCYqKgqSJMm3QQHAqVOnkJCQgPHjx+PatWvw8fFBYWEh+vbtq3Nu3759UVBQoFN2/vx5rF+/HjNmzJDLCgoK0KdPH/l45syZGD58uM5D4kREREREpsYkw0wiIiKgUqmgUqkAACNHjoSVVf0XsLu4uMDZ2RknT57E8uXLMWXKFPj6+gIA2rdvj8WLF+Ojjz7S6/PTTz/FtGnT5H727NkDOzs7TJ8+XW7TunXrpr40IiIiImrhrMw9Aao3YsQIfPbZZ7h8+TIAIDw8HBUVFQCAF198EfHx8WjXrh0UCgWWLFmC7du36/VRXV2N5ORkvP3223LZ+PHjERcXh3nz5qGqqgoXLlzA/Pnz789FEREREVGLpAAgzD0Jal5Onr2ClalndMoWTXAz02yIiIiIqDlSq9Xw8/MzWMfbpYiIiIiIyKR4uxTp+a00F4smGM5KiYiIiIgaw50MIiIiIiIyKSYZRERERERkUkwyiIiIiIjIpPhMBunx7OaLb/9TKR+/OLGTGWdDRERERA8a7mQQEREREZFJMckgIiIiIiKTYpJBREREREQm9adPMq5duwZJkpCdnY2srCwMGDDAaNuXXnoJeXl5yM3NhUajQVhYGAAgISEBAQEBOm1ramp0jmfNmoVLly7B3t5eLhsyZAh+//13SJKEoqIixMbG6pzj7OyM2tpaTJs2TafcxcUFGzZsQElJCTIzM7Fjxw707t0bkiRBkiRUV1ejrKwMkiQhJSUFSqUSFy9elOslScJLL70EANBqtcjNzUVOTg5SU1PRtWvXO38RiYiIiIjukPgzR01NjfzziBEjRGpqqsF2o0aNEllZWcLV1VUAEDY2NuK1114TAERCQoIICAgw2i8AkZ6eLtLS0sSUKVPksiFDhoht27YJAMLOzk4UFRUJf39/uX769OkiLS1Nb04HDhwQ06ZNk4979+4tBg0aJB/fOh+lUiny8vIMXpdWqxXOzs4CgFi4cKFYuXJlo6/ZmbO14tv/VMph7jVkMBgMBoPBYDS/UKvVRuv+9DsZN7O3t8e5c+cM1r3zzjuYO3cuTp06BQCora3FqlWrbqvfhx9+GG3btkVERASCg4MNtrl8+TKys7Ph5uYmlwUHByMsLAxubm5y+dChQ3H16lWsWLFCbpebm4tffvnltubSkIMHD+qMT0RERETUFP70H2HbqlUrSJIEOzs7uLq6YtiwYQbb+fj4ICsry2g/sbGxiIiIMFgXFBSExMRE7Nu3D15eXujUqRMqKyt12jg6OqJHjx5IS0sDALi7u8PV1RVqtRqbNm1CYGAgPv3000bnYUy3bt0gSZJ8/Oabb+olJqNGjcLWrVsNnh8aGorXX38dAGBr06JyTyIiIiIysT/9X5OXLl2CSqVCr169MGrUKHz99dd31U94eDhUKpUcNwsODkZiYiKEEEhKSsKkSZPkusGDByM7OxsnT57E7t27UVFRAQAIDAzEpk2bAACJiYlGd0BuV2lpqc78bk4w9u7di/LycowePRobNmwweH58fDz8/Pzg5+eHK7XX72kuRERERNSy/emTjJulp6ejQ4cO6NixI6KiouSHpAGgoKAAffv2veM+fXx80KNHD6SkpECr1SIoKEgnYdi3bx8ee+wxeHt7Y+rUqfD19QVQn5hMmTIFWq0W33//PXr37o3u3bvf9TwaMnToUCiVSmRnZ2PRokUm7ZuIiIiI6FYtKsnw8vKCpaUlqqurERERobMrER0djdjYWLi4uAAArK2tMXXq1Eb7DA4OxsKFC+Hp6QlPT0+4ubnhoYce0vsUp19//RUxMTGYP38+evTogbZt28Ld3V0+Lzo6GsHBwdizZw9sbW0RGhoqn/voo49i0KBB93TtdXV1mD17Nl5++WU4OTndU19ERERERA350ycZN57JkCQJGzduREhICK5f178daOfOnVi2bBl++ukn5OfnQ6PR6HwcrTFBQUFITk7WKUtOTkZQUJBe2y+//BJPPPEEgoOD9c5JSkqSd0Cee+45PPXUUygpKUF+fj6io6Nx+vTpBudx45mMG/Hmm2/qtTl9+jQ2bNiAGTNmNHpdRERERER3S4H6j5kikp05W4tde36Xj1+c2MmMsyEiIiKi5kitVsPPz89g3Z9+J4OIiIiIiO6vP/1H2NKd05bm4MWJhrNSIiIiIqLGcCeDiIiIiIhMikkGERERERGZFJMMIiIiIiIyKT6TQXp6evoiZV2lfPz0ZH66FBERERHdPu5kEBERERGRSTHJICIiIiIik2KSQUREREREJtXikoyamhqjdS+99BLy8vKQm5sLjUaDsLAwAEBCQgICAgIa7GfWrFm4dOkS7O3t5bIhQ4bg999/hyRJKCoqQmxsrM45zs7OqK2txbRp03TKXVxcsGHDBpSUlCAzMxM7duxA7969IUkSJElCdXU1ysrKIEkSUlJSoFQqcfHiRblekiS89NJLAACtVovc3Fzk5OQgNTUVXbt2vfMXjYiIiIjoDomWFDU1NQbLR40aJbKysoSrq6sAIGxsbMRrr70mAIiEhAQREBDQYD/p6ekiLS1NTJkyRS4bMmSI2LZtmwAg7OzsRFFRkfD395frp0+fLtLS0kRqaqpOXwcOHBDTpk2Tj3v37i0GDRokH986H6VSKfLy8gxel1arFc7OzgKAWLhwoVi5cmWjr9H5M7UiZV2lHOZeMwaDwWAwGAxG8wu1Wm20rsXtZBjzzjvvYO7cuTh16hQAoLa2FqtWrbqtcx9++GG0bdsWERERCA4ONtjm8uXLyM7Ohpubm1wWHByMsLAwuLm5yeVDhw7F1atXsWLFCrldbm4ufvnll7u9NNnBgwd1xiciIiIiagpMMv4/Hx8fZGVlGa2PjY3VuR3pZkFBQUhMTMS+ffvg5eWFTp30P/LV0dERPXr0QFpaGgDA3d0drq6uUKvV2LRpEwIDA29rHsZ069ZNZ36DBg3SazNq1Chs3brV4PmhoaFQq9VQq9WwtuWvBRERERHdPf41eZvCw8OhUqnkuFlwcDASExMhhEBSUhImTZok1w0ePBjZ2dk4efIkdu/ejYqKCgBAYGAgNm3aBABITEw0ugNyu0pLS3Xmd/POx969e1FeXo7Ro0djw4YNBs+Pj4+Hn58f/Pz8cPXK9XuaCxERERG1bC02yYiKitLZlSgoKEDfvn3vuB8fHx/06NEDKSkp0Gq1CAoK0kkY9u3bh8ceewze3t6YOnUqfH19AdQnJlOmTIFWq8X333+P3r17o3v37nc9j4YMHToUSqUS2dnZWLRokUn7JiIiIiK6VYtNMiIiInR2JaKjoxEbGwsXFxcAgLW1NaZOndpoP8HBwVi4cCE8PT3h6ekJNzc3PPTQQ3qf4vTrr78iJiYG8+fPR48ePdC2bVu4u7vL50VHRyM4OBh79uyBra0tQkND5XMfffRRg7c/3Ym6ujrMnj0bL7/8MpycnO6pLyIiIiKihrTYJONWO3fuxLJly/DTTz8hPz8fGo1G5+NojQkKCkJycrJOWXJyMoKCgvTafvnll3jiiScQHBysd05SUpK8A/Lcc8/hqaeeQklJCfLz8xEdHY3Tp083OI9bn8l488039dqcPn0aGzZswIwZMxq9LiIiIiKiu6VA/cdMEcnOn6nFod2/y8dPT9Z/kJ2IiIiIWja1Wg0/Pz+DddzJICIiIiIik7Iy9wSo+TmqzcHTkw1npUREREREjeFOBhERERERmRSTDCIiIiIiMikmGUREREREZFJMMkjPI0pfaOIroImvMPdUiIiIiOgBxCSDiIiIiIhMikkGERERERGZFJMMIiIiIiIyKSYZTezatWuQJAnZ2dnIysrCgAEDDLbr2bMn9u7dC0mSUFhYiBUrVsh1AwcOREZGBoqKilBUVITQ0FC5LjIyEhcuXEDHjh3lspqaGvnnTp06Yd26dSgtLUVmZiYOHDiA8ePHN8GVEhERERH9j2A0XdTU1Mg/jxgxQqSmphpst2vXLvHss8/Kxz4+PgKAcHFxEceOHRMqlUoAEM7OziIzM1OMGTNGABCRkZHi2LFjIiYmxuCYBw4cENOmTZOPu3btKmbOnNngnC9U1gpNfIXQxFeY/fVjMBgMBoPBYDTPUKvVRuu4k3Ef2dvb49y5cwbrXF1dUV5eLh/n5+cDAGbMmIG1a9dCkiQAQHV1NebNm4cFCxbIbdesWYPAwEA4OTnp9Dls2DDU1tbq7IocP34cy5YtM9k1ERERERHdiklGE2vVqhUkSUJRURFWrVqF999/32C7uLg47NmzBz/88ANmz54NBwcHAIC3tzeysrJ02mZmZsLb21s+/uOPP7BmzRrMmjVLp523tzc0Gs1tzTM0NBRqtRpqtRqWdvy1ICIiIqK7x78mm9ilS5egUqnQq1cvjBo1Cl9//bXBdmvXrkWvXr2wefNmPPnkk0hPT4eNjc1tj/P5558jJCQEbdu2Ndpm2bJlyM7OxqFDh/Tq4uPj4efnBz8/P9Rdvn7b4xIRERER3YpJxn2Unp6ODh06oGPHjoiKioIkSfJtUABw6tQpJCQkYPz48bh27Rp8fHxQWFiIvn376vTTt29fFBQU6JSdP38e69evx4wZM+SygoIC9OnTRz6eOXMmhg8frvOQOBERERGRqTHJuI+8vLxgaWmJ6upqREREQKVSQaVSAQBGjhwJKysrAICLiwucnZ1x8uRJLF++HFOmTIGvry8AoH379li8eDE++ugjvf4//fRTTJs2Te5nz549sLOzw/Tp0+U2rVu3burLJCIiIqIWzsrcE/izu/FMBgAoFAqEhITg+nX925FGjBiBzz77DJcvXwYAhIeHo6KiAgDw4osvIj4+Hu3atYNCocCSJUuwfft2vT6qq6uRnJyMt99+Wy4bP3484uLiMG/ePFRVVeHChQuYP39+U1wqEREREREAQIH6j5kikl2orMWR7+o/BatPqIuZZ0NEREREzZFarYafn5/BOt4uRUREREREJsXbpUhP4bEc+IUazkqJiIiIiBrDnQwiIiIiIjIpJhlERERERGRSTDKIiIiIiMikmGSQnt7uvjj+8Skc//iUuadCRERERA8gJhlERERERGRSTDKIiIiIiMikmGQQEREREZFJMcm4S9euXYMkScjOzkZWVhYGDBhgsF1kZCTKy8shSRIKCgoQFBQk1yUkJKCsrAySJEGSJOzfvx9KpRInTpyAQqHQ6UeSJPTr1w+RkZEICwszej4AVFVVwdHREQDQuXNnCCEwcOBAua/Kykq0b9/epK8HEREREdENTDLu0qVLl6BSqfDYY4/hnXfeQXR0tNG2cXFxUKlUGDduHFasWAErq/99B2J4eDhUKhVUKhUGDhyIY8eO4fjx4xg8eLDcxsvLC+3atcOhQ4f0+r71fABIT0+Xkx5/f39oNBr4+/sDAHr27Inq6mqcPXvWJK8DEREREdGtmGSYgL29Pc6dO9dou5KSEly8eBFOTk4NttuwYYPOjkdQUBASExNvez4HDhyQkwp/f3/ExcXpJB03djyIiIiIiJoCk4y71KpVK0iShKKiIqxatQrvv/9+o+eoVCoUFxejqqpKLouNjZVvd/r2228BAJs2bcL48eNhaWkJAAgMDMSGDRsM9mno/P3798tJRr9+/ZCcnIwuXboAqE8yDhw4oNdPaGgo1Go11Go1LFop9OqJiIiIiG6XVeNNyJAbt0sBQP/+/fH111/Dx8fHYNs5c+bglVdeQc+ePfHMM8/o1IWHhyMpKUmnrLKyEvn5+Rg+fDgqKipw7do1FBQUGOzb0PlqtRoqlQqtW7eGtbU1Lly4gLKyMnTr1g3+/v745JNP9PqJj49HfHw8AODKqdrbexGIiIiIiAzgToYJpKeno0OHDujYsSOioqLknYUb4uLi4OPjg4CAAKxevRq2traN9nnjlqmgoCCjuxjGXLp0CcXFxXj11Veh0WjkOY4ZMwadOnXCkSNH7uwCiYiIiIjuAJMME/Dy8oKlpSWqq6sREREhP4h9q23btiEzMxMhISGN9rllyxaMGTMGgYGBd/Q8xg0HDhzA7NmzcfDgQQDAwYMHMWvWLKSnp99xX0REREREd4JJxl268UyGJEnYuHEjQkJCcP369UbPe++99/D222/LH1F78zMVkiTB2toaAHD+/HkcPHgQFRUV0Gq1Rvszdv7+/fvRrVs3OcnQaDRwd3c3+DwGEREREZEpKQAIc0+Cmpcrp2pRsa4aANB1rquZZ0NEREREzZFarYafn5/BOu5kEBERERGRSfHTpUhPbnkO/OYazkqJiIiIiBrDnQwiIiIiIjIpJhlERERERGRSTDKIiIiIiMikmGSQHl93b5z6xPA3jBMRERERNYZJBhERERERmRSTDCIiIiIiMikmGUREREREZFJMMkykpqbGYHlkZCTKy8shSRIKCgoQFBQk1yUkJKCsrAySJEGSJOzfvx9KpRInTpyAQqHQ6UeSJPTr1w+RkZEICwszej4AVFVVwdHREQDQuXNnCCEwcOBAua/Kykq0b9/epNdPRERERHQDk4z7IC4uDiqVCuPGjcOKFStgZfW/70AMDw+HSqWCSqXCwIEDcezYMRw/fhyDBw+W23h5eaFdu3Y4dOiQXt+3ng8A6enpGDBgAADA398fGo0G/v7+AICePXuiuroaZ8+ebcpLJiIiIqIWjEnGfVRSUoKLFy/CycmpwXYbNmzQ2fEICgpCYmLibY9z4MABOanw9/dHXFycTtJxY8eDiIiIiKgpMMm4j1QqFYqLi1FVVSWXxcbGyrc7ffvttwCATZs2Yfz48bC0tAQABAYGYsOGDQb7NHT+/v375SSjX79+SE5ORpcuXQDUJxkHDhzQ6yc0NBRqtRpqtRqK1pamu2giIiIianGsGm9C92rOnDl45ZVX0LNnTzzzzDM6deHh4UhKStIpq6ysRH5+PoYPH46Kigpcu3YNBQWGv7fC0PlqtRoqlQqtW7eGtbU1Lly4gLKyMnTr1g3+/v745JNP9PqJj49HfHw8AKD21MV7uVwiIiIiauG4k2FiUVFR8s7CDXFxcfDx8UFAQABWr14NW1vbRvu5cctUUFCQ0V0MYy5duoTi4mK8+uqr0Gg0AOqf0xgzZgw6deqEI0eO3NlFERERERHdASYZJhYRESE/iH2rbdu2ITMzEyEhIY32s2XLFowZMwaBgYF39DzGDQcOHMDs2bNx8OBBAMDBgwcxa9YspKen33FfRERERER3gknGffbee+/h7bfflj+i9uZnKiRJgrW1NQDg/PnzOHjwICoqKqDVao32Z+z8/fv3o1u3bnKSodFo4O7ubvB5DCIiIiIiU1IAEOaeBDUvtacuonq9Fq5h3uaeChERERE1U2q1Gn5+fgbruJNBREREREQmxSSD9OSUF3AXg4iIiIjuGpMMIiIiIiIyKSYZRERERERkUkwyiIiIiIjIpJhkkB7fLn8x9xSIiIiI6AHGJIOIiIiIiEyKSQYREREREZkUkwwiIiIiIjIpJhl34dq1a5AkCdnZ2cjKysKAAQN06qdMmQJJkiBJEq5cuYLc3FxIkoTo6GiEhISgsrISkiShqKgIs2fP1jnX19cXQgiMHDlSp7ympgYAoFQqkZeXpzenhIQElJWVyePu378fABodj4iIiIioKQjGnUVNTY3884gRI0RqaqrRtlqtVjg7O8vHISEhYunSpQKAaN++vaiqqhLu7u5yfUxMjEhLSxNr1641OKZSqRR5eXl64yQkJIiAgAC98sbGMxS1p2sarGcwGAwGg8FgMNRqtdE67mTcI3t7e5w7d+6uzj179ixKSkrg6uoql02aNAlTpkzB008/DVtbW1NN0+h4RERERESmZmXuCTyIWrVqBUmSYGdnB1dXVwwbNuyu+unSpQvs7OyQm5sLAPD394dWq0VZWRlSU1MxduxYbNmy5bb7i42NRUREBACgoKAAL774YoPj3Sw0NBSvv/46AMCiFX8tiIiIiOjemH2r5UGLm2+X6t+/v8jPzzfa1tDtUpWVlSInJ0dcuXJFhIaGynVLly4Vr732mgAgnnnmGbF582a9Me/mdilj4xkL3i7FYDAYDAaDwWgseLtUE0pPT0eHDh3QsWNHREVFyQ9eN2Tjxo3w9fWFv78/YmJi4OLiAgsLCwQEBODdd9+FVqvF0qVLMWrUKLRt2/ae52hoPCIiIiKipsIk4x55eXnB0tIS1dXViIiIgEqlgkqluq1zs7Ky8M0332DWrFkYPnw4cnNz0bVrV3h6esLDwwNJSUl47rnnTDbXm8cjIiIiImoqvPn+Ltx4JgMAFAoFQkJCcP369bvqa/HixdBoNOjcuTOSk5N16pKSkvDGG2/gm2++0Sn38vLCiRMn5OM5c+YA0H0mAwD69etndLwPP/wQf/zxx13NmYiIiIioIQrU3zdFJKs9XQObzu3MPQ0iIiIiasbUajX8/PwM1vF2KSIiIiIiMikmGaQn58Rhc0+BiIiIiB5gTDKIiIiIiMikmGQQEREREZFJMckgIiIiIiKTYpJBREREREQmxSSDiIiIiIhMikkGERERERGZFJMMIiIiIiIyKSYZJlBTU6NzPGXKFEiSBEmScOXKFeTm5kKSJERHRyMkJASVlZWQJAlFRUWYPXu2zrm+vr4QQmDkyJEGx1AqlcjLy9ObQ0JCAsrKyuRx9+/fDwCNjkdERERE1BQE496ipqbGaJ1WqxXOzs7ycUhIiFg7c7PGAAAT4UlEQVS6dKkAINq3by+qqqqEu7u7XB8TEyPS0tLE2rVrDY6hVCpFXl6e3jgJCQkiICBAr7yx8QyFWq02+2vKYDAYDAaDwWje0dDfjNzJMKOzZ8+ipKQErq6uctmkSZMwZcoUPP3007C1tW3y8YiIiIiITI1Jhhl16dIFdnZ2yM3NBQD4+/tDq9WirKwMqampGDt27B31FxsbK98u9e233zY63s1CQ0OhVquhVqvRoUOHu7sgIiIiIiIwyTCLwMBA5OTkoKSkBP/+979x5coVAEBwcDASExMBAImJiQgODr6jfsPDw6FSqaBSqfDiiy82Ot7N4uPj4efnBz8/P5w5c+Yero6IiIiIWjomGSYUFRUl7yQ0ZOPGjfD19YW/vz9iYmLg4uICCwsLBAQE4N1334VWq8XSpUsxatQotG3b9p7nZWg8IiIiIqKmwiTDhCIiIuSdhNuRlZWFb775BrNmzcLw4cORm5uLrl27wtPTEx4eHkhKSsJzzz1nsvndPB4RERERUVNhkmFmixcvxiuvvILg4GAkJyfr1CUlJRm8ZcrLywsnTpyQY+LEiQB0n8mQJAnW1tZGxzPFDgkRERERkSEK1H/MFJFMrVbDz8/P3NMgIiIiomasob8ZuZNBREREREQmxSSDiIiIiIhMikkGERERERGZFJ/JID3//e9/ceTIEXNPg27RoUMHfodJM8R1aZ64Ls0T16V54ro0Tw/CuiiVSnTq1MlovWAwbg61Wm32OTC4Lg9KcF2aZ3BdmmdwXZpncF2aZzzo68LbpYiIiIiIyKSYZBARERERkUlZAlho7klQ86PRaMw9BTKA69I8cV2aJ65L88R1aZ64Ls3Tg7wufPCbiIiIiIhMirdLERERERGRSTHJICIiIiIik2KSQTpGjhyJw4cPo7i4GPPnzzf3dP6UVq9ejYqKCuTl5cllTk5O+PHHH3H06FH8+OOPcHR0lOs+++wzFBcXIycnByqVSi5/+eWXcfToURw9ehQvv/yyXN6nTx/k5uaiuLgYn3322f25qAecu7s79uzZg4KCAuTn5+Ott94CwHUxN1tbW2RkZCA7Oxv5+flYuHAhAMDDwwPp6ekoLi5GYmIirK2tAQA2NjZITExEcXEx0tPToVQq5b4WLFiA4uJiHD58GCNGjJDL+Z539ywsLKDRaLBt2zYAXJfmQKvVIjc3F5IkQa1WA+D7WHPg4OCAzZs3o6ioCIWFhejfv3+LWRezf44uo3mEhYWFKCkpEZ6ensLa2lpkZ2eLXr16mX1ef7YYPHiwUKlUIi8vTy5bvHixmD9/vgAg5s+fL2JiYgQAMXr0aPHDDz8IAOLxxx8X6enpAoBwcnISpaWlwsnJSTg6OorS0lLh6OgoAIiMjAzx+OOPCwDihx9+EKNGjTL7NTf36Ny5s1CpVAKAaNu2rThy5Ijo1asX16UZRJs2bQQAYWVlJdLT08Xjjz8uNm7cKAIDAwUA8cUXX4jp06cLAOKNN94QX3zxhQAgAgMDRWJiogAgevXqJbKzs4WNjY3w8PAQJSUlwsLCgu959xhz5swR69atE9u2bRMAuC7NILRarXB2dtYp4/uY+WPt2rVi6tSpAoCwtrYWDg4OLWVdzD4BRjOJ/v37i127dsnHCxYsEAsWLDD7vP6MoVQqdZKMw4cPi86dOwug/g/ew4cPCwDiyy+/FEFBQXrtgoKCxJdffimX32jXuXNnUVRUJJff2o5xe7F161bx1FNPcV2aUbRq1UpkZWWJfv36iaqqKmFpaSkA3fetXbt2if79+wsAwtLSUlRVVQlA/73sRju+5919uLm5iZ9++kkMHTpUTjK4LuYPQ0kG38fMG/b29qKsrEyvvCWsC2+XIpmbmxtOnDghH5eXl8PNzc2MM2o5XFxccPr0aQDA6dOn4eLiAsD4mjRUXl5erldOt0+pVEKlUiEjI4Pr0gxYWFhAkiRUVlYiJSUFpaWl+P3331FXVwdA97W8+fWvq6vD+fPn4ezsfMfrRY1bsmQJ5s2bh+vXrwMAnJ2duS7NgBACP/74IzIzMxEaGgqA/38xN09PT1RVVSEhIQEajQbx8fFo3bp1i1gXJhlEzZAQwtxTaJHatGmDpKQkzJ49GzU1NXr1XJf77/r161CpVHB3d0e/fv3wl7/8xdxTavHGjh2LysrKB/rz+/+sBg0ahL59+2L06NGYMWMGBg8erNeG72P3l5WVFfr06YMvvvgCffr0wYULF7BgwQK9dn/GdWGSQbKTJ0+iS5cu8rG7uztOnjxpxhm1HBUVFejcuTMAoHPnzqisrARgfE0aKnd3d9crp8ZZWVkhKSkJ69atQ3JyMgCuS3Ny/vx57N27FwMGDICjoyMsLS0B6L6WN7/+lpaWcHBwQHV19R2vFzVs4MCBePbZZ6HVapGYmIhhw4bhs88+47o0A7/99hsAoKqqCsnJyejXrx/fx8ysvLwc5eXlOHToEADgP//5D/r06dNi1sXs92wxmkdYWlqK0tJS4eHhIT9s98gjj5h9Xn/GuPWZjI8++kjnAbDFixcLAGLMmDE6D4BlZGQIoP4BsLKyMuHo6CgcHR1FWVmZcHJyEoD+A2CjR482+/U+CPHVV1+JuLg4nTKui3mjQ4cOwsHBQQAQdnZ2Ii0tTYwdO1Zs2rRJ5wHjN954QwAQf//733UeMN64caMAIB555BGdB4xLS0uFhYUF3/NMEEOGDJGfyeC6mDdat24t2rZtK/+8f/9+MXLkSL6PNYNIS0sTPXv2FABEZGSk+Oijj1rKuph9AoxmFKNHjxZHjhwRJSUl4h//+IfZ5/NnjPXr14vffvtN1NbWihMnTohXX31VtG/fXvz000/i6NGjIiUlRX7jACCWLVsmSkpKRG5urujbt69c/sorr4ji4mJRXFwspkyZIpf37dtX5OXliZKSErF06VKzX++DEAMHDhRCCJGTkyMkSRKSJInRo0dzXcwcjz76qNBoNCInJ0fk5eWJf/3rXwKA8PT0FBkZGaK4uFhs2rRJ2NjYCADC1tZWbNq0SRQXF4uMjAzh6ekp9/WPf/xDlJSUiMOHD+t88grf8+4tbk4yuC7mDU9PT5GdnS2ys7NFfn6+/Lrxfcz84evrK9RqtcjJyRHJycnC0dGxRayL4v//QEREREREZBJ8JoOIiIiIiEyKSQYREREREZkUkwwiIiIiIjIpJhlERERERGRSTDKIiIiIiMikmGQQETUDQgh8/PHH8nFYWBgiIyNN0ndCQgICAgJM0ldDJk6ciMLCQuzZs0enXKlUIjg4uMnHb8kWLVqE4cOHm3saBvn6+mL06NHmngYR3WdMMoiImoHLly9jwoQJcHZ2NvdUdNz4BufbMXXqVISGhmLYsGE65R4eHnjhhRdMPTWzUSgUDR6bQ2RkJH7++WdzT8Ogxx57DGPGjDH3NIjoPmOSQUTUDFy7dg0rV67EnDlz9Opu3YmoqakBAAwZMgSpqanYunUrSktLER0djRdeeAEZGRnIzc3Fww8/LJ/z1FNPQa1W48iRIxg7diwAwMLCAh999BEOHTqEnJwcvP7663K/aWlp+O6771BYWKg3n6CgIOTm5iIvLw8xMTEAgH/9618YNGgQVq9ejY8++kinfUxMDAYPHgxJkjB79mwolUqkpaUhKysLWVlZGDBgAID6P9aXL1+OoqIi/Pjjj9ixY4d83dHR0SgoKEBOTg5iY2P15uTk5ITk5GTk5OTg4MGDePTRRwEAbdq0wZo1a5Cbm4ucnBxMmDABADBy5EhkZWUhOzsbP/30E4D6P9TDwsLkPvPy8qBUKqFUKnH48GF89dVXyM/Px+DBg3WOu3Tpgrlz58qv48KFCwHU7+AUFhZi5cqVyM/Px+7du2FnZwcA6NatG1JSUpCdnY2srCx5rQz107p1a2zfvh3Z2dnIy8vD888/3+DviFarxcKFC5GVlYXc3Fx4eXnptTe2BjczNm6fPn2QmpqKzMxM7Nq1C507dwYA7N27FzExMcjIyMCRI0cwaNAgWFtb47333kNgYCAkScLzzz+P1q1bY/Xq1cjIyIBGo8Gzzz4LAAgJCUFSUhJ27tyJo0ePYvHixfJcDK2XsX6IqPkw+zcCMhgMRkuPmpoa0a5dO6HVaoW9vb0ICwsTkZGRAoBISEgQAQEBOm2B+m9bPnfunOjcubOwsbER5eXlYuHChQKAeOutt0RcXJx8/s6dO4VCoRDdu3cXJ06cELa2tiI0NFT885//FACEjY2NUKvVwsPDQwwZMkT88ccfwsPDQ2+erq6u4tixY6JDhw7C0tJS/Pzzz2LcuHECgNi7d6/Ot9PeiJu/FRqAaNWqlbC1tRUARPfu3YVarRYAREBAgNixY4dQKBTCxcVFnD17VgQEBIj27duLw4cPy+c7ODjojfH555+Ld999VwAQQ4cOFZIkCQAiJiZGfh0ACEdHR9GhQwdx/Phx+fpufNNuZGSkCAsLk9vm5eUJpVIplEqlqKurE48//rgAoHf89NNPixUrVggAQqFQiG3btonBgwcLpVIprl69Knx9fQUAsXHjRjF58mQBQKSnp4vx48cLoP4bsVu1amW0nwkTJoiVK1fK87K3t9e7/pt/R7RarZg5c6YAIN544w0RHx+v197YGtwchsa1srIS+/fvFx06dBAAxPPPPy9Wr14tr//HH38sgPpv7E5JSREAREhIiM63EH/wwQfy6+Dg4CCOHDkiWrduLUJCQkRpaamwt7cXtra24tdffxXu7u5G18tYP+b+b5nBYNSHFYiIqFmoqanB119/jbfeeguXLl26rXPUajVOnz4NACgtLcWPP/4IoP5f4YcOHSq327RpE4QQKCkpQVlZGf7yl79gxIgR6N27NyZOnAgAcHBwQI8ePVBbW4tDhw7h119/1RvPz88PqampOHPmDABg3bp1eOKJJ/Ddd9/d9nVaW1tj2bJleOyxx1BXV4eePXsCAAYNGoTNmzdDCIGKigrs3bsXAHD+/HlcvnwZq1evxvbt27F9+3a9PgcNGiT/S/7evXvh7OyMdu3a4amnnkJQUJDc7vfff8ff/vY3pKWlydd37ty5Rud87NgxZGRkGDweMWIERowYAUmSAABt27ZFjx49cPz4cWi1WuTk5AAAsrKy4OHhgbZt28LNzQ1bt24FAFy5cqXBfvbt24dPPvkEMTEx2L59O3755ZdG57tlyxZ5zBu7NzcztgY3y8vL0xvX29sbPj4+SElJAVB/O92pU6cMjuvh4WFwbiNGjMCzzz6LuXPnAgDs7OzQtWtXAMDPP/+M//73vwCAwsJCKJVKODk5GVwvY/0cPny40deHiJoekwwiomZkyZIl0Gg0SEhIkMuuXbsGC4v6u1sVCgVsbGzkuht/oALA9evX5ePr16/Dyup/b/FCCJ1xhBBQKBR488035cTkhiFDhuDChQumu6hbzJkzBxUVFfD19YWFhQUuX77cYPu6ujr069cPw4cPx8SJEzFz5swmecj55tcZgHxrEwC91+PmY4VCgejoaKxcuVKnjVKp1Fmfuro6tGrVyuj4xvoB6m9RGjNmDKKiovDzzz/j/fffb/BaboxbV1en83tww+2sQXFxsd64ycnJKCgogL+//12Ne+M6AwICcPToUZ3yxx9/XO/1MtZHQ/0QUfPAZzKIiJqRc+fOYdOmTZg6dapc9uuvv6Jv374AgGeffVYnybhdkyZNgkKhwMMPP4yHH34YR44cwe7du/HGG2/If8j16NEDrVu3brCfQ4cOYciQIXB2doaFhQWCg4Pxf//3fw2eU1NTg3bt2snHDg4OOHXqFIQQeOmll+Tx9+/fj4CAACgUCnTq1AlPPvkkgPrnKhwcHLBz507MmTMHvr6+emPs27cPkydPBlCfJJ05cwY1NTVISUnBjBkz5HaOjo5IT0/HE088If9Lu5OTE4D617lPnz4AAJVKBU9Pzwav64bdu3fj1VdfRZs2bQAADz30EDp27Gi0/R9//IHy8nKMGzcOAGBjY4NWrVoZ7cfV1RUXL17EunXrEBsbK8/xXhhbg5sZGvfIkSPo2LEj+vfvDwCwsrLCI4880uBYt67/7t278eabb8rHjz32WIPnG1uvO+2HiO4v7mQQETUzn3zyCWbOnCkfx8fH47vvvkN2djZ27dqFP/744477PH78OA4dOgR7e3tMnz4dV65cwapVq+Dh4QGNRgOFQoGqqiqMHz++wX5Onz6NBQsWYO/evVAoFNixYwe+//77Bs/Jzc1FXV0dsrOzsXbtWvz73/9GUlISXn75ZZ3rSUpKwvDhw1FYWIgTJ05Ao9Hg/PnzaNeuHb777jvY2dlBoVDg7bff1htj4cKFWLNmDXJycnDx4kWEhIQAAKKiorB8+XLk5eWhrq4OixYtQnJyMl5//XVs2bIFFhYWqKysxIgRI+Q55efnIyMj47b/hTwlJQW9evXCwYMHAdQnES+++CLq6uqMnvPSSy9hxYoVeO+993D16lVMmjTJaD/du3dHbGwsrl+/jqtXr+KNN964rXk1xNga3OzRRx/VG/fq1auYOHEiPv/8czg4OMDKygpLliwx+AEBN+zduxcLFiyAJEmIjo7G+++/jyVLliA3NxcWFhbQarV45plnjJ5/5swZg+t1p/0Q0f2lQP3DGURERGbXpk0bXLhwAe3bt8ehQ4cwcOBAVFRUmHtaRER0h7iTQUREzcb27dvh6OgIGxsbvP/++0wwiIgeUNzJICIiIiIik+KD30REREREZFJMMoiIiIiIyKSYZBARERERkUkxySAiIiIiIpNikkFERERERCb1/wCaBpYG4v6GCQAAAABJRU5ErkJggg==\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "YPCRkkm768sx" + }, + "source": [ + "\n", + "## Conclusion after analysis" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "BhDBB8847Dev" + }, + "source": [ + "### Vocab\n", + "1. The vocab size is 6710. This is not a large dataset.\n", + "2. Our dataset it comprised of short sentences, with average length of 10 and median length of 9, and the length ranges from 1 to 47. Only a small amount of sentences have the length greater than 30 => set the max length equal to 30 => need a lot of padding tokens.\n", + "3. 50% of the vocabulary only occur once but they could be person's names so let's keep them.\n", + "4. Year: can be replaced by a common\n", + "5. Number in text: can be replaced by a common\n", + "6. Lemmatization: *films* to *film*\n", + "7. All words are in lowercase.\n", + "8. No punctuations.\n", + "9. No informal text.\n", + "10. No abbreviation (like *'ll, can't, 2morrow*)\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "fRaEnZIe8UWy" + }, + "source": [ + "### Tags\n", + "1. Most of tags are short, with median length and average length of 2 words. The longest tag has 16 words. This could be a movie name.\n", + "2. Most sentences are about movies and ratings.\n", + "3. There are 25 classes of entities, divided to 3 categories: B Tags (Beginning of an entity), I Tag (Intermediate Entity), Or None Tag (O). Proportions of B,I,O are about 21.40%, 17.28%, 61.32% respectively.\n", + "4. Most of the tags are in the minority and O is the most common entity => need to over-sample the tags from the minority groups.\n", + "5. As under the section 2.7 Check the dataset imbalance, the percentage of sentences that only contain O tags -> 0.59% => It's small amount so we don't need to delete those sentences.\n", + "6. ALso under the section 2.7 Check the dataset imbalance, the percentage of OOV tokens in test set -> 3.53%\n", + "\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "Dtr2P3ZE3sE6" + }, + "source": [ + "\n", + "# Part 2: Pre-process the data\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "MRofcZfAUd9g" + }, + "source": [ + "\n", + "## 2.1 Stemming" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "1J6BsedjUeOf" + }, + "source": [ + "def stem_sentence(sentence):\n", + " sentence = sentence.split(' ')\n", + " stemmer = PorterStemmer()\n", + " result = [stemmer.stem(word) for word in sentence]\n", + " stemmed_sentence = ' '.join(result)\n", + " return stemmed_sentence" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "4p6nLSAePE9U" + }, + "source": [ + "\n", + "## 2.2 Lemmatization" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "reh2ym9rO2PM", + "outputId": "e58500ab-a58e-4abe-ab2c-48e905ed18d5" + }, + "source": [ + "nltk.download('punkt')\n", + "nltk.download('wordnet')\n", + "def lemmatize_sentence(sentence):\n", + " tokenization = nltk.word_tokenize(sentence)\n", + " lemmatizer = WordNetLemmatizer()\n", + " result = [lemmatizer.lemmatize(word) for word in tokenization]\n", + " lemmatized_sentence = ' '.join(result)\n", + " return lemmatized_sentence" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "[nltk_data] Downloading package punkt to /root/nltk_data...\n", + "[nltk_data] Unzipping tokenizers/punkt.zip.\n", + "[nltk_data] Downloading package wordnet to /root/nltk_data...\n", + "[nltk_data] Unzipping corpora/wordnet.zip.\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "Hm9UnKBqUeZA" + }, + "source": [ + "\n", + "## 2.3 Replacement" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "EZzhc1zBUeiv" + }, + "source": [ + "import re\n", + "\n", + "def replace(sentence, to_replace, replace_by):\n", + " replaced = sentence.replace(to_replace, replace_by)\n", + " return replaced\n", + "\n", + "def replace_num(sentence):\n", + " replaced = re.sub(r'^\\d{1,2}$', \"NUM\", sentence) # replace 1, 2 digits\n", + " replaced = re.sub(r'^\\d{4}$', \"YEAR\", replaced) # replace year\n", + " replaced = re.sub(r'^\\d{4}s$', \"YEAR\", replaced) # replace year\n", + " return replaced\n" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "vfswr-bAXwl5" + }, + "source": [ + "\n", + "## 2.4 Pre-processing pipeline" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "uvGcbsdOXw3N" + }, + "source": [ + "def apply_preproc(data_generator):\n", + " data_generator = list(map(lambda x: replace(x, \"ca n t\",\"cannot\"), data_generator))\n", + " data_generator = list(map(lambda x: replace(x, \"ll\",\"will\"), data_generator))\n", + " data_generator = list(map(lambda x: replace_num(x), data_generator))\n", + " data_generator = list(map(lambda x: lemmatize_sentence(x), data_generator))\n", + " return data_generator\n", + "\n", + "processed_sentences = apply_preproc(sentences)\n", + "processed_test_sentences = apply_preproc(test_sentences)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "WMK0gaxoOi97" + }, + "source": [ + "\n", + "## 2.5 Split to train/val datasets" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "yjPJLq1I0_CX" + }, + "source": [ + "split_ratio = 0.8\n", + "\n", + "def train_val_split(data, label, ratio, shuffle=True, random_seed=33):\n", + " length = len(data)\n", + " lines_index = [*range(length)] \n", + " # shuffle the indexes if shuffle is set to True\n", + " rnd.seed(random_seed)\n", + " if shuffle:\n", + " rnd.shuffle(lines_index)\n", + " split_point = int(length * ratio)\n", + "\n", + " train_data = []\n", + " train_label = []\n", + " val_data = []\n", + " val_label = []\n", + " for i in range(length):\n", + " if i <= split_point:\n", + " train_data.append(data[lines_index[i]])\n", + " train_label.append(label[lines_index[i]])\n", + " else:\n", + " val_data.append(data[lines_index[i]])\n", + " val_label.append(label[lines_index[i]])\n", + " return train_data, train_label, val_data, val_label\n", + "\n", + "\n", + "train_sentences, train_tags, val_sentences, val_tags = \\\n", + " train_val_split(processed_sentences, tags, split_ratio)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "3Sot8kCJltNu" + }, + "source": [ + "\n", + "## 2.6 Tokenization and Padding" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "GDfsVDq9KxgV" + }, + "source": [ + "oov_tok = \"\"\n", + "trunc_type='post'\n", + "pad_type='post'\n", + "max_length = 50\n", + "\n", + "vocab_tokenizer = Tokenizer(oov_token=oov_tok)\n", + "vocab_tokenizer.fit_on_texts(train_sentences)\n", + "\n", + "vocab = vocab_tokenizer.word_index\n", + "reverse_vocab = dict([(value, key) for (key, value) in vocab.items()])\n", + "vocab_size = len(vocab)\n", + "\n", + "\n", + "train_sequences = vocab_tokenizer.texts_to_sequences(train_sentences)\n", + "val_sequences = vocab_tokenizer.texts_to_sequences(val_sentences)\n", + "test_sequences = vocab_tokenizer.texts_to_sequences(processed_test_sentences)\n", + "\n", + "train_padded_sequences = pad_sequences(train_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "val_padded_sequences = pad_sequences(val_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "test_padded_sequences = pad_sequences(test_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "CP_DmX44K24X", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "c581b751-a8a5-438a-cdde-53305a7d09a9" + }, + "source": [ + "tag_tokenizer = Tokenizer(filters=\".\", lower=False, oov_token=oov_tok)\n", + "tag_tokenizer.fit_on_texts(train_tags)\n", + "\n", + "tag_map = tag_tokenizer.word_index\n", + "reverse_tag_map = dict([(value, key) for (key, value) in tag_map.items()])\n", + "tag_size = len(tag_map)\n", + "\n", + "train_tag_sequences = tag_tokenizer.texts_to_sequences(train_tags)\n", + "val_tag_sequences = tag_tokenizer.texts_to_sequences(val_tags)\n", + "test_tag_sequences = tag_tokenizer.texts_to_sequences(test_tags)\n", + "\n", + "\n", + "train_padded_tags = pad_sequences(train_tag_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "val_padded_tags = pad_sequences(val_tag_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "test_padded_tags = pad_sequences(test_tag_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "print(\"\\nExample of a a sentence and its tokenized, padded version\")\n", + "print(train_sentences[0])\n", + "print(train_padded_sequences[0])\n", + "print(\"\\nExample of a list of tags in a sentence and its tokenized, padded version\")\n", + "print(train_tags[0])\n", + "print(train_padded_tags[0])\n" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "\n", + "Example of a a sentence and its tokenized, padded version\n", + "what is the movie triangle\n", + "[ 5 7 2 3 1866 0 0 0 0 0 0 0 0 0\n", + " 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n", + " 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n", + " 0 0 0 0 0 0 0 0]\n", + "\n", + "Example of a list of tags in a sentence and its tokenized, padded version\n", + "O O O O B-TITLE\n", + "[2 2 2 2 9 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n", + " 0 0 0 0 0 0 0 0 0 0 0 0 0]\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "VmNBER6vh8fZ" + }, + "source": [ + "\n", + "## 2.7 Check the Imbalance in train/test dataset" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "1JkHts8fDElj", + "outputId": "594fd0fd-2ea0-4688-b005-be211a32076f" + }, + "source": [ + "#Percentage of B, I and O Tags in train dataset\n", + "def get_tag_proportion(series_tags):\n", + " '''\n", + " Input:\n", + " series_tags - pd.Series of tags\n", + " Output:\n", + " [print] - B, I and O tags' proportion\n", + " '''\n", + " tags_list=[tag for tag in series_tags]\n", + " counter=dict(Counter(tags_list))\n", + " beg = 0\n", + " inter = 0\n", + " out = 0\n", + " for key, value in counter.items():\n", + " if key.startswith(\"B\"):\n", + " beg += value\n", + " elif key.startswith(\"I\"):\n", + " inter += value\n", + " else:\n", + " out += value\n", + " total = len(tags_list)\n", + " print(\"B tags proportion = {0:.2%}\".format(round(beg/total,4)))\n", + " print(\"I tags proportion = {0:.2%}\".format(round(inter/total,4)))\n", + " print(\"O tags proportion = {0:.2%}\".format(round(out/total,4)))\n", + "\n", + "get_tag_proportion(df[\"Tag\"])\n", + "\n", + "# Percentage of sentences that only contain O tags\n", + "# If this percentage > 50% => the dataset is imbalanced => drop empty sentences\n", + "def get_empty_tag_sentence_proportion(list_tag_sequence):\n", + " '''\n", + " Input:\n", + " list_tag_sequence - list of tag sequences in train/test set\n", + " Output:\n", + " [print] - Percentage of sentences that only contain O tags\n", + " '''\n", + " count = 0\n", + " for seq in list_tag_sequence:\n", + " if sum(seq) == 2 * len(seq): # if seq contains only 2 (token for O tag)\n", + " count += 1\n", + "\n", + " \n", + " print(\"\\nPercentage of sentences that only contain O tags -> {0:.2%}\".\\\n", + " format(round(count/len(list_tag_sequence),4)))\n", + " \n", + "get_empty_tag_sentence_proportion(train_tag_sequences)\n", + "\n", + "def get_OOV_density(list_token_sequence):\n", + " '''\n", + " Input:\n", + " list_token_sequence - list of token sequences in test set\n", + " Output:\n", + " [print] - Percentage of OOV token in the test set\n", + " '''\n", + " list_token_sequence = [token for seq in list_token_sequence for token in seq]\n", + " counter=dict(Counter(list_token_sequence))\n", + " print(\"\\nPercentage of OOV tokens in test set -> {0:.2%}\".\\\n", + " format(round(counter[1]/len(list_token_sequence),4)))\n", + "\n", + "get_OOV_density(test_sequences)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "B tags proportion = 21.40%\n", + "I tags proportion = 17.28%\n", + "O tags proportion = 61.32%\n", + "\n", + "Percentage of sentences that only contain O tags -> 0.59%\n", + "\n", + "Percentage of OOV tokens in test set -> 3.54%\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "0OZsbuw2iDAp" + }, + "source": [ + "\n", + "## 2.8 One-hot encoding" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "qdsnDiIZwGaC" + }, + "source": [ + "# Ont hot encoding\n", + "train_padded_tags = np.array([to_categorical(tags, num_classes = tag_size+1) \\\n", + " for tags in train_padded_tags])\n", + "val_padded_tags = np.array([to_categorical(tags, num_classes = tag_size+1) \\\n", + " for tags in val_padded_tags])\n", + "test_padded_tags = np.array([to_categorical(tags, num_classes = tag_size+1) \\\n", + " for tags in test_padded_tags])\n", + "\n" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "84RwGepu37jO" + }, + "source": [ + "\n", + "# Part 3: Building the model\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "yX9PMBUtuJmJ" + }, + "source": [ + "\n", + "## 3.1 Glove Embedding" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "3QAO1_GehxQ1", + "outputId": "58d403c6-9566-4f90-d5e2-6e882c568a51" + }, + "source": [ + "!mkdir -p /glove_embedding\n", + "# Download data\n", + "!wget --no-check-certificate \\\n", + "http://nlp.stanford.edu/data/glove.6B.zip -O /glove_embedding/glove.6B.zip\n", + "!unzip /glove_embedding/glove.6B.zip -d /glove_embedding\n" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "--2021-06-03 08:19:24-- http://nlp.stanford.edu/data/glove.6B.zip\n", + "Resolving nlp.stanford.edu (nlp.stanford.edu)... 171.64.67.140\n", + "Connecting to nlp.stanford.edu (nlp.stanford.edu)|171.64.67.140|:80... connected.\n", + "HTTP request sent, awaiting response... 302 Found\n", + "Location: https://nlp.stanford.edu/data/glove.6B.zip [following]\n", + "--2021-06-03 08:19:24-- https://nlp.stanford.edu/data/glove.6B.zip\n", + "Connecting to nlp.stanford.edu (nlp.stanford.edu)|171.64.67.140|:443... connected.\n", + "HTTP request sent, awaiting response... 301 Moved Permanently\n", + "Location: http://downloads.cs.stanford.edu/nlp/data/glove.6B.zip [following]\n", + "--2021-06-03 08:19:25-- http://downloads.cs.stanford.edu/nlp/data/glove.6B.zip\n", + "Resolving downloads.cs.stanford.edu (downloads.cs.stanford.edu)... 171.64.64.22\n", + "Connecting to downloads.cs.stanford.edu (downloads.cs.stanford.edu)|171.64.64.22|:80... connected.\n", + "HTTP request sent, awaiting response... 200 OK\n", + "Length: 862182613 (822M) [application/zip]\n", + "Saving to: ‘/glove_embedding/glove.6B.zip’\n", + "\n", + "/glove_embedding/gl 100%[===================>] 822.24M 5.16MB/s in 2m 42s \n", + "\n", + "2021-06-03 08:22:07 (5.09 MB/s) - ‘/glove_embedding/glove.6B.zip’ saved [862182613/862182613]\n", + "\n", + "Archive: /glove_embedding/glove.6B.zip\n", + " inflating: /glove_embedding/glove.6B.50d.txt \n", + " inflating: /glove_embedding/glove.6B.100d.txt \n", + " inflating: /glove_embedding/glove.6B.200d.txt \n", + " inflating: /glove_embedding/glove.6B.300d.txt \n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "u2SIlF-NeR_o", + "outputId": "b3d92d12-4e8c-470c-e694-33e6fc5c6fbc" + }, + "source": [ + "GLOVE_DIR = \"/glove_embedding\"\n", + "embedding_dim = 300\n", + "hits = 0\n", + "misses = 0\n", + "embeddings_index = {}\n", + "\n", + "with open(os.path.join(GLOVE_DIR, 'glove.6B.300d.txt')) as f:\n", + " for line in f:\n", + " values = line.split()\n", + " word = values[0]\n", + " coefs = np.asarray(values[1:], dtype='float32')\n", + " embeddings_index[word] = coefs\n", + "\n", + "print('Found %s word vectors.' % len(embeddings_index))\n", + "\n", + "embedding_matrix = np.zeros((len(vocab) + 1, embedding_dim))\n", + "for word, i in vocab.items():\n", + " embedding_vector = embeddings_index.get(word)\n", + " if embedding_vector is not None:\n", + " # words not found in embedding index will be all-zeros.\n", + " embedding_matrix[i] = embedding_vector\n", + " hits += 1\n", + " else:\n", + " misses += 1\n", + "print(\"Converted %d words (%d misses)\" % (hits, misses))\n", + "\n", + " " + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Found 400000 word vectors.\n", + "Converted 5008 words (574 misses)\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "3h-sT21kjV2X" + }, + "source": [ + "\n", + "## 3.2 Define the model " + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "tshH4jK03oWM", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "67b6f8ca-5ce7-4a8f-d677-003aa2a2a847" + }, + "source": [ + "# Model architecture\n", + "batch_size = 32\n", + "embedding_dim = 300\n", + "max_length = 50\n", + "\n", + "def BiLSTM(vocab_size=vocab_size, tag_size=tag_size, hidden_size = 32, \n", + " embedding_dim=embedding_dim):\n", + " sequence_input = Input(shape = (max_length,))\n", + "\n", + " model = Embedding(input_dim = vocab_size+1, \n", + " output_dim = embedding_dim, \n", + " input_length = max_length, \n", + " embeddings_initializer=Constant(embedding_matrix),\n", + " trainable=False,\n", + " mask_zero = False)(sequence_input)\n", + " \n", + " model = Bidirectional(LSTM(units = hidden_size,return_sequences=True,\n", + " recurrent_dropout=0.1))(model)\n", + " \n", + " model = TimeDistributed(Dense(hidden_size, activation=\"relu\"))(model)\n", + " outputs = Dense(tag_size+1, activation='softmax')(model)\n", + " #crf = CRF(tag_size+1) # CRF layer\n", + " #outputs = crf(model)\n", + "\n", + " model = Model(inputs=sequence_input, outputs=outputs)\n", + "\n", + " model.compile(optimizer=\"RMSprop\", \n", + " loss = tf.keras.losses.categorical_crossentropy, \n", + " metrics=['accuracy'])\n", + " return model\n", + "\n", + "model = BiLSTM(vocab_size=vocab_size, tag_size=tag_size, hidden_size = 32, \\\n", + " embedding_dim=embedding_dim)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "WARNING:tensorflow:Layer lstm will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n", + "WARNING:tensorflow:Layer lstm will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n", + "WARNING:tensorflow:Layer lstm will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "sXWo98J3jgLQ" + }, + "source": [ + "\n", + "## 3.3 Callbacks" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "d2hDC-w7DGvu" + }, + "source": [ + "# Callback\n", + "class myCallback(tf.keras.callbacks.Callback):\n", + " def on_epoch_end(self, epoch, logs={}):\n", + " if(logs.get('val_accuracy')>0.95):\n", + " print(\"\\nReached 95% accuracy so cancelling training!\")\n", + " self.model.stop_training = True\n", + "\n", + "checkpointer = ModelCheckpoint(filepath = 'NER_BiLSTM.h5',\n", + " verbose = 0,\n", + " mode = 'auto',\n", + " save_best_only = True,\n", + " monitor='val_loss')\n", + "\n", + "earlystopper = EarlyStopping(monitor='val_loss', min_delta=0, patience=3, \n", + " verbose=0, mode='auto', \n", + " baseline=None, restore_best_weights=True)\n", + "\n", + "initial_learning_rate = 0.001\n", + "epochs = 10\n", + "decay = initial_learning_rate / epochs\n", + "def lr_time_based_decay(epoch, lr):\n", + " return lr * 1 / (1 + decay * epoch)\n", + "\n", + "lr_scheduler = tf.keras.callbacks.LearningRateScheduler(lr_time_based_decay, verbose=1)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "gKnkkiSS4EwI" + }, + "source": [ + "\n", + "# Part 4: Train the Model \n" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "vQ7uIYmgX3SC", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "10c23a4d-c31e-45d3-9dbb-51b322b3ef9d" + }, + "source": [ + "num_epochs = 15\n", + "history = model.fit(train_padded_sequences, train_padded_tags, \n", + " batch_size=batch_size, epochs=num_epochs, \n", + " validation_data= (val_padded_sequences, val_padded_tags),\n", + " callbacks=[checkpointer, earlystopper, lr_scheduler])\n", + "\n", + "model.summary()" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Epoch 1/15\n", + "\n", + "Epoch 00001: LearningRateScheduler reducing learning rate to 0.0010000000474974513.\n", + "245/245 [==============================] - 118s 452ms/step - loss: 0.4174 - accuracy: 0.9329 - val_loss: 0.1493 - val_accuracy: 0.9618\n", + "Epoch 2/15\n", + "\n", + "Epoch 00002: LearningRateScheduler reducing learning rate to 0.0009999000574917021.\n", + "245/245 [==============================] - 110s 449ms/step - loss: 0.1187 - accuracy: 0.9688 - val_loss: 0.1074 - val_accuracy: 0.9719\n", + "Epoch 3/15\n", + "\n", + "Epoch 00003: LearningRateScheduler reducing learning rate to 0.000999700106714659.\n", + "245/245 [==============================] - 110s 448ms/step - loss: 0.0902 - accuracy: 0.9760 - val_loss: 0.0922 - val_accuracy: 0.9759\n", + "Epoch 4/15\n", + "\n", + "Epoch 00004: LearningRateScheduler reducing learning rate to 0.0009994003415259673.\n", + "245/245 [==============================] - 110s 447ms/step - loss: 0.0762 - accuracy: 0.9794 - val_loss: 0.0825 - val_accuracy: 0.9782\n", + "Epoch 5/15\n", + "\n", + "Epoch 00005: LearningRateScheduler reducing learning rate to 0.0009990007918579775.\n", + "245/245 [==============================] - 109s 445ms/step - loss: 0.0676 - accuracy: 0.9818 - val_loss: 0.0756 - val_accuracy: 0.9802\n", + "Epoch 6/15\n", + "\n", + "Epoch 00006: LearningRateScheduler reducing learning rate to 0.0009985014876310735.\n", + "245/245 [==============================] - 109s 443ms/step - loss: 0.0614 - accuracy: 0.9836 - val_loss: 0.0713 - val_accuracy: 0.9816\n", + "Epoch 7/15\n", + "\n", + "Epoch 00007: LearningRateScheduler reducing learning rate to 0.0009979026914447063.\n", + "245/245 [==============================] - 110s 449ms/step - loss: 0.0564 - accuracy: 0.9849 - val_loss: 0.0708 - val_accuracy: 0.9817\n", + "Epoch 8/15\n", + "\n", + "Epoch 00008: LearningRateScheduler reducing learning rate to 0.0009972046657933619.\n", + "245/245 [==============================] - 110s 449ms/step - loss: 0.0523 - accuracy: 0.9861 - val_loss: 0.0684 - val_accuracy: 0.9821\n", + "Epoch 9/15\n", + "\n", + "Epoch 00009: LearningRateScheduler reducing learning rate to 0.0009964075567443476.\n", + "245/245 [==============================] - 109s 444ms/step - loss: 0.0491 - accuracy: 0.9869 - val_loss: 0.0671 - val_accuracy: 0.9825\n", + "Epoch 10/15\n", + "\n", + "Epoch 00010: LearningRateScheduler reducing learning rate to 0.0009955116266172385.\n", + "245/245 [==============================] - 108s 440ms/step - loss: 0.0461 - accuracy: 0.9876 - val_loss: 0.0676 - val_accuracy: 0.9823\n", + "Epoch 11/15\n", + "\n", + "Epoch 00011: LearningRateScheduler reducing learning rate to 0.0009945171376267872.\n", + "245/245 [==============================] - 107s 436ms/step - loss: 0.0433 - accuracy: 0.9884 - val_loss: 0.0704 - val_accuracy: 0.9814\n", + "Epoch 12/15\n", + "\n", + "Epoch 00012: LearningRateScheduler reducing learning rate to 0.000993424351882975.\n", + "245/245 [==============================] - 107s 437ms/step - loss: 0.0412 - accuracy: 0.9889 - val_loss: 0.0664 - val_accuracy: 0.9832\n", + "Epoch 13/15\n", + "\n", + "Epoch 00013: LearningRateScheduler reducing learning rate to 0.0009922336476668565.\n", + "245/245 [==============================] - 107s 435ms/step - loss: 0.0389 - accuracy: 0.9896 - val_loss: 0.0649 - val_accuracy: 0.9835\n", + "Epoch 14/15\n", + "\n", + "Epoch 00014: LearningRateScheduler reducing learning rate to 0.00099094540310837.\n", + "245/245 [==============================] - 108s 440ms/step - loss: 0.0373 - accuracy: 0.9900 - val_loss: 0.0642 - val_accuracy: 0.9838\n", + "Epoch 15/15\n", + "\n", + "Epoch 00015: LearningRateScheduler reducing learning rate to 0.0009895599961864132.\n", + "245/245 [==============================] - 109s 444ms/step - loss: 0.0352 - accuracy: 0.9905 - val_loss: 0.0668 - val_accuracy: 0.9838\n", + "Model: \"model\"\n", + "_________________________________________________________________\n", + "Layer (type) Output Shape Param # \n", + "=================================================================\n", + "input_1 (InputLayer) [(None, 50)] 0 \n", + "_________________________________________________________________\n", + "embedding (Embedding) (None, 50, 300) 1674900 \n", + "_________________________________________________________________\n", + "bidirectional (Bidirectional (None, 50, 64) 85248 \n", + "_________________________________________________________________\n", + "time_distributed (TimeDistri (None, 50, 32) 2080 \n", + "_________________________________________________________________\n", + "dense_1 (Dense) (None, 50, 27) 891 \n", + "=================================================================\n", + "Total params: 1,763,119\n", + "Trainable params: 88,219\n", + "Non-trainable params: 1,674,900\n", + "_________________________________________________________________\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "M3vHe1vaC-fG", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "6a982b63-5732-4ef2-a3b2-db581ec329d8" + }, + "source": [ + "history.history.keys()" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy', 'lr'])" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 24 + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "K-LuVEjYZ1qg", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 73 + }, + "outputId": "8c62eb5c-3ef0-4630-829d-8b52aaea2086" + }, + "source": [ + "acc = history.history['accuracy']\n", + "val_acc = history.history['val_accuracy']\n", + "loss = history.history['loss']\n", + "val_loss = history.history['val_loss']\n", + "plt.figure(figsize = (8,8))\n", + "epochs = range(1, len(acc) + 1)\n", + "plt.plot(epochs, acc, 'wo', label='Training acc')\n", + "plt.plot(epochs, val_acc, 'w', label='Validation acc')\n", + "plt.title('Training and validation accuracy')\n", + "plt.legend()" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 25 + }, + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "_orawE5havtv", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 499 + }, + "outputId": "fc17f13c-d589-45a5-aa37-17dab1091f93" + }, + "source": [ + "plt.figure(figsize = (8,8))\n", + "plt.plot(epochs, loss, 'wo', label='Training loss')\n", + "plt.plot(epochs, val_loss, 'w', label='Validation loss')\n", + "plt.title('Training and validation loss')\n", + "plt.legend()\n", + "plt.show()" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "-ygvHcHU4Nkj" + }, + "source": [ + "\n", + "# Part 5: Test the model\n" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "SYIAQDjhsHNQ" + }, + "source": [ + "# Convert from one-hot encoding (3D array) to 2D array \n", + "test_padded_tags_pred = model.predict(test_padded_sequences)\n", + "test_padded_tags_pred = np.argmax(test_padded_tags_pred, axis=-1)\n", + "test_padded_tags_true = np.argmax(test_padded_tags, axis=-1)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "aS4oaqkBE4XX" + }, + "source": [ + "# Convert the index to tag\n", + "test_tags_pred =[0]*len(test_padded_tags_pred)\n", + "for idx, row in enumerate(test_padded_tags_pred):\n", + " add = []\n", + " for i in row:\n", + " add.append(reverse_tag_map[i]) if i != 0 else add.append(\"PAD\")\n", + " test_tags_pred[idx] = add\n", + "\n", + "test_tags_true =[0]*len(test_padded_tags_true)\n", + "for idx, row in enumerate(test_padded_tags_true):\n", + " add = []\n", + " for i in row:\n", + " add.append(reverse_tag_map[i]) if i != 0 else add.append(\"PAD\")\n", + " test_tags_true[idx] = add" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "OO5Pg9r2MmV8", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "5b9d58d6-77dc-4381-8a74-3e26c8dea358" + }, + "source": [ + "print(\"Micro F1-score is : {:.1%}\".format(f1_score(test_tags_true, test_tags_pred)))\n", + "print(\"Micro Precision-score is : {:.1%}\".format(precision_score(test_tags_true, test_tags_pred)))\n", + "print(\"Micro Recall-score is : {:.1%}\".format(recall_score(test_tags_true, test_tags_pred)))\n" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.7/dist-packages/seqeval/metrics/sequence_labeling.py:171: UserWarning: PAD seems not to be NE tag.\n", + " warnings.warn('{} seems not to be NE tag.'.format(chunk))\n" + ], + "name": "stderr" + }, + { + "output_type": "stream", + "text": [ + "Micro F1-score is : 82.8%\n", + "Micro Precision-score is : 82.0%\n", + "Micro Recall-score is : 83.7%\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "znKBJcadMr9l", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "6952c03e-ac24-44e3-8cba-752c94c6f27d" + }, + "source": [ + "!pip install sklearn_crfsuite\n", + "from sklearn_crfsuite.metrics import flat_classification_report\n", + "\n", + "report = flat_classification_report(y_pred=test_tags_pred, y_true=test_tags_true)\n", + "print(report)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Requirement already satisfied: sklearn_crfsuite in /usr/local/lib/python3.7/dist-packages (0.3.6)\n", + "Requirement already satisfied: python-crfsuite>=0.8.3 in /usr/local/lib/python3.7/dist-packages (from sklearn_crfsuite) (0.9.7)\n", + "Requirement already satisfied: tqdm>=2.0 in /usr/local/lib/python3.7/dist-packages (from sklearn_crfsuite) (4.41.1)\n", + "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from sklearn_crfsuite) (1.15.0)\n", + "Requirement already satisfied: tabulate in /usr/local/lib/python3.7/dist-packages (from sklearn_crfsuite) (0.8.9)\n" + ], + "name": "stdout" + }, + { + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1272: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n", + " _warn_prf(average, modifier, msg_start, len(result))\n" + ], + "name": "stderr" + }, + { + "output_type": "stream", + "text": [ + " precision recall f1-score support\n", + "\n", + " B-ACTOR 0.89 0.92 0.91 812\n", + " B-CHARACTER 0.70 0.49 0.58 90\n", + " B-DIRECTOR 0.90 0.83 0.87 456\n", + " B-GENRE 0.91 0.94 0.92 1117\n", + " B-PLOT 0.74 0.58 0.65 491\n", + " B-RATING 0.97 0.97 0.97 500\n", + "B-RATINGS_AVERAGE 0.91 0.88 0.90 451\n", + " B-REVIEW 0.67 0.04 0.07 56\n", + " B-SONG 0.85 0.52 0.64 54\n", + " B-TITLE 0.79 0.75 0.77 562\n", + " B-TRAILER 0.84 0.90 0.87 30\n", + " B-YEAR 0.95 0.94 0.94 720\n", + " I-ACTOR 0.91 0.89 0.90 862\n", + " I-CHARACTER 0.57 0.39 0.46 75\n", + " I-DIRECTOR 0.89 0.85 0.87 496\n", + " I-GENRE 0.86 0.72 0.78 222\n", + " I-PLOT 0.75 0.40 0.52 496\n", + " I-RATING 0.97 0.87 0.92 226\n", + "I-RATINGS_AVERAGE 0.86 0.81 0.84 403\n", + " I-REVIEW 0.25 0.02 0.04 45\n", + " I-SONG 0.76 0.67 0.71 119\n", + " I-TITLE 0.74 0.81 0.77 856\n", + " I-TRAILER 0.00 0.00 0.00 8\n", + " I-YEAR 0.96 0.97 0.96 610\n", + " O 0.95 0.97 0.96 14929\n", + " PAD 1.00 1.00 1.00 97514\n", + "\n", + " accuracy 0.98 122200\n", + " macro avg 0.79 0.70 0.72 122200\n", + " weighted avg 0.98 0.98 0.98 122200\n", + "\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "HE7Q2oD2M9rL", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "d2c6e203-d751-4b46-c631-b5b4c165fdc3" + }, + "source": [ + "# At every execution model picks some random test sample from test set.\n", + "i = np.random.randint(0,test_padded_sequences.shape[0]) # choose a random number between 0 and len(X_te)b\n", + "p = model.predict(np.array([test_padded_sequences[i]]))\n", + "p = np.argmax(p, axis=-1)\n", + "true = np.argmax(test_padded_tags[i], -1)\n", + "\n", + "print(\"Sample number {} of {} (Test Set)\".format(i, test_padded_sequences.shape[0]))\n", + "# Visualization\n", + "print(\"{:20}||{:20}||{}\".format(\"Word\", \"True\", \"Pred\"))\n", + "print(60 * \"=\")\n", + "for word, tag, pred in zip(test_padded_sequences[i], true, p[0]):\n", + " if word != 0:\n", + " print(\"{:20}: {:20} {}\".format(reverse_vocab[word], reverse_tag_map[tag], reverse_tag_map[pred]))\n", + "\n" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Sample number 1175 of 2444 (Test Set)\n", + "Word ||True ||Pred\n", + "============================================================\n", + "did : O O\n", + "steven : B-DIRECTOR B-DIRECTOR\n", + "spielberg : I-DIRECTOR I-DIRECTOR\n", + "direct : O O\n", + "any : O O\n", + "horror : B-GENRE B-GENRE\n", + "movie : O O\n", + "in : O O\n", + "the : O O\n", + "1980 : B-YEAR B-YEAR\n", + "s : I-YEAR I-YEAR\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "sNtJLetx4Sb_" + }, + "source": [ + "\n", + "# Part 6: Test with your own sentence" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "V0niUrOnVskX", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 722 + }, + "outputId": "51453fb1-958b-4b99-c405-6b8e4a705b57" + }, + "source": [ + "# if this cell fails => run the 2nd time, it will work\n", + "\n", + "original_to_test = [\"is michael scofield the protagonist in prison break\", \n", + " \"what is the highest rated romantic movie in all time\"]\n", + "\n", + "to_test = apply_preproc(original_to_test)\n", + "\n", + "vocab_tokenizer.fit_on_texts(to_test)\n", + "to_test = vocab_tokenizer.texts_to_sequences(to_test)\n", + "\n", + "to_test = pad_sequences(to_test,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "to_test_tag_pred = model.predict(to_test)\n", + "to_test_tag_pred = np.argmax(to_test_tag_pred, axis=-1)\n", + "\n", + "for i, row in enumerate(to_test_tag_pred):\n", + " print(\"\\n{:20}||{}\".format(\"Word\", \"Pred\"))\n", + " print(40 * \"=\")\n", + " for j, pred in enumerate(row):\n", + " words = original_to_test[i].split(' ')\n", + " length = len(words)\n", + " if pred != 0 and j < length:\n", + " print(\"{:20}: {}\".format(words[j], reverse_tag_map[pred]))\n", + " " + ], + "execution_count": null, + "outputs": [ + { + "output_type": "error", + "ename": "InvalidArgumentError", + "evalue": "ignored", + "traceback": [ + "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", + "\u001b[0;31mInvalidArgumentError\u001b[0m Traceback (most recent call last)", + "\u001b[0;32m\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 14\u001b[0m padding=pad_type)\n\u001b[1;32m 15\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 16\u001b[0;31m \u001b[0mto_test_tag_pred\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpredict\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mto_test\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 17\u001b[0m \u001b[0mto_test_tag_pred\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margmax\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mto_test_tag_pred\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 18\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py\u001b[0m in \u001b[0;36mpredict\u001b[0;34m(self, x, batch_size, verbose, steps, callbacks, max_queue_size, workers, use_multiprocessing)\u001b[0m\n\u001b[1;32m 1725\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mstep\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mdata_handler\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msteps\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1726\u001b[0m \u001b[0mcallbacks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mon_predict_batch_begin\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mstep\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1727\u001b[0;31m \u001b[0mtmp_batch_outputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpredict_function\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0miterator\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 1728\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mdata_handler\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshould_sync\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1729\u001b[0m \u001b[0mcontext\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0masync_wait\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/def_function.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *args, **kwds)\u001b[0m\n\u001b[1;32m 887\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 888\u001b[0m \u001b[0;32mwith\u001b[0m \u001b[0mOptionalXlaContext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_jit_compile\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 889\u001b[0;31m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 890\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 891\u001b[0m \u001b[0mnew_tracing_count\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexperimental_get_tracing_count\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/def_function.py\u001b[0m in \u001b[0;36m_call\u001b[0;34m(self, *args, **kwds)\u001b[0m\n\u001b[1;32m 922\u001b[0m \u001b[0;31m# In this case we have not created variables on the first call. So we can\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 923\u001b[0m \u001b[0;31m# run the first trace but we should fail if variables are created.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 924\u001b[0;31m \u001b[0mresults\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_stateful_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 925\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_created_variables\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 926\u001b[0m raise ValueError(\"Creating variables on a non-first call to a function\"\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/function.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m 3022\u001b[0m filtered_flat_args) = self._maybe_define_function(args, kwargs)\n\u001b[1;32m 3023\u001b[0m return graph_function._call_flat(\n\u001b[0;32m-> 3024\u001b[0;31m filtered_flat_args, captured_inputs=graph_function.captured_inputs) # pylint: disable=protected-access\n\u001b[0m\u001b[1;32m 3025\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 3026\u001b[0m \u001b[0;34m@\u001b[0m\u001b[0mproperty\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/function.py\u001b[0m in \u001b[0;36m_call_flat\u001b[0;34m(self, args, captured_inputs, cancellation_manager)\u001b[0m\n\u001b[1;32m 1959\u001b[0m \u001b[0;31m# No tape is watching; skip to running the function.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1960\u001b[0m return self._build_call_outputs(self._inference_function.call(\n\u001b[0;32m-> 1961\u001b[0;31m ctx, args, cancellation_manager=cancellation_manager))\n\u001b[0m\u001b[1;32m 1962\u001b[0m forward_backward = self._select_forward_and_backward_functions(\n\u001b[1;32m 1963\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/function.py\u001b[0m in \u001b[0;36mcall\u001b[0;34m(self, ctx, args, cancellation_manager)\u001b[0m\n\u001b[1;32m 594\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 595\u001b[0m \u001b[0mattrs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mattrs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 596\u001b[0;31m ctx=ctx)\n\u001b[0m\u001b[1;32m 597\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 598\u001b[0m outputs = execute.execute_with_cancellation(\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/execute.py\u001b[0m in \u001b[0;36mquick_execute\u001b[0;34m(op_name, num_outputs, inputs, attrs, ctx, name)\u001b[0m\n\u001b[1;32m 58\u001b[0m \u001b[0mctx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mensure_initialized\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 59\u001b[0m tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,\n\u001b[0;32m---> 60\u001b[0;31m inputs, attrs, num_outputs)\n\u001b[0m\u001b[1;32m 61\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0mcore\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_NotOkStatusException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 62\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mname\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;31mInvalidArgumentError\u001b[0m: 2 root error(s) found.\n (0) Invalid argument: indices[0,2] = 5583 is not in [0, 5583)\n\t [[node model/embedding/embedding_lookup (defined at :2) ]]\n (1) Invalid argument: indices[0,2] = 5583 is not in [0, 5583)\n\t [[node model/embedding/embedding_lookup (defined at :2) ]]\n\t [[model/embedding/embedding_lookup/_6]]\n0 successful operations.\n0 derived errors ignored. [Op:__inference_predict_function_17969]\n\nErrors may have originated from an input operation.\nInput Source operations connected to node model/embedding/embedding_lookup:\n model/embedding/embedding_lookup/17449 (defined at /usr/lib/python3.7/contextlib.py:112)\n\nInput Source operations connected to node model/embedding/embedding_lookup:\n model/embedding/embedding_lookup/17449 (defined at /usr/lib/python3.7/contextlib.py:112)\n\nFunction call stack:\npredict_function -> predict_function\n" + ] + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "WZJApvzvaJcy" + }, + "source": [ + "\n", + "# Part 7: Analyse the incorrect predictions" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "fb8q_rD4aNRH" + }, + "source": [ + "def get_incorrect(y_pred, y_true, X_test):\n", + " y_pred.flatten()\n", + " y_true.flatten()\n", + " X_test.flatten()\n", + " where_incorrect = y_true != y_pred\n", + " incorrect_idxes = np.where(where_incorrect==1)[0]\n", + " incorrect_tokens = X_test[incorrect_idxes]\n", + " incorrect_tokens = dict(Counter(incorrect_tokens.flatten()))\n", + " incorrect_tags = y_true[incorrect_idxes]\n", + " incorrect_tags = dict(Counter(incorrect_tags.flatten()))\n", + " return incorrect_tokens, incorrect_tags\n", + "\n", + "\n", + "incorrect_tokens, incorrect_tags = get_incorrect(test_padded_tags_pred, \n", + " test_padded_tags_true, \n", + " test_padded_sequences) \n", + "\n", + "incorrect_tokens = sorted(incorrect_tokens.items(), key=lambda x:x[1], reverse=True)\n", + "incorrect_tags = sorted(incorrect_tags.items(), key=lambda x:x[1], reverse=True)\n", + "\n", + "print(\"{:^20}||{:^15}\".format(\"Incorrect word\", \"Frequency\"))\n", + "print(37 * \"=\")\n", + "for idx, count in incorrect_tokens[:20]:\n", + " if idx != 0:\n", + " print(\"{:20}: {:15}\".format(reverse_vocab[idx], count))\n", + "\n", + "print(\"\\n{:^20}||{:^15}\".format(\"Incorrect tag\", \"Frequency\"))\n", + "print(37 * \"=\")\n", + "for idx, count in incorrect_tags[:20]:\n", + " if idx != 0:\n", + " print(\"{:20}: {:15}\".format(reverse_tag_map[idx], count))\n" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "5rOw_59Ztzy7" + }, + "source": [ + "## Conclusion after analysis " + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "NiAN2Zy--j-_" + }, + "source": [ + "1. A lot of incorrect words are stopwords, due to their occurence in the movie titles/plots sometimes. This is the problem of token label inconsistency.\n", + "2. The values of macro-average are much lower than micro-average of Precision, Recall and F1-score, as a result of imbalanced classes.\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "N2J2LDiBRkyq" + }, + "source": [ + "## Potential improvements" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "9aBUF7SIHVhd" + }, + "source": [ + "1. To tackle imbalanced classes problem, use over-sample to gain more examples of tags from the minority groups.\n", + "2. To tackle label inconsistency, there're 3 solutions:\n", + "\n", + " * Use larger context. For example, use longer sentences, or combine 2 or more sentences that have similar/corelated meaning.\n", + " * Use CRF decoder layer.\n", + " * Use Character/Subword-level encoders like ELMO, Flair, CNN and BERT. \n", + "\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "NkJ8Go7BVVgz" + }, + "source": [ + "\n", + "# Export result to .tsv file" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "3EQxLZgzRdL9" + }, + "source": [ + "# write result to a new txt file\n", + "with open('/tmp/pred.tsv', 'wt') as out_file:\n", + " tsv_writer = csv.writer(out_file, delimiter='\\t')\n", + " test_size = len(test_padded_sequences)\n", + " for i in range(test_size):\n", + " for pred, word in zip(test_padded_tags_pred[i], test_padded_sequences[i]):\n", + " if pred != 0 and word != 0:\n", + " tsv_writer.writerow([reverse_tag_map[pred], reverse_vocab[word]])\n", + " tsv_writer.writerow([])" + ], + "execution_count": null, + "outputs": [] + } + ] +} \ No newline at end of file diff --git a/NER_trivia_BiLSTM.ipynb b/NER_trivia_BiLSTM.ipynb new file mode 100644 index 0000000..2c46a3c --- /dev/null +++ b/NER_trivia_BiLSTM.ipynb @@ -0,0 +1,2104 @@ +{ + "nbformat": 4, + "nbformat_minor": 0, + "metadata": { + "colab": { + "name": "NER_trivia_BiLSTM.ipynb", + "provenance": [], + "collapsed_sections": [], + "toc_visible": true + }, + "kernelspec": { + "name": "python3", + "display_name": "Python 3" + }, + "accelerator": "GPU" + }, + "cells": [ + { + "cell_type": "markdown", + "metadata": { + "id": "YsSB3CDt2NxF" + }, + "source": [ + "## Outline\n", + "- [Introduction](#0)\n", + " - [Import libraries](#0.1)\n", + "- [Part 1: Explore the data](#1)\n", + " - [1.1 Import the datasets](#1.1)\n", + " - [1.2 Exploratory Analysis](#1.2)\n", + " - [Conclusion after analysis](#1.3)\n", + " \n", + "- [Part 2: Pre-process the data](#2)\n", + " - [2.1 Stemming](#2.1)\n", + " - [2.2 Lemmatization](#2.2)\n", + " - [2.3 Replacement](#2.3)\n", + " - [2.4 Pre-processing pipeline](#2.4)\n", + " - [2.5 Split to train/val datasets](#2.5)\n", + " - [2.6 Tokenization and Padding](#2.6)\n", + " - [2.7 Check the Imbalance in train dataset](#2.7)\n", + " - [2.8 One-hot encoding](#2.8)\n", + "\n", + "- [Part 3: Build the model](#3)\n", + " - [3.1 Glove Embedding](#3.1)\n", + " - [3.2 Define the model](#3.2)\n", + " - [3.3 Callbacks](#3.3)\n", + " \n", + "\n", + "- [Part 4: Train the model](#4)\n", + "- [Part 5: Test the model](#5)\n", + "- [Part 6: Test with your own sentence](#6)\n", + "\n", + "- [Part 7: Analyse the incorrect predictions](#7)\n", + " - [Potential improvement](#7.1)\n", + "\n", + "- [Export result to .tsv file](#8)\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "Uj9t5Iav2nzR" + }, + "source": [ + "\n", + "# Introduction\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "pZDTBRDfi6u7" + }, + "source": [ + "\n", + "## Import libraries" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "Ntqkdg4N3HW4", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "e3960f95-492d-40c7-f57a-1d0e39c11fcd" + }, + "source": [ + "!python --version\n", + "import os\n", + "\n", + "%tensorflow_version 2.x\n", + "import tensorflow as tf\n", + "print(tf.__version__)\n", + "\n", + "# build the tokenized sentences and tags\n", + "from tensorflow.keras.preprocessing.text import Tokenizer\n", + "from tensorflow.keras.preprocessing.sequence import pad_sequences\n", + "\n", + "from tensorflow.keras.utils import to_categorical\n", + "from tensorflow.keras.initializers import Constant\n", + "from tensorflow.keras import Model\n", + "from tensorflow.keras.layers import Input, Embedding, Bidirectional, LSTM, \\\n", + "TimeDistributed, Dense, Dropout\n", + "from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping, \\\n", + "LearningRateScheduler\n", + "\n", + "import numpy as np # linear algebra\n", + "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n", + "import matplotlib.pyplot as plt\n", + "%matplotlib inline\n", + "import seaborn as sns\n", + "from collections import Counter\n", + "import random as rnd\n", + "from nltk.corpus import stopwords\n", + "import nltk\n", + "nltk.download('stopwords')\n", + "from nltk.stem import WordNetLemmatizer \n", + "from nltk.stem import PorterStemmer\n", + "\n", + "!pip install sklearn_crfsuite\n", + "from sklearn_crfsuite.metrics import flat_classification_report\n", + "!pip install seqeval\n", + "from seqeval.metrics import precision_score, recall_score, f1_score, classification_report\n", + "import csv\n" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Python 3.7.10\n", + "2.5.0\n", + "[nltk_data] Downloading package stopwords to /root/nltk_data...\n", + "[nltk_data] Unzipping corpora/stopwords.zip.\n", + "Collecting sklearn_crfsuite\n", + " Downloading https://files.pythonhosted.org/packages/25/74/5b7befa513482e6dee1f3dd68171a6c9dfc14c0eaa00f885ffeba54fe9b0/sklearn_crfsuite-0.3.6-py2.py3-none-any.whl\n", + "Requirement already satisfied: tqdm>=2.0 in /usr/local/lib/python3.7/dist-packages (from sklearn_crfsuite) (4.41.1)\n", + "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from sklearn_crfsuite) (1.15.0)\n", + "Requirement already satisfied: tabulate in /usr/local/lib/python3.7/dist-packages (from sklearn_crfsuite) (0.8.9)\n", + "Collecting python-crfsuite>=0.8.3\n", + "\u001b[?25l Downloading https://files.pythonhosted.org/packages/79/47/58f16c46506139f17de4630dbcfb877ce41a6355a1bbf3c443edb9708429/python_crfsuite-0.9.7-cp37-cp37m-manylinux1_x86_64.whl (743kB)\n", + "\u001b[K |████████████████████████████████| 747kB 8.7MB/s \n", + "\u001b[?25hInstalling collected packages: python-crfsuite, sklearn-crfsuite\n", + "Successfully installed python-crfsuite-0.9.7 sklearn-crfsuite-0.3.6\n", + "Collecting seqeval\n", + "\u001b[?25l Downloading https://files.pythonhosted.org/packages/9d/2d/233c79d5b4e5ab1dbf111242299153f3caddddbb691219f363ad55ce783d/seqeval-1.2.2.tar.gz (43kB)\n", + "\u001b[K |████████████████████████████████| 51kB 4.5MB/s \n", + "\u001b[?25hRequirement already satisfied: numpy>=1.14.0 in /usr/local/lib/python3.7/dist-packages (from seqeval) (1.19.5)\n", + "Requirement already satisfied: scikit-learn>=0.21.3 in /usr/local/lib/python3.7/dist-packages (from seqeval) (0.22.2.post1)\n", + "Requirement already satisfied: scipy>=0.17.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.21.3->seqeval) (1.4.1)\n", + "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.21.3->seqeval) (1.0.1)\n", + "Building wheels for collected packages: seqeval\n", + " Building wheel for seqeval (setup.py) ... \u001b[?25l\u001b[?25hdone\n", + " Created wheel for seqeval: filename=seqeval-1.2.2-cp37-none-any.whl size=16184 sha256=9799e8a9d85ef24cf7d61633bec3a02b9574dda85b1785d301d2f0d8a3bbc28d\n", + " Stored in directory: /root/.cache/pip/wheels/52/df/1b/45d75646c37428f7e626214704a0e35bd3cfc32eda37e59e5f\n", + "Successfully built seqeval\n", + "Installing collected packages: seqeval\n", + "Successfully installed seqeval-1.2.2\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "T58oiLSU25Za" + }, + "source": [ + "\n", + "# Part 1: Explore the data" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "WP5M3GnHxY-9" + }, + "source": [ + "\n", + "## 1.1 Import the datasets" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "iHhxVylamxCG", + "outputId": "61a675f5-464b-4cf3-caf9-1f7ece62d62a" + }, + "source": [ + "# Create new directories\n", + "!mkdir -p /data/eng\n", + "!mkdir -p /data/trivia10k13\n", + "!mkdir -p /model\n", + "\n", + "# Download data\n", + "!wget --no-check-certificate \\\n", + "https://groups.csail.mit.edu/sls/downloads/movie/engtrain.bio \\\n", + "-O /data/eng/train.tsv\n", + "\n", + "!wget --no-check-certificate \\\n", + "https://groups.csail.mit.edu/sls/downloads/movie/engtest.bio \\\n", + "-O /data/eng/test.tsv\n", + "\n", + "!wget --no-check-certificate \\\n", + "https://groups.csail.mit.edu/sls/downloads/movie/trivia10k13train.bio \\\n", + "-O /data/trivia10k13/train.tsv\n", + "\n", + "!wget --no-check-certificate \\\n", + "https://groups.csail.mit.edu/sls/downloads/movie/trivia10k13test.bio \\\n", + "-O /data/trivia10k13/test.tsv" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "--2021-06-02 17:10:01-- https://groups.csail.mit.edu/sls/downloads/movie/engtrain.bio\n", + "Resolving groups.csail.mit.edu (groups.csail.mit.edu)... 128.30.2.44\n", + "Connecting to groups.csail.mit.edu (groups.csail.mit.edu)|128.30.2.44|:443... connected.\n", + "HTTP request sent, awaiting response... 200 OK\n", + "Length: 1013492 (990K)\n", + "Saving to: ‘/data/eng/train.tsv’\n", + "\n", + "/data/eng/train.tsv 100%[===================>] 989.74K 2.23MB/s in 0.4s \n", + "\n", + "2021-06-02 17:10:02 (2.23 MB/s) - ‘/data/eng/train.tsv’ saved [1013492/1013492]\n", + "\n", + "--2021-06-02 17:10:02-- https://groups.csail.mit.edu/sls/downloads/movie/engtest.bio\n", + "Resolving groups.csail.mit.edu (groups.csail.mit.edu)... 128.30.2.44\n", + "Connecting to groups.csail.mit.edu (groups.csail.mit.edu)|128.30.2.44|:443... connected.\n", + "HTTP request sent, awaiting response... 200 OK\n", + "Length: 252636 (247K)\n", + "Saving to: ‘/data/eng/test.tsv’\n", + "\n", + "/data/eng/test.tsv 100%[===================>] 246.71K 826KB/s in 0.3s \n", + "\n", + "2021-06-02 17:10:03 (826 KB/s) - ‘/data/eng/test.tsv’ saved [252636/252636]\n", + "\n", + "--2021-06-02 17:10:03-- https://groups.csail.mit.edu/sls/downloads/movie/trivia10k13train.bio\n", + "Resolving groups.csail.mit.edu (groups.csail.mit.edu)... 128.30.2.44\n", + "Connecting to groups.csail.mit.edu (groups.csail.mit.edu)|128.30.2.44|:443... connected.\n", + "HTTP request sent, awaiting response... 200 OK\n", + "Length: 1785558 (1.7M)\n", + "Saving to: ‘/data/trivia10k13/train.tsv’\n", + "\n", + "/data/trivia10k13/t 100%[===================>] 1.70M 3.35MB/s in 0.5s \n", + "\n", + "2021-06-02 17:10:04 (3.35 MB/s) - ‘/data/trivia10k13/train.tsv’ saved [1785558/1785558]\n", + "\n", + "--2021-06-02 17:10:04-- https://groups.csail.mit.edu/sls/downloads/movie/trivia10k13test.bio\n", + "Resolving groups.csail.mit.edu (groups.csail.mit.edu)... 128.30.2.44\n", + "Connecting to groups.csail.mit.edu (groups.csail.mit.edu)|128.30.2.44|:443... connected.\n", + "HTTP request sent, awaiting response... 200 OK\n", + "Length: 438729 (428K)\n", + "Saving to: ‘/data/trivia10k13/test.tsv’\n", + "\n", + "/data/trivia10k13/t 100%[===================>] 428.45K 1.13MB/s in 0.4s \n", + "\n", + "2021-06-02 17:10:04 (1.13 MB/s) - ‘/data/trivia10k13/test.tsv’ saved [438729/438729]\n", + "\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "-D8_xEVDN-dj", + "outputId": "97f1f98b-0097-46e2-fabd-800939a857d0" + }, + "source": [ + "def get_sentence(file_path):\n", + " '''\n", + " Input:\n", + " file_path - path to the tsv file\n", + " Output:\n", + " sentences - list of sentences in string format\n", + " tags - list associated tags in string format\n", + " '''\n", + " sentences = []\n", + " tags = []\n", + " with open(file_path) as f:\n", + " contents = f.read()\n", + " sens_tags = contents.split(\"\\n\\n\")\n", + " for sen_tag in sens_tags:\n", + " words_tags = sen_tag.split(\"\\n\")\n", + " while (\"\" in words_tags):\n", + " words_tags.remove(\"\")\n", + " sen = ' '.join([word_tag.split(\"\\t\")[1] for word_tag in words_tags])\n", + " tag = ' '.join([word_tag.split(\"\\t\")[0] for word_tag in words_tags])\n", + " sentences.append(sen)\n", + " tags.append(tag)\n", + "\n", + " return sentences, tags\n", + "\n", + "\n", + "train_path = \"/data/trivia10k13/train.tsv\"\n", + "test_path = \"/data/trivia10k13/test.tsv\"\n", + "\n", + "sentences, tags = get_sentence(train_path)\n", + "test_sentences, test_tags = get_sentence(test_path)\n", + "\n", + "print(\"The train dataset has {} sentences.\".format(len(sentences)))\n", + "print(\"The test dataset has {} sentences.\".format(len(test_sentences)))" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "The train dataset has 7817 sentences.\n", + "The test dataset has 1954 sentences.\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "SJyYqFkQNCoW" + }, + "source": [ + "\n", + "## 1.2 Exploratory Analysis" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 198 + }, + "id": "Q6KzO6mq8326", + "outputId": "1e12ffca-e377-4ad9-a60c-a5aa2df4a1bb" + }, + "source": [ + "# Take a look at the data\n", + "df = pd.read_csv(train_path, delimiter=\"\\t\", names=[\"Tag\", \"Word\"])\n", + "df.head()" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
TagWord
0B-Actorsteve
1I-Actormcqueen
2Oprovided
3Oa
4B-Plotthrilling
\n", + "
" + ], + "text/plain": [ + " Tag Word\n", + "0 B-Actor steve\n", + "1 I-Actor mcqueen\n", + "2 O provided\n", + "3 O a\n", + "4 B-Plot thrilling" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 4 + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "tBVUmUQtFRJ1" + }, + "source": [ + "\n", + "### 1.2.1 Sentence Length \n" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 584 + }, + "id": "h2J38zVa-wkD", + "outputId": "c4f9ef1d-a3ab-46e5-9ba3-380b9f3f4495" + }, + "source": [ + "plt.style.use(\"dark_background\")\n", + "# How long are the sentences?\n", + "def plot_sentence_length_histogram(list_sentences):\n", + " '''\n", + " Input:\n", + " list_sentences - a list of sentences\n", + " Output:\n", + " [print] - Min, Max, Median and Average value of sentence length\n", + " [plot] - Histogram plot of sentence length\n", + " '''\n", + " lengths = [len(sen.split(' ')) for sen in list_sentences]\n", + " a4_dims = (11.7, 8.27)\n", + " fig, ax = plt.subplots(figsize=a4_dims)\n", + " sns.histplot(lengths)\n", + " plt.xlabel(\"Number of tokens in a sentence\")\n", + " plt.ylabel(\"Number of occurrences\")\n", + " print(\"Min: \",np.min(lengths))\n", + " print(\"Max: \",np.max(lengths))\n", + " \n", + " print(\"Median: \",np.median(lengths))\n", + " print(\"Average: \",round(np.mean(lengths),2))\n", + "\n", + "plot_sentence_length_histogram(sentences)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Min: 1\n", + "Max: 71\n", + "Median: 19.0\n", + "Average: 20.32\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "qY8VCLDQhCbJ" + }, + "source": [ + "### 1.2.2 Entity Length" + ] + }, + { + "cell_type": "code", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 584 + }, + "id": "XQrFzlrkhCm3", + "outputId": "592c48e6-e95d-40d0-ee98-5f0b3fd25b3a" + }, + "source": [ + "# How long are the entities?\n", + "def plot_entity_length_histogram(series):\n", + " '''\n", + " Input:\n", + " series - a pandas series of the tags\n", + " Output:\n", + " [print] - Min, Max, Median and Average value of entity length\n", + " [plot] - Histogram plot of entity length\n", + " '''\n", + " tags_list=[tag for tag in series]\n", + " tag_length = []\n", + " current_length = 0\n", + " for tag in tags_list:\n", + " if tag.startswith(\"B\"):\n", + " tag_length.append(current_length)\n", + " current_length = 1\n", + " elif tag.startswith(\"I\"):\n", + " current_length += 1\n", + " tag_length = tag_length[1:]\n", + " \n", + " a4_dims = (11.7, 8.27)\n", + " fig, ax = plt.subplots(figsize=a4_dims)\n", + " sns.histplot(tag_length)\n", + " plt.xlabel(\"Number of tokens in a tag\")\n", + " plt.ylabel(\"Number of occurrences\")\n", + " print(\"Min: \",np.min(tag_length))\n", + " print(\"Max: \",np.max(tag_length))\n", + " print(\"Median: \",np.median(tag_length))\n", + " print(\"Average: \",round(np.mean(tag_length),2))\n", + "\n", + "plot_entity_length_histogram(df[\"Tag\"])" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Min: 1\n", + "Max: 44\n", + "Median: 2.0\n", + "Average: 4.47\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "OcoUQIfwfOAe" + }, + "source": [ + "### 1.2.3 Token frequency" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "b2kvMJJ3KvXu", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 567 + }, + "outputId": "e4a479bf-ef36-40d7-8a47-d2003e0eae9c" + }, + "source": [ + "def plot_top_non_stopwords_barchart(series, top=20, word=True):\n", + " '''\n", + " Input:\n", + " series - a pd.Series of words or tags\n", + " top - number of most common words to plot\n", + " Output:\n", + " [print] - No of distinct words in train dataset\n", + " [plot] - Barchart of most common words' occurrence\n", + " '''\n", + " stop=set(stopwords.words('english'))\n", + " value = 'words' if word == True else 'tags'\n", + " corpus=[word for word in series]\n", + " counter=Counter(corpus)\n", + " print(\"There are {} distinct {} in dataset\".format(len(dict(counter)), value))\n", + " print(dict(counter))\n", + "\n", + " most=counter.most_common()\n", + " x, y=[], []\n", + " for word,count in most:\n", + " if (word not in stop):\n", + " x.append(count)\n", + " y.append(word)\n", + " if len(x) == top:\n", + " break\n", + " a4_dims = (11.7, 8.27)\n", + " fig, ax = plt.subplots(figsize=a4_dims)\n", + " sns.barplot(x=x,y=y)\n", + " plt.xlabel(\"Number of {} occurrences in a sentence\".format(value))\n", + " plt.ylabel(\"Most common {}s\".format(value))\n", + " return dict(counter)\n", + "\n", + "word_counter = plot_top_non_stopwords_barchart(df[\"Word\"], top=40)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "There are 10987 distinct words in dataset\n", + "{'steve': 85, 'mcqueen': 12, 'provided': 6, 'a': 7798, 'thrilling': 7, 'motorcycle': 8, 'chase': 19, 'in': 3435, 'this': 1962, 'greatest': 25, 'of': 4255, 'all': 177, 'ww': 5, '2': 43, 'prison': 52, 'escape': 35, 'movies': 80, 'liza': 3, 'minnelli': 7, 'and': 4049, 'joel': 16, 'gray': 5, 'won': 100, 'oscars': 22, 'for': 781, 'their': 346, 'roles': 25, '1972': 23, 'movie': 3924, 'that': 2194, 'follows': 64, 'nightclub': 2, 'entertainers': 1, 'berlin': 8, 'as': 1373, 'the': 8372, 'nazis': 16, 'come': 49, 'to': 2536, 'power': 25, 'what': 4273, 'is': 3785, 'tom': 147, 'hanks': 81, 'julia': 30, 'roberts': 26, 'about': 1849, 'who': 1716, 'plays': 175, 'down': 87, 'on': 1130, 'his': 1169, 'luck': 5, 'average': 9, 'guy': 118, 'goes': 117, 'back': 137, 'college': 50, 'gets': 125, 'taught': 7, 'by': 1411, 'making': 22, 'fun': 21, 'macgyver': 2, 're': 20, 'enacting': 1, 'scenes': 14, 'similar': 5, 'i': 706, 'am': 259, 'thinking': 595, 'an': 981, 'animated': 459, 'film': 2298, 'based': 554, 'classic': 661, 'theodor': 2, 'geisel': 2, 'children': 124, 's': 1590, 'novel': 168, 'young': 377, 'boy': 228, 'quest': 32, 'save': 98, 'trees': 8, '1981': 21, 'feature': 66, 'starring': 1044, 'mel': 62, 'gibson': 41, 'takes': 140, 'place': 109, 'post': 35, 'apocalyptic': 27, 'world': 269, 'australia': 5, 'steven': 119, 'speilberg': 8, 'supernatural': 21, 'story': 377, 'haunted': 12, 'house': 52, 'made': 117, 'many': 64, 'filmgoers': 1, 'afraid': 7, 'clowns': 1, '1997': 21, 'there': 49, 'scene': 53, 'featuring': 146, 'muscians': 1, 'playing': 55, 'deck': 1, 'ship': 47, 'with': 1113, 'julie': 21, 'andre': 4, 'where': 724, 'she': 136, 'sings': 5, 'flies': 2, 'umbrella': 9, 'dick': 13, 'van': 10, 'dyke': 2, 'jbiebs': 1, 'does': 191, 'concerts': 1, 'was': 757, 'banksy': 2, 'star': 186, 'director': 76, 'doing': 12, 'before': 43, 'he': 468, 'created': 28, 'john': 216, 'travolta': 38, 'had': 113, 'one': 275, 'jam': 1, 'saving': 18, 'rod': 4, 'into': 237, 'woman': 224, 'chest': 5, '1959': 18, 'american': 356, 'thriller': 178, 'directed': 695, 'alfred': 39, 'hitchcock': 52, 'cary': 30, 'grant': 31, 'eva': 7, 'marie': 3, 'saint': 4, 'm': 360, '80': 48, 'comedy': 646, 'stars': 589, 'dustin': 20, 'hoffman': 23, 'terri': 3, 'garr': 4, 'character': 278, 'dresses': 5, 'up': 265, 'land': 53, 'job': 37, 'teenage': 24, 'getting': 21, 'bit': 6, 'radioactive': 2, 'arachnid': 2, 'gaining': 2, 'super': 50, 'powers': 26, 'blockbuster': 39, 'science': 116, 'fiction': 116, 'will': 79, 'smith': 22, 'tommy': 13, 'lee': 91, 'jones': 37, 'undercover': 20, 'agents': 11, 'fight': 68, 'intergalactic': 5, 'beings': 7, 'which': 776, 'little': 95, 'girl': 270, 'accidentally': 23, 'lands': 12, 'top': 22, 'witch': 24, 'witches': 10, 'sister': 32, 'vows': 2, 'revenge': 41, 'baseball': 46, 'dennis': 10, 'quaid': 7, 'true': 74, 'pitcher': 3, 'jim': 47, 'morris': 4, '1939': 31, 'adaptation': 67, '1000': 4, 'plus': 2, 'page': 2, 'book': 182, 'clark': 25, 'gable': 14, 'vivien': 6, 'leigh': 11, 'inspired': 34, 'very': 70, 'famous': 271, 'song': 47, 'beatles': 10, 'written': 128, 'paul': 83, 'mccartney': 2, 'wimpy': 1, 'main': 80, 'befriends': 29, 'fierce': 2, 'animal': 31, 'dismay': 1, 'viking': 13, 'community': 17, 'did': 126, 'billy': 38, 'bob': 18, 'thornton': 4, 'win': 43, 'oscar': 136, 'writing': 11, '1979': 24, 'musical': 164, 'loosely': 26, 'life': 297, 'career': 16, 'fosse': 3, 'while': 79, 'shooting': 5, 'pivotal': 3, 'counterculture': 2, 'newman': 22, 'said': 8, 'have': 153, 'eaten': 1, 'over': 93, 'fifty': 2, 'boiled': 1, 'eggs': 3, 'single': 17, 'afternoon': 1, '2012': 295, 'revived': 1, 'pie': 2, 'chain': 3, 'bringing': 6, 'jason': 66, 'biggs': 1, 'alyson': 1, 'hannigan': 1, 'most': 95, 'rest': 5, 'cast': 60, 'from': 625, 'original': 33, 'graphic': 11, 'actor': 71, 'mickey': 22, 'rourke': 7, 'features': 458, 'others': 24, 'such': 16, 'jessica': 16, 'alba': 4, 'carla': 1, 'gugino': 1, 'bruce': 81, 'willis': 65, 'tobey': 5, 'maguire': 6, '2000': 33, 'two': 294, 'warriors': 13, 'chasing': 9, 'stolen': 10, 'sword': 16, 'run': 30, 'skilled': 4, 'nobleman': 2, 'daughter': 52, 'leo': 9, 'dicaprio': 49, 'enters': 5, 'human': 47, 'mind': 22, 'through': 74, 'dreams': 34, 'lets': 4, 'three': 104, 'strangers': 8, 'escort': 1, 'her': 401, 'see': 31, 'strange': 26, 'gentleman': 2, 'fond': 2, 'magic': 22, 'tricks': 2, 'released': 88, '1967': 21, 'featured': 175, 'major': 19, 'portraying': 11, 'group': 207, 'enlisted': 3, 'convicts': 3, 'wwii': 24, 'led': 11, 'marvin': 4, 'award': 95, 'winning': 64, 'spielberg': 113, 'army': 24, 'captain': 28, 'mission': 26, 'rescue': 23, 'sole': 1, 'survivor': 6, 'four': 49, 'brothers': 77, 'after': 202, 'invasion': 27, 'normandy': 7, 'luke': 23, 'skywalker': 21, 'breaking': 3, 'free': 22, 'aunt': 2, 'uncle': 18, 'princess': 60, 'leia': 6, 'off': 142, 'rhyme': 2, 'couple': 73, 'go': 97, 'hill': 44, 'middle': 33, 'aged': 11, 'decides': 33, 'embark': 8, 'adventure': 113, 'bali': 2, 'im': 23, 'manhattan': 9, 'are': 217, 'suddenly': 7, 'dealing': 14, 'issue': 2, 'employment': 2, 'move': 25, 'love': 279, 'rules': 7, 'fuck': 1, 'you': 132, 'thunder': 10, 'your': 17, 'just': 29, 'gods': 13, 'farts': 1, 'seemingly': 4, 'unrelated': 1, 'characters': 82, 'being': 111, 'isolated': 9, 'tortured': 3, 'villain': 40, 'jigsaw': 6, 'name': 870, 'clown': 6, 'fish': 20, 'has': 436, 'lost': 41, 'son': 90, 'fannie': 1, 'flagg': 1, 'tandy': 2, 'kathy': 4, 'bates': 11, 'tells': 161, 'tale': 61, 'strong': 11, 'bond': 30, 'between': 69, 'depression': 16, 'era': 33, 'women': 53, 'want': 32, 'jonah': 28, 'bring': 22, 'alcoholic': 6, 'english': 19, 'rockstar': 2, 'playboy': 6, 'nuisance': 1, 'destination': 2, 'set': 193, 'time': 194, 'psychopath': 3, 'norman': 4, 'shower': 14, 'cinematic': 2, 'history': 15, 'street': 25, 'urchin': 2, 'meets': 35, 'jasmine': 3, 'city': 83, 'they': 173, 'each': 60, 'other': 100, 'but': 143, 'can': 117, 'only': 79, 'marry': 10, 'prince': 48, 'myth': 4, 'gangster': 33, 'keiser': 1, 'soze': 1, 'recounted': 2, 'different': 49, 'perspectives': 3, 'jeremy': 12, 'renner': 9, 'sergeant': 4, 'bomb': 13, 'squad': 6, 'unit': 10, 'during': 161, 'iraq': 10, 'war': 260, 'iconic': 25, 'epic': 101, 'crime': 144, 'man': 514, 'offer': 4, 'ca': 17, 'n': 92, 't': 97, 'refuse': 4, 'literary': 7, 'emily': 25, 'bronte': 2, 'named': 132, 'heathcliff': 2, 'gene': 44, 'kelly': 15, 'friends': 159, 'struggling': 13, 'find': 139, 'work': 60, 'paris': 17, 'things': 26, 'become': 67, 'more': 42, 'complicated': 3, 'when': 145, 'them': 86, 'fall': 49, 'same': 103, 'chris': 42, 'pine': 9, 'hardy': 12, 'cia': 42, 'fighting': 61, 'adventures': 33, 'both': 50, 'giants': 1, 'small': 85, 'people': 180, 'these': 23, 'came': 35, 'out': 216, '2010': 363, 'starred': 305, 'jake': 29, 'gyllenhaal': 13, 'pg': 2, '13': 2, 'disney': 259, 'unappreciated': 1, 'video': 26, 'games': 12, 'leaves': 10, 'game': 43, 'rambunctious': 1, 'rock': 59, 'spin': 15, 'forgetting': 3, 'sarah': 13, 'marshall': 15, 'quentin': 47, 'tarantino': 58, 'jewelry': 4, 'heist': 17, 'wrong': 27, 'j': 52, 'r': 31, 'tolkien': 15, 'hobbits': 4, 'wizards': 8, 'elves': 2, 'dark': 33, 'suspense': 15, 'horror': 238, 'blowing': 6, 'portrait': 3, 'mentally': 10, 'ill': 10, 'retelling': 17, 'fairytale': 8, 'falling': 28, 'at': 216, 'end': 40, '1982': 37, 'also': 69, 'james': 160, 'earl': 6, 'thulsa': 2, 'doom': 5, 'first': 207, 'box': 25, 'office': 25, 'success': 9, 'austrian': 3, 'bodybuilder': 1, 'arnold': 49, 'schwarzenegger': 46, 'taking': 21, 'involves': 76, 'priest': 6, 'old': 125, 'removing': 2, 'demon': 6, 'spirit': 8, 'theater': 13, 'experience': 10, 'equally': 2, 'loved': 6, 'hated': 1, 'pop': 8, 'icon': 3, 'canada': 3, 'some': 65, 'say': 11, 'looks': 6, 'like': 61, 'ellen': 10, 'degeneres': 3, 'filmed': 22, 'around': 150, 'chicago': 18, 'acclaimed': 15, 'leonardo': 54, 'tries': 80, 'fix': 3, '1995': 53, 'detectives': 16, 'following': 20, 'serial': 53, 'killer': 128, 'showcases': 6, 'dante': 3, 'deadly': 31, 'sins': 15, '2011': 301, 'stoner': 6, 'latest': 13, 'instalment': 2, 'series': 208, 'cho': 2, 'kal': 2, 'penn': 15, 'perfect': 15, 'christmas': 76, 'tree': 6, 'scary': 21, 'returns': 25, 'town': 99, 'kill': 51, 'everyone': 17, 'born': 9, 'day': 89, 'supposedly': 3, 'died': 5, 'sequel': 132, '1987': 28, 'stock': 1, 'broker': 6, 'seeking': 16, 'restore': 3, 'empire': 16, 'manufacturer': 3, 'initially': 3, 'inadvertently': 2, 'saves': 15, 'thousands': 8, 'jews': 20, 'certain': 14, 'death': 75, 'holocaust': 21, 'west': 27, 'brooks': 22, 'robert': 199, 'de': 62, 'niro': 48, 'boxer': 71, 'martin': 69, 'scorsese': 47, 'matt': 55, 'damon': 57, 'highly': 8, 'popular': 100, 'espionage': 2, 'trials': 6, 'effects': 10, 'civil': 42, 'rich': 38, 'southern': 16, 'family': 218, 'russell': 65, 'brand': 24, 'rocker': 5, 'aldous': 6, 'snow': 14, 'teen': 35, 'whose': 64, 'halloween': 8, 'night': 49, 'ruined': 3, 'brother': 61, 'missing': 17, 'contains': 22, 'moment': 3, 'deniro': 32, 'looking': 51, 'mirror': 8, 'asking': 7, 'me': 42, 'carell': 35, 'dated': 2, 'awhile': 1, 'learning': 14, 'how': 72, 'pick': 7, 'girls': 29, 'cyberpunk': 1, 'action': 262, 'carpenter': 11, 'kurt': 8, 'snake': 4, 'plissken': 2, '1988': 32, 'nypd': 3, 'officer': 26, 'mcclane': 5, 'jimmy': 23, 'stewart': 57, 'wholesome': 1, 'involved': 22, 'corrupt': 12, 'politicians': 3, 'often': 15, 'referred': 2, 'female': 37, 'companion': 8, 'piece': 17, 'hangover': 5, 'competing': 6, 'determine': 1, 'maid': 11, 'honor': 8, 'glass': 7, 'trophy': 1, 'race': 33, 'winners': 1, 'picture': 73, 'colin': 19, 'firth': 11, 'academy': 91, 'best': 212, 'historical': 19, 'drama': 279, 'yasujiro': 1, 'ozu': 1, 'directs': 6, 'spare': 1, 'emotionally': 2, 'devastating': 1, 'coming': 22, 'visit': 5, 'japan': 9, '1948': 11, 'judy': 20, 'garland': 21, 'fred': 20, 'astaire': 10, 'dancing': 28, 'team': 76, 'drift': 2, 'apart': 13, 'new': 179, 'partners': 6, 'seen': 6, 'exposed': 3, 'chemical': 4, 'spill': 1, 'essentially': 1, 'turning': 6, 'zombies': 15, 'inch': 2, 'living': 56, '14': 4, 'year': 58, 'home': 91, 'tiger': 17, 'get': 141, 'shipwrecked': 4, 'stranded': 6, 'boat': 23, 'models': 3, 'hamlet': 3, 'lions': 8, 'african': 24, 'sahara': 1, 'soft': 2, 'spoken': 3, 'writer': 25, 'completes': 1, 'consider': 3, 'be': 221, 'another': 45, 'title': 112, 'might': 10, 'cold': 12, 'blood': 5, 'author': 18, 'completing': 2, '90': 33, 'blew': 1, 'careers': 2, 'keanu': 33, 'reeves': 38, 'sandra': 11, 'bullock': 9, 'reindeer': 2, 'north': 15, 'pole': 9, 'teased': 3, 'because': 28, 'color': 10, 'nose': 8, 'humphrey': 30, 'bogart': 37, 'considered': 21, 'ever': 28, 'introduced': 13, 'us': 31, 'phrase': 15, 'here': 6, 'kid': 32, '2006': 24, 'awards': 29, 'documentary': 64, 'winner': 18, 'arctic': 9, 'yearly': 1, 'migration': 2, 'grossed': 1, '5': 10, 'nominees': 1, 'sisters': 17, '1800': 6, 'sigourney': 26, 'weaver': 28, 'ridley': 33, 'scott': 48, 'spawned': 22, 'sequels': 15, 'bear': 43, 'pig': 19, '1954': 24, 'stanley': 55, 'donen': 2, 'focused': 9, 'courting': 1, 'females': 2, 'related': 2, 'males': 1, 'itself': 8, 'jack': 119, 'black': 140, 'klumsy': 2, 'lovable': 11, 'miyazaki': 10, '1968': 34, 'sparked': 1, 'zombie': 19, 'craze': 4, 'rabbit': 23, 'kangaroo': 3, 'tigger': 1, 'hitmen': 3, 'gangsters': 12, 'wife': 107, 'intertwine': 2, 'throughout': 6, 'various': 14, 'parts': 12, 'played': 138, 'geena': 4, 'davis': 16, 'susan': 18, 'sarandon': 16, 'take': 84, 'thunderbird': 3, 'drew': 10, 'barrymore': 9, 'justin': 21, 'long': 49, 'relationship': 52, 'even': 22, 'though': 13, 'live': 61, 'opposite': 5, 'coasts': 1, 'grumpy': 7, 'semi': 2, 'bald': 2, 'relives': 3, 'masterpiece': 30, 'cameron': 62, 'crashing': 4, 'iceberg': 9, 'police': 51, 'investigating': 6, 'asylum': 6, 'it': 300, 'seemed': 1, 'friendly': 4, 'adult': 22, 'breakout': 5, 'role': 104, 'melissa': 4, 'mccarthy': 4, 'tomboy': 1, 'ish': 1, 'sexed': 1, 'crazed': 7, 'groom': 2, 'melville': 2, 'monomaniacal': 1, 'sea': 18, 'hunting': 14, 'giant': 45, 'white': 99, 'whale': 12, 'ang': 22, 'nominated': 31, 'foreign': 23, 'language': 17, 'christopher': 31, 'reeve': 8, 'playwright': 10, 'finds': 100, 'way': 69, 'past': 36, 'loves': 15, 'coen': 35, 'moving': 10, 'california': 18, 'producers': 4, '70': 29, 'sylvester': 30, 'stallone': 41, 'philadelphia': 6, 'hal': 15, '9000': 5, 'kubrick': 43, 'him': 178, 'arthur': 17, 'c': 29, 'clarke': 2, '1989': 11, 'close': 12, 'circle': 6, 'shirley': 17, 'maclaine': 9, 'olympia': 2, 'dukakis': 2, 'french': 49, 'jacques': 1, 'yves': 2, 'cousteau': 1, 'louis': 5, 'malle': 1, 'exploring': 1, 'underwater': 2, 'depths': 2, 'oceans': 2, 'shakespeare': 58, 'married': 45, 'kills': 24, 'chinese': 10, 'hong': 8, 'kong': 9, 'triad': 1, 'departed': 2, 'machines': 8, 'believe': 6, 'ryan': 66, 'reynolds': 28, 'test': 10, 'pilot': 17, 'given': 14, 'alien': 95, 'ring': 21, 'gross': 2, 'early': 31, 'lord': 24, 'rings': 13, 'genius': 9, 'peter': 72, 'jackson': 44, 'creates': 5, 'theme': 13, 'park': 13, 'dinosaurs': 8, 'loose': 6, 'downey': 25, 'jr': 38, 'shipped': 2, 'theaters': 5, 'under': 38, 'code': 5, 'maternity': 1, 'recent': 51, 'sience': 1, 'edgerton': 4, 'lead': 41, 'male': 31, 'astronauts': 4, 'crash': 26, 'earth': 70, 'planet': 47, 'humans': 36, 'caged': 1, 'treated': 1, 'zoo': 16, 'animals': 53, 'singleton': 9, 'ice': 14, 'cube': 4, 'cuba': 11, 'gooding': 9, 'really': 34, 'hair': 15, 'stuck': 16, 'high': 89, 'tower': 17, 'lady': 13, 'keeps': 4, 'remade': 30, 'william': 47, 'friedkin': 2, '1977': 23, 'sorcerer': 12, '1953': 13, 'depicts': 52, 'men': 88, 'trying': 157, 'drive': 6, 'truck': 3, 'load': 2, 'explosives': 2, 'across': 35, 'dangerous': 17, 'terrain': 1, 'need': 38, 'says': 19, 'or': 49, 'not': 87, 'armstrong': 2, 'half': 7, 'shots': 1, 'barbra': 14, 'streisand': 23, 'drag': 8, 'ultimate': 7, 'leading': 29, 'soap': 7, 'opera': 12, 'childrens': 5, 'beloved': 12, 'growing': 13, 'pixar': 80, 'animation': 43, 'robot': 56, 'ends': 45, 'future': 82, 'cop': 56, 'crawling': 1, 'building': 19, 'terrorists': 10, 'comes': 49, 'metal': 12, 'ultimatum': 2, 'beat': 11, 'private': 21, 'late': 22, 'nineties': 2, 'involving': 36, 'steal': 21, 'great': 47, 'lois': 2, 'wes': 26, 'anderson': 15, 'adaption': 7, 'roald': 7, 'dahl': 7, 'voices': 38, 'george': 107, 'clooney': 22, 'meryl': 18, 'streep': 19, 'hemsworth': 14, 'dan': 7, 'bradley': 16, 'danny': 25, 'trejo': 5, 'ex': 43, 'federale': 1, 'seek': 15, 'hired': 22, 'assassinate': 5, 'important': 4, 'government': 18, 'official': 2, 'invaded': 1, 'local': 13, 'teenagers': 35, 'natalie': 35, 'wood': 16, 'rita': 3, 'moreno': 3, 'broadway': 22, 'lights': 5, 'owen': 21, 'wilson': 17, 'blonde': 11, 'barbara': 15, '1920': 12, 'fanny': 4, 'brice': 4, 'preteen': 1, 'changed': 10, 'befriending': 3, 'school': 108, 'date': 9, '1998': 33, 'jeff': 27, 'bridges': 19, 'goodman': 8, 'dude': 12, 'mistaken': 8, 'millionaire': 7, 'seeks': 15, 'restitution': 1, 'rug': 2, 'nicholson': 46, 'crazy': 33, 'than': 24, 'usual': 2, 'mental': 18, 'institution': 8, 'nicholas': 36, 'cage': 57, 'rides': 8, 'fire': 9, 'head': 32, 'mad': 18, 'think': 26, 'comic': 57, 'coach': 13, 'assistant': 13, 'using': 14, 'reject': 1, 'players': 9, 'form': 8, 'dorothy': 16, 'dog': 63, 'toto': 7, 'tin': 7, 'cowardly': 2, 'lion': 31, 'scarecrow': 9, 'journey': 35, 'yellow': 20, 'brick': 12, 'road': 35, 'eccentric': 10, 'musician': 7, 'much': 19, 'anticipated': 1, 'shot': 30, 'bunch': 22, 'guys': 43, 'went': 9, 'thailand': 7, 'worst': 8, 'oliver': 21, 'megaton': 1, 'followed': 9, 'fallout': 1, 'bryan': 7, 'mills': 6, 'kidnapped': 26, 'tell': 24, 'russel': 20, 'crow': 4, 'smart': 6, 'deaf': 4, 'baby': 34, 'do': 113, 'know': 45, 'sparks': 11, 'turned': 52, 'miley': 4, 'cyrus': 4, 'fiance': 7, 'liam': 49, 'scorcese': 5, 'lamotta': 10, 'clive': 8, 'barker': 2, 'shows': 53, 'urban': 5, 'legend': 16, 'brought': 24, 'saying': 4, 'ben': 41, 'stiller': 19, 'leader': 25, 'employees': 9, 'cheated': 1, 'money': 30, 'crooked': 1, 'businessman': 15, '1951': 10, 'morley': 1, 'katharine': 8, 'hepburn': 38, 'samuel': 21, 'rose': 11, 'saye': 1, 'reporter': 12, 'keep': 27, 'remarrying': 2, 'country': 48, 'playful': 3, 'romantic': 200, 'sex': 17, 'therapist': 8, 'counsels': 1, 'put': 30, 'excitement': 3, 'marriage': 12, 'ron': 21, 'howard': 27, 'lucas': 30, 'evil': 91, 'queen': 19, 'arrests': 1, 'pregnant': 12, 'montgomery': 1, 'clift': 1, 'murders': 17, 'shelley': 4, 'winters': 2, 'desperately': 3, 'elizabeth': 22, 'taylor': 20, 'stevens': 2, 'holly': 17, 'hunter': 32, 'kidnappers': 2, 'dominic': 5, 'toretto': 2, 'crew': 29, 'plan': 12, 'massive': 8, 'secure': 3, 'franchise': 46, 'teenager': 25, 'york': 75, 'monster': 36, 'been': 50, 'hesitant': 1, 'reprise': 4, 'ripley': 6, 'rejected': 1, 'numerous': 7, 'offers': 3, 'fox': 27, 'studios': 10, 'any': 8, 'fearing': 3, 'would': 43, 'poorly': 1, 'sub': 2, 'par': 1, 'could': 19, 'hurt': 6, 'legacy': 2, 'however': 2, 'so': 37, 'impressed': 2, 'quality': 3, 'script': 5, 'finally': 8, 'agreed': 2, 'archeologist': 3, 'adventurer': 5, 'something': 20, 'hand': 19, 'mid': 10, 'knights': 14, 'portman': 29, 'revolves': 13, 'murder': 72, 'dakota': 4, 'give': 19, 'finding': 21, 'drug': 53, 'change': 16, 'whole': 7, 'must': 97, 'away': 36, 'ruthless': 5, 'thugs': 4, 'documents': 5, 'disfigured': 5, 'famously': 10, 'yells': 2, 'peacefully': 1, 'destroyed': 6, 'danger': 7, 'planets': 1, 'woods': 15, 'lives': 74, 'katherine': 11, 'missionary': 3, 'convinces': 4, 'drunken': 6, 'dilapidated': 1, 'steamboat': 1, 'help': 73, 'torpedo': 1, 'german': 36, '60': 14, 'cult': 50, 'romero': 9, 'attempt': 28, 'survive': 20, 'low': 6, 'brow': 2, 'jealous': 8, 'friend': 69, '2003': 20, 'search': 25, 'several': 35, 'including': 40, 'peanuts': 4, 'gang': 24, 'holiday': 47, 'season': 4, 'kristen': 35, 'bell': 18, 'josh': 12, 'duhamel': 3, 'opposites': 3, 'european': 3, 'lawman': 3, 'wild': 24, 'brings': 14, 'outlaws': 7, '1990': 58, 'savini': 1, 'remake': 62, '1984': 50, 'milos': 5, 'foreman': 3, 'f': 9, 'murray': 24, 'abraham': 6, 'louise': 3, 'fletcher': 3, '1975': 28, 'scientists': 7, 'attacked': 17, 'creature': 27, 'mimic': 1, 'thing': 4, 'squares': 2, 'against': 100, 'alan': 7, 'rickman': 2, 'kids': 47, 'play': 141, 'board': 9, 'found': 19, 'release': 11, 'trapped': 29, 'decades': 4, 'host': 7, 'dangers': 4, 'stopped': 5, 'finishing': 1, 'plot': 73, 'robin': 15, 'williams': 16, 'david': 52, 'naughton': 1, 'u': 22, 'bitten': 3, 'mythical': 11, 'beast': 16, 'backpacks': 1, 'abroad': 2, '3': 53, 'rd': 5, 'selling': 8, 'books': 18, 'british': 84, 'k': 22, 'rowling': 9, 'brad': 59, 'pitt': 57, 'anthony': 27, 'hopkins': 19, 'father': 124, 'remote': 7, 'wilderness': 8, '1900': 6, 'transported': 18, 'nature': 10, 'separated': 9, 'its': 61, 'then': 39, 'adopted': 7, 'transplanted': 1, 'mars': 9, 'vet': 4, 'discovers': 32, 'lush': 1, 'inhabited': 5, '12': 4, 'foot': 3, 'tall': 6, 'barbarians': 1, 'rated': 7, 'spoofs': 1, 'flicks': 2, 'sparkling': 1, 'heart': 19, 'throbs': 1, 'werewolves': 1, '1970': 34, 'altman': 7, 'successful': 16, 'television': 33, 'space': 52, 'latifah': 7, 'dolly': 6, 'parton': 6, 'leaders': 2, 'church': 7, 'choir': 1, 'tyler': 18, 'perry': 15, 'big': 71, 'exported': 1, 'protect': 25, 'herself': 18, 'surely': 3, '1980': 84, 'zucker': 2, 'abrahams': 1, 'if': 38, 'lacking': 1, 'edgar': 7, 'rice': 2, 'burroughs': 2, 'short': 23, 'goldie': 3, 'hawn': 3, 'bette': 14, 'midler': 5, 'diane': 15, 'keaton': 10, '1996': 27, 'divorce': 6, 'strip': 7, 'large': 27, 'dane': 2, 'girlfriend': 19, 'villianous': 1, 'boyfriends': 4, 'order': 45, 'soldier': 33, 'clint': 55, 'eastwood': 59, 'sean': 30, 'tim': 63, 'robins': 1, 'stop': 50, 'nothing': 8, 'catch': 15, 'third': 52, 'installment': 52, 'continues': 4, 'intertwined': 5, 'stories': 17, 'andy': 11, 'woody': 40, 'buzz': 12, 'jackie': 12, 'chan': 9, 'agent': 72, 'babysit': 3, 'eddie': 18, 'murphy': 18, 'doctor': 15, 'talk': 18, 'second': 63, 'trilogy': 49, 'mockumentary': 7, 'loudest': 1, 'band': 28, 'tour': 12, 'evans': 7, 'ready': 6, 'sock': 1, 'ol': 2, 'adolf': 3, 'jaw': 2, 'marvel': 29, 'those': 16, 'spotted': 1, 'dogs': 9, 'running': 27, 'senator': 3, 'sense': 2, 'justice': 5, 'clashes': 2, 'washington': 47, 'system': 5, 'ingrid': 4, 'bergman': 10, 'reunite': 7, 'africa': 16, 'interlude': 1, 'judd': 6, 'apatow': 4, 'rudd': 32, 'knocked': 5, 'zach': 21, 'galifianakis': 15, 'soon': 6, 'having': 26, 'hitch': 2, 'ride': 11, 'aspiring': 9, 'make': 73, 'birth': 13, 'child': 63, 'nuremberg': 2, 'rally': 2, 'leni': 1, 'riefenstahl': 1, '1940': 25, 'adapted': 38, '1950': 21, '000': 5, 'performances': 7, 'violent': 24, 'fincer': 1, 'edward': 26, 'norton': 20, 'mysterious': 31, 'pair': 17, 'infertile': 1, 'decide': 22, 'kidnap': 7, 'andrew': 10, 'garfield': 6, 'our': 14, 'web': 2, 'slinging': 1, 'hero': 77, 'fedora': 1, 'wearing': 14, 'professor': 21, 'max': 7, 'lumet': 3, 'al': 43, 'pacino': 37, 'few': 11, 'cops': 23, 'orphan': 13, 'train': 25, 'station': 13, 'marlon': 27, 'brando': 29, 'talking': 38, 'contender': 3, 'premise': 4, 'mermaid': 6, 'makes': 34, 'faustian': 1, 'bargain': 2, 'mean': 5, 'seahag': 1, 'meet': 20, 'rises': 4, 'avenge': 11, 'parents': 36, 'roman': 26, 'polanski': 18, 'x': 7, 'crumb': 2, 'redford': 32, 'version': 34, 'neil': 13, 'simon': 9, 'outrageous': 2, 'president': 23, 'vampires': 17, 'fantasy': 107, 'garnered': 18, 'nomination': 9, 'direcor': 1, 'allen': 40, '2001': 33, 'league': 7, '1937': 11, 'jean': 17, 'renoir': 9, 'frequently': 6, 'films': 49, 'collection': 3, 'music': 48, 'done': 5, 'tv': 33, 'show': 75, 'club': 14, 'indiana': 14, 'harrison': 47, 'ford': 64, 'track': 26, 'recover': 6, 'legendary': 19, 'biblical': 8, 'bernard': 6, 'malamud': 2, '1952': 9, 'eats': 6, 'poisoned': 2, 'apple': 5, 'fairest': 3, 'benedict': 2, 'slade': 1, 'receives': 5, 'spirits': 11, 'channing': 35, 'tatum': 32, 'adrien': 7, 'brody': 10, 'warsaw': 4, 'ghetto': 7, 'whitney': 1, 'houston': 1, 'jordin': 1, 'steinbeck': 4, 'kate': 31, 'beckinsale': 4, 'returning': 8, 'selene': 2, 'fourth': 17, 'nurse': 11, 'tends': 3, 'badly': 3, 'burned': 5, 'plane': 25, 'victim': 15, '9': 2, 'now': 19, 'shark': 43, 'benchley': 2, 'forbidden': 7, 'secretive': 2, 'cowboys': 7, 'years': 38, 'huge': 8, 'hit': 87, 'blue': 39, 'costumes': 3, '2009': 40, 'records': 3, 'happening': 1, '1942': 13, 'errol': 7, 'flynn': 5, 'portrays': 19, 'real': 72, 'boxing': 22, 'champion': 9, '1971': 22, 'ultra': 5, 'produced': 51, 'burgess': 5, '1962': 25, 'novella': 5, 'richard': 51, 'dreyfuss': 14, 'sees': 15, 'sky': 10, 'compelled': 3, 'devil': 11, 'wyoming': 1, 'tales': 4, 'bad': 48, 'recreates': 1, 'image': 6, 'abolisher': 1, 'bloodsucking': 1, 'monsters': 13, 'bus': 16, 'drivers': 1, 'stay': 17, 'above': 3, '50': 17, 'mph': 3, 'exploding': 3, 'stephen': 26, 'king': 70, 'wrongfully': 2, 'imprisoned': 7, 'yet': 10, 'patiently': 1, 'awaits': 2, 'chance': 15, 'sci': 89, 'fi': 89, 'kingsley': 5, 'indian': 10, 'lawyer': 21, 'activist': 6, 'pacifist': 3, 'prize': 6, 'fighter': 10, 'longshoreman': 4, 'struggles': 26, 'stand': 7, 'union': 13, 'bosses': 8, 'unemployed': 7, 'reputation': 5, 'difficult': 6, 'disguises': 3, 'himself': 57, 'sick': 4, 'wish': 6, 'fountain': 8, 'rome': 12, 'wedding': 15, 'steel': 7, 'accepted': 2, 'peers': 1, 'santa': 24, 'clause': 6, 'hemingway': 3, 'terrence': 2, 'military': 23, 'personnel': 3, 'frog': 3, 'magical': 47, 'spell': 17, 'marion': 2, 'crane': 3, 'room': 2, 'motel': 10, 'crowe': 42, 'mathematician': 7, 'issues': 6, 'describes': 2, 'theory': 4, 'ignoring': 1, 'con': 18, 'settle': 4, 'score': 10, 'mob': 28, 'boss': 23, 'controversial': 15, 'well': 42, 'received': 15, 'homosexuality': 1, 'wayne': 29, 'sons': 3, 'avenging': 3, 'honoring': 1, 'memory': 11, 'mother': 54, 'themselves': 27, 'side': 16, 'kidnapping': 1, 'kurosawa': 13, 'relays': 1, 'independent': 11, 'listener': 1, 'audience': 4, 'showing': 9, 'multiple': 20, 'versions': 7, 'truth': 5, 'adam': 47, 'sandler': 39, 'got': 15, 'teacher': 8, 'tracks': 4, 'clones': 1, 'cheesy': 1, 'titans': 3, 'owl': 4, '2008': 21, 'humorous': 3, 'postmodern': 1, 'genre': 15, 'neighboring': 2, 'gardens': 2, 'montague': 1, 'capulet': 1, 'preparing': 3, 'appearance': 6, 'slasher': 20, 'mask': 15, 'upon': 25, 'shatner': 2, 'likeness': 1, 'line': 40, 'smell': 2, 'napalm': 2, 'morning': 4, 'sheen': 12, 'bright': 2, 'computer': 43, 'devito': 4, 'zac': 9, 'efron': 9, 'swift': 5, 'depicting': 10, 'inhabitants': 8, 'thneedville': 1, 'craven': 19, 'defending': 4, 'invaders': 5, '1994': 32, 'uma': 14, 'thurman': 12, 'hitman': 18, 'bandits': 11, 'cher': 7, 'javier': 6, 'bardem': 6, 'roy': 8, 'scheider': 6, 'seven': 21, 'connected': 2, 'stalked': 9, 'totalitarian': 2, 'society': 16, 'neurotic': 2, 'bug': 2, 'adventurous': 7, 'archaeologist': 4, 'searches': 9, 'egypt': 1, 'ancient': 15, 'relic': 2, '1993': 34, 'mountain': 8, 'climber': 4, 'lithgow': 3, 'wizard': 36, 'look': 21, 'destroy': 20, 'charms': 1, 'contain': 1, 'soul': 9, 'magician': 9, 'alongside': 8, 'actresses': 3, 'sally': 6, 'field': 14, 'macclaine': 1, 'helping': 9, 'carrell': 13, 'adopts': 4, 'g': 6, 'robinson': 9, 'became': 20, 'rico': 3, 'bandello': 1, '1931': 12, 'emma': 30, 'stone': 46, 'gosling': 15, 'embarrassing': 1, 'impediment': 7, 'full': 22, '1957': 16, 'hardened': 1, 'general': 21, 'egged': 1, 'ambitious': 7, 'works': 29, 'fulfill': 6, 'prophecy': 1, 'spider': 6, 'castle': 8, 'weird': 11, 'indie': 4, 'screenwriter': 5, 'case': 18, 'block': 2, 'may': 16, 'next': 16, 'political': 21, 'co': 31, 'neeson': 46, 'claiming': 1, 'institutionalized': 2, 'represented': 1, 'court': 5, 'farce': 1, 'dopey': 3, 'teens': 8, 'phenomena': 1, 'twilight': 6, 'bounty': 18, 'rooster': 5, 'cogburn': 5, 'reunited': 9, 'circumstance': 1, 'loses': 17, 'childhood': 10, 'tragedy': 13, 'overshadowed': 1, 'looses': 3, 'grip': 1, 'reality': 15, 'hockey': 14, 'player': 31, 'control': 21, 'antics': 6, 'creation': 11, 'athletes': 4, 'determined': 11, 'jew': 3, 'devout': 2, 'christian': 18, '1924': 3, 'olympics': 4, 'chameleon': 10, 'winds': 4, 'dirt': 5, 'pete': 2, 'yates': 2, 'fast': 12, 'paced': 4, 'grandfather': 6, 'shakespere': 1, 'pretends': 9, 'slut': 3, 'gain': 7, 'popularity': 1, 'kevin': 81, 'bacon': 10, 'butts': 1, 'heads': 7, 'allow': 2, 'listen': 2, 'roll': 11, 'updated': 5, 'fairy': 26, 'orleans': 7, 'reptile': 1, 'again': 27, 'snyder': 7, 'armor': 2, 'clad': 3, 'owls': 2, 'battle': 42, 'guards': 5, 'row': 7, 'affected': 1, 'charges': 1, 'accused': 24, 'rape': 8, 'gift': 7, 'unusual': 9, 'scout': 12, 'floating': 1, 'newspaper': 10, 'paper': 1, 'sometimes': 3, 'touted': 1, 'falls': 98, 'turns': 32, 'called': 58, 'psychosomatically': 1, 'dumb': 4, 'blind': 6, 'becomes': 56, 'master': 14, 'pinball': 2, '1961': 15, 'pool': 5, 'tear': 6, 'jerker': 3, 'defense': 4, 'undeserved': 1, 'charge': 7, 'final': 40, 'showdown': 3, 'face': 37, 'mount': 8, 'rushmore': 2, 'overcomes': 3, 'captured': 14, 'ken': 4, '1986': 20, 'mystery': 23, 'centers': 55, 'student': 43, 'visiting': 3, 'hospital': 16, 'ear': 5, 'hometown': 4, 'lumberton': 1, 'lucy': 5, 'walker': 6, 'garbage': 1, 'pickers': 1, 'franco': 15, 'zeffirelli': 3, 'crossed': 21, 'lovers': 38, 'families': 22, 'd': 48, 'dreamworks': 27, 'sanders': 2, 'dean': 10, 'deblois': 1, 'association': 1, 'vampire': 47, 'tod': 2, 'browning': 3, 'bela': 1, 'lugosi': 1, 'farm': 18, 'bullied': 10, 'misunderstood': 1, 'ferrel': 3, 'mark': 50, 'walhberg': 1, 'goofy': 4, 'mess': 5, 'johnny': 93, 'depp': 90, 'gentle': 2, 'winona': 7, 'ryder': 9, 'burton': 36, 'favorite': 17, 'skellington': 5, 'misguided': 1, 'transform': 4, 'halloweentown': 1, 'quite': 6, 'farrelly': 3, 'update': 4, 'program': 10, 'concerns': 11, 'clumsy': 4, '6': 4, 'th': 29, 'highest': 11, 'grossing': 9, 'domesticating': 1, 'learned': 3, 'appreciate': 2, 'once': 12, 'fearsome': 1, 'ponders': 1, 'question': 2, 'slyvester': 4, 'heavyweight': 4, 'warren': 13, 'beatty': 17, 'youth': 9, 'intelligent': 12, 'h': 8, 'l': 30, 'superhero': 85, 'style': 5, 'beginning': 11, 'organizations': 4, 'supermutants': 1, 'coast': 2, 'connecticut': 1, 'holden': 2, 'vied': 1, 'affections': 3, 'audrey': 24, 'wilder': 30, 'mexican': 7, 'alfonso': 2, 'cuar': 2, 'modern': 43, 'henry': 21, 'entered': 1, 'witness': 7, 'protection': 6, 'unfulfilling': 1, 'depressing': 1, 'lifestyle': 4, 'solace': 1, 'starting': 6, 'fights': 22, 'underground': 8, 'slums': 2, 'brazil': 5, 'michael': 79, 'crichton': 1, 'masked': 13, 'horrifying': 2, 'my': 27, 'lewis': 27, 'wiig': 23, 'parody': 10, 'macguyver': 1, 'federico': 4, 'fellini': 7, 'marcello': 2, 'mastroianni': 2, 'anita': 1, 'ekberg': 1, 'biographical': 26, 'nash': 7, 'brilliant': 12, 'portrayed': 8, 'wolves': 19, 'oil': 9, 'rig': 1, 'workers': 13, 'alaskan': 3, 'setting': 12, 'september': 4, 'aftermath': 2, 'five': 16, 'chosen': 8, 'eliminate': 5, 'ones': 7, 'responsible': 6, 'fateful': 3, 'travels': 29, 'entirely': 8, 'computers': 1, 'household': 6, 'reprises': 3, 'cartoon': 39, 'hunted': 6, 'according': 5, 'lesbian': 3, 'throws': 1, 'turmoil': 2, 'exclusive': 3, 'golf': 6, 'course': 9, 'deal': 24, 'brash': 2, 'member': 16, 'destructive': 3, 'gopher': 4, 'remarkable': 1, 'narrated': 4, 'morgan': 38, 'freeman': 35, 'kubric': 6, 'ghost': 7, 'rider': 6, 'survivors': 5, 'locked': 3, 'mall': 3, 'undead': 5, '1974': 20, 'lucille': 2, 'ball': 5, 'bea': 1, '2004': 22, 'terrance': 1, 'bolluck': 1, 'focuses': 30, 'racial': 4, 'tension': 1, 'los': 15, 'angeles': 12, 'saga': 9, 'good': 36, 'western': 83, '7': 6, 'gunman': 3, 'protecting': 5, 'village': 13, 'donkey': 10, 'green': 28, 'grandmother': 2, 'wolf': 7, 'stalking': 2, 'jamie': 28, 'curtis': 26, 'babysitter': 4, 'psycho': 2, 'older': 10, 'downing': 1, 'forced': 35, 'trip': 38, 'stranger': 5, 'arrival': 3, 'ii': 49, 'parisian': 2, 'actress': 41, 'hide': 6, 'jewish': 19, 'husband': 28, 'meek': 2, 'hobbit': 4, 'shire': 2, 'eight': 3, 'companions': 3, 'sauron': 4, 'loser': 7, 'fortune': 4, 'changes': 6, 'europe': 3, 'wants': 45, 'hemworth': 1, 'titular': 20, 'orwell': 1, 'parodies': 7, 'communism': 1, 'pet': 12, 'courage': 4, 'prequel': 16, '21': 1, 'st': 7, 'century': 14, 'rat': 4, 'attempts': 15, 'beautiful': 21, 'fonda': 15, 'temple': 1, 'indians': 7, 'southwest': 1, 'francis': 19, 'mcdormand': 7, 'macy': 6, 'minnesota': 2, 'accents': 1, 'patrick': 11, 'harris': 9, 'pesky': 1, 'robbers': 6, 'stray': 6, 'sets': 24, 'japanese': 42, 'camp': 30, 'prisoners': 5, 'revealing': 3, 'tech': 3, 'runs': 17, 'gorilla': 9, 'damsel': 2, 'hostage': 10, 'climbs': 3, 'buildings': 2, 'brendan': 4, 'fraser': 4, 'rachel': 27, 'weisz': 8, 'treasure': 10, 'hamunaptra': 2, 'anime': 7, 'scientist': 17, 'dynamic': 2, 'trio': 15, 'spacey': 22, 'examination': 2, 'mans': 3, 'rivalry': 1, 'sharks': 7, 'jets': 6, 'christie': 5, 'bow': 4, 'wow': 2, 'wins': 16, 'raffle': 1, 'neighborhood': 9, 'flick': 41, 'propelled': 1, 'englund': 1, 'stardom': 4, 'deformed': 6, 'hope': 7, 'never': 25, 'dream': 26, 'superiors': 6, 'hosting': 1, 'dinner': 8, 'celebrating': 2, 'idiocy': 2, 'guests': 4, 'rising': 5, 'executive': 11, 'questions': 6, 'invited': 6, 'guest': 3, 'direction': 5, 'cubric': 1, 'claims': 5, 'acting': 8, 'fame': 7, 'scottish': 11, 'highlands': 1, 'anne': 35, 'hathaway': 25, 'learns': 29, 'heir': 5, 'throne': 7, 'genovia': 2, 'charles': 37, 'dickens': 20, 'lightyear': 7, 'agree': 1, 'accompany': 4, 'submarine': 2, 'travel': 36, 'pepperland': 2, 'hating': 2, 'meanies': 2, 'spooky': 1, 'benicio': 4, 'del': 14, 'toro': 13, 'lawrence': 25, 'talbot': 1, 'aha': 1, 'twisted': 2, 'ballet': 11, 'swan': 3, 'lake': 12, 'havoc': 5, 'arcade': 5, 'cowboy': 7, 'spy': 37, 'length': 5, 'walt': 14, 'wahlburg': 2, 'retired': 20, 'criminal': 27, 'law': 30, 'debts': 2, 'dealers': 3, 'nicolas': 28, 'knight': 7, 'charged': 5, 'task': 4, 'escorting': 1, 'monastery': 2, 'part': 39, 'vehicles': 3, 'personality': 2, 'innocent': 3, 'jail': 11, 'escapes': 12, 'always': 20, 'dreamt': 1, 'fully': 2, 'cgi': 7, 'going': 34, 'rainbow': 1, 'along': 18, 'ralph': 15, 'fiennes': 12, 'gweneth': 1, 'paltrow': 11, 'period': 12, 'fictionalization': 1, 'fishing': 1, 'held': 5, 'freak': 3, 'goldblum': 3, 'ian': 5, 'malcolm': 8, 'romeo': 9, 'juliet': 9, 'concerned': 5, 'efforts': 8, 'phone': 6, 'god': 26, 'facing': 9, 'loki': 6, 'asgard': 4, 'ingmar': 6, 'von': 7, 'sydow': 2, 'chess': 2, 'telling': 14, 'surgeons': 1, 'korean': 6, 'shares': 5, 'available': 1, 'infamous': 14, 'cruella': 2, 'glenn': 5, 'attenborough': 3, 'failed': 8, 'operation': 6, 'market': 7, 'garden': 9, 'talked': 3, 'attack': 18, 'enemy': 16, 'warship': 1, 'wwi': 4, 'bing': 6, 'crosby': 6, 'elephant': 13, 'jennifer': 55, 'aniston': 28, 'gerard': 14, 'butler': 17, 'centered': 43, 'retrieve': 9, 'cure': 2, 'mutation': 1, 'margaret': 6, 'mitchell': 7, 'surviving': 5, 'leslie': 8, 'nielsen': 6, 'jet': 10, 'everything': 7, 'food': 16, 'rains': 3, 'rain': 7, 'witnesses': 8, 'apartment': 13, 'andrews': 15, 'nanny': 9, 'flying': 13, '1965': 19, 'russian': 13, 'poet': 4, 'omar': 6, 'sharif': 5, 'geraldine': 1, 'chaplin': 10, 'shocking': 3, 'water': 22, 'filthiest': 1, 'extremely': 7, 'alike': 3, 'martha': 2, 'vineyard': 2, 'gold': 15, 'fever': 2, 'huston': 12, 'someone': 19, 'dies': 17, 'causing': 4, 'secrets': 11, 'philip': 9, 'androids': 7, 'electric': 4, 'sheep': 9, 'capturing': 2, 'plague': 7, 'soldiers': 25, 'traveling': 39, 'countryside': 3, 'toy': 7, 'bay': 11, 'americans': 8, 'prospector': 2, 'wilds': 3, 'central': 5, 'mexico': 14, '1973': 22, 'linda': 10, 'blair': 7, 'demonic': 5, 'base': 5, 'path': 4, 'closet': 2, 'helen': 8, 'hunt': 16, 'mute': 9, 'pianist': 4, 'sent': 46, 'zealand': 4, 'greek': 21, 'mythology': 5, 'possession': 7, 'catholic': 1, 'defend': 10, 'naive': 5, 'bucked': 1, 'tooth': 3, 'break': 21, 'pornography': 1, 'industry': 6, '1941': 19, 'orson': 16, 'welles': 14, 'last': 49, 'words': 7, 'tycoon': 6, 'ability': 9, 'turn': 32, 'invisible': 8, 'burt': 12, 'jon': 19, 'voight': 5, 'ned': 7, 'ronnie': 1, 'cox': 6, 'river': 13, 'hotel': 26, 'winter': 12, 'spiritual': 5, 'presence': 3, 'influences': 2, 'violence': 16, 'psychic': 3, 'horrific': 5, 'forebodings': 2, 'errant': 1, 'united': 19, 'states': 20, 'bomber': 1, 'drops': 1, 'nuclear': 16, 'warhead': 1, 'russia': 9, 'due': 13, 'schemes': 3, 'cooper': 23, 'ray': 22, 'liotta': 12, 'body': 13, 'puppetry': 1, 'performed': 1, '10': 11, 'stuntman': 4, 'kitchen': 2, 'were': 56, 'without': 18, 'legs': 4, 'expert': 6, 'walking': 3, 'hands': 17, 'eventually': 17, 'nabbed': 1, 'originally': 26, 'wrote': 20, 'mitchum': 2, 'courageous': 2, 'tracking': 7, '4': 14, 'hijackers': 1, 'staring': 41, 'kirsten': 1, 'soderbergh': 12, 'gina': 1, 'carano': 1, 'panda': 12, 'po': 4, 'voiced': 18, 'martial': 34, 'arts': 33, 'lopez': 4, 'alex': 15, 'o': 35, 'loughlin': 1, 'bosley': 1, 'carrey': 22, 'shaved': 1, 'hairline': 1, 'thought': 7, 'cerebral': 1, 'person': 25, 'childish': 1, 'eighteenth': 1, 'vienna': 7, 'composer': 11, 'drunk': 6, 'car': 34, 'threw': 2, 'andersen': 1, 'hackman': 20, 'schwartzman': 1, 'started': 7, 'menswear': 1, 'fashion': 5, 'trend': 1, 'wars': 5, 'flim': 2, 'skipped': 1, 'bail': 6, 'filled': 8, 'mercenaries': 3, 'thinks': 7, 'door': 10, 'denzel': 41, 'charm': 2, 'filmmaker': 5, 'fondly': 1, 'remembers': 2, 'boyhood': 1, 'fascination': 1, 'heath': 14, 'ledger': 14, 'stiles': 4, 'emotional': 5, 'twist': 11, 'dead': 38, 'candy': 9, 'reese': 17, 'pieces': 8, 'ensemble': 24, 'eve': 13, 'ashton': 10, 'kutcher': 9, 'mendes': 8, 'bon': 2, 'jovi': 2, 'among': 16, 'together': 58, 'leads': 23, 'rick': 4, 'deckard': 2, 'cyborg': 13, 'uttered': 1, 'phrases': 1, 'll': 8, 'dave': 7, 'lizewski': 1, 'unnoticed': 2, 'fan': 6, 'boris': 3, 'karloff': 3, '1935': 7, 'twins': 2, 'thanksgiving': 1, 'understudy': 1, 'betrayed': 8, 'margo': 3, 'hung': 1, 'lived': 9, 'louisa': 2, 'alcott': 1, '1860': 1, 'ethereal': 1, 'searching': 27, 'golden': 10, 'retriever': 2, 'rabies': 4, 'bite': 2, 'sharon': 3, 'joe': 18, 'pesci': 14, 'las': 9, 'vegas': 9, 'macbeth': 2, 'meg': 19, 'widowed': 5, 'horrible': 10, 'outlaw': 10, 'dolphin': 4, 'underlining': 1, 'eternal': 2, 'connection': 6, 'warrior': 15, 'murdered': 23, 'rapunzel': 14, 'voice': 47, 'chuck': 5, 'exacted': 1, 'counselors': 4, 'allowed': 1, 'drown': 2, 'known': 54, 'anton': 5, 'corbijn': 4, 'exercise': 2, '1922': 3, 'expressionism': 1, 'w': 5, 'murnau': 2, 'unauthorized': 2, 'bram': 5, 'stoker': 4, 'dracula': 4, 'store': 10, 'avent': 2, 'booksellers': 1, 'production': 7, 'greenwood': 1, 'ms': 4, 'allied': 5, 'pows': 3, 'evacuate': 1, 'tipping': 1, 'hakunamatata': 1, 'samurai': 7, 'point': 7, 'view': 12, 'topher': 4, 'grace': 9, 'anna': 17, 'faris': 8, 'awkward': 6, 'capture': 9, 'attention': 4, 'crush': 3, '1947': 4, 'prominently': 4, 'department': 4, 'integral': 1, 'storyline': 1, '1978': 17, 'revisited': 1, 'zack': 9, 'comedic': 7, 'nia': 3, 'vardalos': 3, 'upcoming': 2, 'nuptials': 2, 'causes': 4, 'stress': 1, 'nobel': 2, 'laureate': 1, 'economics': 1, 'milton': 5, 'stapler': 3, 'threat': 7, 'burn': 1, '1946': 15, 'typically': 1, 'shown': 9, 'investigate': 3, 'virus': 10, 'killed': 39, 'mankind': 8, 'concluded': 2, 'frodo': 11, 'baggin': 1, 'vanquish': 1, 'frank': 28, 'baum': 3, 'far': 7, 'mcadams': 14, 'daniel': 55, 'craig': 36, 'futuristic': 30, 'roddy': 5, 'mcdowall': 4, '1943': 5, 'horse': 24, 'raise': 7, 'boyle': 5, 'convict': 6, 'terrorizing': 4, 'da': 2, 'henreid': 3, 'claude': 4, 'kathryn': 3, 'mccormick': 1, 'dance': 28, 'costner': 20, 'elliot': 5, 'ness': 4, 'connery': 13, 'capone': 6, 'mock': 1, 'rob': 34, 'reiner': 13, 'fictional': 21, 'heavy': 12, 'signorney': 1, 'betty': 5, 'missouri': 2, 'louisiana': 6, 'purchase': 1, 'exposition': 1, 'fair': 8, '1904': 4, 'prostitute': 9, 'plots': 5, 'presidential': 5, 'candidate': 5, 'akira': 10, 'built': 5, 'tellings': 1, 'savannah': 1, 'especially': 1, 'darren': 4, 'aronofsky': 5, 'descending': 1, 'madness': 5, 'pack': 9, 'savior': 5, 'nazi': 15, 'credited': 1, 'killing': 23, 'notorious': 3, 'actually': 20, 'reilly': 4, 'satire': 8, 'fans': 8, 'marx': 13, 'groucho': 3, 'liners': 1, 'seaside': 1, 'comrades': 1, 'funny': 26, 'katie': 10, 'holmes': 5, 'dramatic': 22, 'effect': 1, 'bowie': 6, 'goblin': 3, 'henson': 10, 'fifth': 9, 'outwit': 2, 'swedish': 10, 'paying': 2, 'respects': 1, 'dying': 11, 'corresponds': 1, 'giving': 8, 'moses': 9, 'stage': 10, 'castaway': 2, 'pretty': 10, 'krackens': 1, 'showed': 7, 'medusa': 2, 'douglas': 9, 'financial': 8, 'district': 2, 'kline': 7, 'mismatched': 7, 'interests': 3, 'art': 7, 'reclusive': 1, 'artist': 12, 'puppet': 9, 'lies': 5, 'grow': 8, 'basketball': 23, 'championship': 6, 'sly': 9, 'teams': 9, 'aging': 27, 'merceneries': 2, 'tragic': 19, 'claire': 6, 'danes': 6, 'biopic': 22, 'founder': 3, 'facebook': 13, 'actual': 10, 'events': 26, 'prohibition': 6, 'dystopian': 17, 'debilitating': 1, 'stutter': 4, 'process': 6, 'overcome': 6, 'waters': 7, 'remain': 3, 'pure': 2, 'temptations': 1, 'previously': 6, 'still': 6, 'harm': 1, 'intended': 1, 'victims': 14, 'sleep': 13, 'translated': 1, 'means': 4, '1960': 33, 'journalist': 15, 'week': 6, 'longs': 4, 'convinced': 3, 'cheat': 8, 'entertainment': 5, 'happen': 9, 'spies': 3, 'carey': 5, 'escapee': 2, '1964': 15, 'fleming': 3, 'delves': 1, 'deep': 11, 'smuggling': 5, 'protagonist': 9, 'kyle': 4, 'machlachlan': 1, 'solve': 16, 'england': 22, 'interest': 15, 'literature': 3, 'harper': 10, 'saved': 15, 'employing': 2, 'factory': 12, 'scissors': 3, 'buys': 1, 'tennessee': 3, 'fragile': 1, 'blanche': 3, 'dubois': 2, 'moves': 10, 'watches': 1, 'disappear': 1, 'confederate': 5, 'includes': 27, 'aliens': 28, 'optimus': 2, 'prime': 10, 'bumblebee': 2, 'start': 22, 'gifted': 4, 'youngsters': 2, 'discovery': 8, 'onboard': 1, 'decided': 4, 'send': 7, 'angels': 2, 'humanity': 13, 'cafe': 2, 'rooney': 6, 'mara': 4, 'forty': 2, 'divorced': 10, 'husbands': 6, 'left': 23, 'younger': 11, 'timberlake': 13, 'amanda': 11, 'seyfried': 8, '25': 6, '1933': 5, 'ape': 6, 'state': 17, 'rampage': 3, 'zemeckis': 6, 'touchstone': 1, 'pictures': 10, 'alzheimer': 1, 'develops': 7, 'experimental': 5, 'chimpanzee': 1, 'surreal': 2, 'italian': 27, 'decadent': 2, '300': 2, 'spartans': 1, 'times': 19, 'tina': 6, 'fey': 6, 'identity': 10, 'bored': 3, 'evening': 5, 'ordinary': 4, 'bastian': 1, 'atreyu': 1, 'criminals': 13, 'begin': 5, 'suspect': 5, 'informant': 3, 'simple': 9, 'jewelery': 2, 'terribly': 3, 'chevy': 14, 'renew': 1, 'nevada': 2, 'neo': 15, 'noir': 20, 'roger': 5, 'kint': 2, 'soundtrack': 5, 'avril': 1, 'lavigne': 1, 'hole': 6, 'busta': 1, 'rhymes': 1, 'uses': 26, 'own': 46, 'methods': 2, 'writes': 10, 'hardships': 4, 'ghosts': 13, 'complications': 3, 'arise': 2, 'kansas': 15, 'wicked': 5, 'whales': 3, 'remember': 21, '1930': 19, 'powell': 1, 'nick': 16, 'nora': 4, 'detective': 41, 'crimes': 4, 'animators': 2, 'classical': 11, 'leopold': 1, 'stokowski': 1, 'conducts': 1, 'orchestra': 1, 'continuation': 2, 'students': 12, 'pursuing': 1, 'singing': 22, 'stanwyck': 4, 'macmurray': 2, 'insurance': 11, 'plans': 6, 'robotic': 5, 'creatures': 34, 'enemies': 6, 'lovely': 3, 'spring': 2, 'shrubbery': 2, 'murderer': 13, 'snoopy': 3, 'learn': 18, 'meaning': 9, 'holidays': 12, 'decorating': 1, 'sad': 7, 'adults': 7, 'wasteland': 3, 'kindhearted': 1, 'storyteller': 1, 'remus': 1, 'trickster': 1, 'br': 3, 'er': 3, 'outwits': 1, 'slow': 7, 'witted': 1, 'rule': 10, 'pride': 3, 'universe': 8, 'personalities': 1, 'blurred': 1, 'dramatises': 1, 'pearl': 3, 'harbor': 2, 'mike': 29, 'myers': 17, 'jay': 17, 'roach': 3, 'crystal': 16, 'lend': 5, 'immature': 1, 'encountering': 1, 'crisis': 10, 'moved': 5, 'cross': 14, 'glamorous': 2, 'ed': 14, 'helms': 6, 'wisconsin': 2, 'disillusionment': 2, 'suffers': 7, 'horrors': 6, 'vision': 5, 'builds': 8, 'ingredients': 1, 'jude': 8, 'forest': 13, 'whitaker': 7, 'repossession': 1, 'mambo': 1, 'eric': 8, 'garcia': 1, 'return': 21, 'canadian': 10, 'deals': 20, 'experiments': 2, 'genetic': 4, 'engineering': 1, 'too': 14, 'dc': 13, 'comics': 11, 'transformation': 1, 'batman': 12, 'provides': 7, 'island': 27, 'berk': 1, 'marks': 5, 'thompson': 13, 'attaches': 5, 'crewman': 1, 'ended': 4, 'charlie': 23, 'kareem': 3, 'abdul': 3, 'jabar': 1, 'bill': 24, 'weather': 6, 'days': 12, 'repeat': 2, 'controls': 1, 'touched': 2, 'ufo': 3, 'sighting': 1, 'fantastical': 5, 'instruct': 1, 'build': 5, 'diamond': 5, 'imprisonment': 3, 'budding': 2, 'romance': 57, 'spoiled': 2, 'cocker': 2, 'spaniel': 2, 'scruffy': 2, 'poor': 23, 'mutt': 1, 'hat': 2, 'freddie': 2, 'krueger': 5, 'descent': 2, 'becoming': 13, 'caretaker': 4, 'lots': 7, 'marshmallow': 3, 'goo': 1, 'shaving': 2, 'cream': 2, 'covered': 2, 'actors': 28, 'cassavetes': 1, 'rag': 3, 'tag': 12, 'drugs': 11, 'cats': 5, 'rabbits': 2, 'realizes': 7, 'unhappy': 5, 'needs': 10, 'bestseller': 1, 'self': 16, 'hypnosis': 3, 'hangs': 3, 'grand': 11, 'red': 31, 'freidkin': 1, 'chronicles': 20, 'possessed': 13, 'cunningham': 2, 'dirty': 7, 'politics': 5, 'segel': 10, 'blunt': 17, 'proposal': 1, 'walk': 6, 'aisle': 1, 'winslet': 19, 'passengers': 2, 'doomed': 8, 'voyage': 1, 'wudang': 1, 'swordsman': 4, 'fabled': 3, 'unlikely': 15, 'opponent': 1, 'required': 5, 'extensively': 1, 'ballerina': 11, 'soulmate': 1, 'mammoth': 2, 'sabertooth': 1, 'sloth': 2, 'saw': 6, 'godfather': 3, 'screen': 13, 'davy': 1, 'better': 6, 'dutchman': 1, 'heartbreaking': 1, 'tribulations': 3, 'chow': 11, 'yun': 11, 'fat': 16, 'affleck': 15, 'dramatizes': 1, 'secret': 33, 'iranian': 2, 'hostages': 2, 'create': 13, 'anything': 6, 'races': 4, 'ark': 1, 'priests': 2, 'tcm': 1, 'kind': 13, nan: 1, '1': 6, 'cat': 22, 'owners': 2, 'thief': 18, 'woo': 8, 'genie': 7, 'scared': 4, 'wits': 3, 'audiences': 2, 'simply': 1, 'repeating': 2, 'no': 42, 'dull': 3, 'blooms': 1, 'robots': 18, 'consumerist': 1, 'nightmare': 2, 'pearce': 4, 'uncovers': 6, 'policeman': 5, 'hollywood': 13, 'spawning': 1, 'remakes': 2, 'leatherface': 3, 'football': 13, 'stricken': 4, 'cancer': 26, 'support': 4, 'distance': 5, 'idealism': 1, 'twin': 5, 'watch': 5, 'dual': 3, 'fraternal': 1, 'wacky': 5, 'gary': 17, 'sheriff': 19, 'refuses': 4, 'lines': 10, 'songs': 7, '1985': 20, 'dr': 23, 'emmett': 1, 'brown': 8, 'remarry': 1, 'exhusband': 1, 'behind': 16, 'vacation': 17, 'bumbling': 5, 'thieves': 6, 'stepsisters': 2, 'compete': 6, 'kris': 3, 'kristofferson': 2, 'interwoven': 1, 'redemption': 6, 'overdose': 1, 'restaurant': 6, 'stick': 2, 'reigniting': 1, 'matthew': 25, 'broderick': 13, 'senior': 3, 'cutting': 2, 'unlucky': 4, 'jinxes': 1, 'stealing': 4, 'coins': 1, 'alone': 7, 'vigilante': 3, 'mass': 6, 'jules': 3, 'verne': 2, 'vanessa': 7, 'hudgens': 6, '1969': 17, 'joesph': 1, 'gordon': 32, 'levitt': 21, 'patient': 7, 'maintaining': 1, 'sam': 31, 'obsession': 3, 'juliette': 3, 'binoche': 3, 'willem': 5, 'dafoe': 5, '2002': 31, 'soccer': 4, 'parminder': 1, 'nagra': 1, 'keira': 3, 'knightley': 2, 'idea': 7, 'lot': 12, 'mutated': 1, 'lynching': 1, 'believed': 3, 'cattle': 3, 'rustlers': 1, 'cursed': 6, 'try': 50, 'ending': 13, 'radcliffe': 9, 'special': 29, 'critically': 11, 'flopped': 1, 'gather': 2, 'information': 3, 'wiped': 2, 'population': 5, 'fathers': 5, '2005': 24, 'corpses': 2, 'oskar': 2, 'schindler': 3, 'discovering': 4, 'beliefs': 1, 'carroll': 7, 'agency': 1, 'double': 3, 'faye': 9, 'dunaway': 10, 'psychological': 19, 'convicted': 13, 'rapist': 4, 'attorney': 7, 'helps': 25, 'wahlberg': 23, 'boston': 6, 'campbell': 10, 'replaces': 1, 'chainsaw': 5, 'stripper': 8, 'burlesque': 1, 'emerges': 1, 'domineering': 2, 'perseus': 13, 'andromeda': 1, 'intervention': 1, 'darabont': 2, 'robbins': 14, 'altered': 1, 'dramatically': 1, 'seeing': 8, 'ufos': 2, 'beard': 1, 'told': 25, 'kennedy': 1, 'nobody': 3, 'eat': 6, 'elderly': 8, 'portray': 5, 'hiker': 2, 'literally': 1, 'hard': 13, 'chick': 5, 'seem': 7, 'glasses': 4, 'dress': 10, 'comicbook': 2, 'twelve': 1, 'exes': 4, 'jesse': 13, 'eisenberg': 10, 'zuckerburg': 1, 'harvard': 3, 'manager': 12, 'hogwarts': 5, 'height': 3, 'disco': 4, 'remembered': 2, 'catchy': 1, 'courtesy': 1, 'bee': 1, 'gees': 1, 'suit': 5, 'jockey': 6, 'overwhelming': 1, 'odds': 7, 'oversized': 1, 'primate': 4, 'residing': 2, 'skull': 3, 'lusts': 1, 'fincher': 7, 'singer': 20, 'conmen': 1, 'hughes': 9, 'able': 6, 'admitted': 3, 'psychiatric': 5, 'institute': 9, 'patients': 4, 'aurthur': 1, 'round': 2, 'table': 2, 'hilarious': 10, 'scrapes': 1, 'brighter': 1, 'campy': 3, 'lycanthrope': 1, 'backward': 1, 'timeline': 1, 'dvd': 1, 'mastermind': 9, 'pawns': 5, 'scheme': 13, 'profoundly': 2, 'changing': 8, 'emilio': 4, 'estevez': 4, 'harry': 13, 'stanton': 3, 'neal': 5, 'ali': 5, 'macgraw': 2, 'fated': 5, 'scoot': 2, 'tough': 13, 'foul': 3, 'mouthed': 2, 'concoction': 1, 'delivers': 5, 'chilling': 3, 'performance': 18, 'sequestered': 1, 'innkeeper': 1, 'fagin': 1, 'ringleader': 2, 'pickpockets': 1, 'surrounding': 3, 'pow': 5, 'hell': 10, 'wasted': 1, 'horro': 1, 'roommate': 2, 'obsessed': 12, 'perform': 4, 'seasons': 1, 'fell': 5, 'wagon': 1, 'tanked': 1, 'canon': 3, 'leonard': 7, 'nemoy': 1, 'sang': 4, 'protagonists': 1, 'precursor': 1, 'accident': 15, 'memories': 16, '1992': 34, 'virginia': 1, 'madsen': 1, 'tony': 32, 'todd': 4, 'replaced': 3, 'hook': 4, 'ferrell': 27, 'galifianaki': 1, 'congressman': 3, 'incredible': 2, 'taken': 19, 'tornado': 11, 'lifetime': 3, 'six': 13, 'chipmunks': 2, 'marooned': 1, 'cruise': 32, 'egotistical': 1, 'stockbroker': 1, 'gekko': 2, 'america': 31, 'triple': 8, 'crown': 8, '1966': 13, 'cub': 14, 'simba': 14, 'grows': 15, 'freddy': 6, 'kruger': 2, 'murdering': 6, 'francois': 3, 'truffaut': 5, 'recognized': 1, 'nouvelle': 1, 'vague': 1, 'movement': 3, 'filmmakers': 1, 'focusing': 6, 'antoine': 2, 'doinel': 1, 'continent': 1, 'begins': 12, 'charlton': 13, 'heston': 13, 'act': 7, 'retire': 3, 'anonymously': 2, 'tombstone': 1, 'arizona': 2, 'disrupted': 2, 'eliminating': 1, 'acts': 5, 'air': 10, 'traffic': 1, 'controllers': 1, 'atari': 1, '2600': 1, 'spoof': 12, 'forlorn': 1, 'owner': 17, 'supply': 7, 'absorbing': 1, 'adjustments': 1, 'messy': 2, 'schooler': 2, 'promiscuous': 1, 'rumors': 1, 'beverly': 4, 'angelo': 3, 'comically': 1, 'walley': 2, 'sociopathic': 1, 'punk': 7, 'spree': 4, 'bernardo': 3, 'bertolucci': 3, 'swept': 5, 'lones': 1, 'china': 9, 'ruler': 6, 'ethan': 11, 'hawke': 5, 'whereby': 1, 'oprah': 4, 'winfrey': 3, 'glover': 6, 'vietnam': 31, 'berenger': 3, 'familiar': 1, 'shakespear': 4, 'slightly': 1, 'silent': 20, 'tetsuo': 1, 'biker': 4, 'samberg': 4, 'preacher': 1, 'charlize': 3, 'theron': 3, 'source': 3, 'professional': 17, 'squid': 1, 'android': 2, 'sunglasses': 1, 'lemmon': 19, 'marilyn': 13, 'monroe': 13, 'funniest': 2, 'took': 24, 'pandora': 6, 'satellite': 1, 'orbiting': 2, 'alpha': 2, 'centauri': 1, 'savage': 3, 'treat': 1, 'horses': 3, 'sewell': 1, 'monstrous': 2, 'curse': 5, 'handsome': 2, 'lin': 2, 'furious': 2, 'descends': 2, 'foolishly': 1, 'disposing': 1, 'estate': 3, 'daughters': 4, 'flattery': 1, 'consequences': 3, 'penguins': 7, 'cuban': 12, 'refugee': 3, 'powerful': 22, 'ultimately': 8, 'basically': 3, 'artificial': 9, 'insemination': 2, 'mobsters': 3, 'robbing': 5, 'bank': 25, 'monkey': 4, 'suits': 4, 'revolutionary': 7, 'species': 10, 'ways': 17, '1999': 38, 'gave': 6, 'rise': 12, 'shyamalan': 4, 'endings': 1, 'subtitled': 1, 'episode': 5, 'iv': 4, 'opening': 6, 'scroll': 1, 'ratso': 1, 'rizzo': 1, 'appear': 8, 'trunk': 1, 'hallows': 1, 'public': 3, 'explode': 1, 'india': 8, 'hippie': 7, 'commune': 4, 'garrett': 1, 'hedlund': 1, 'digital': 3, 'abrams': 7, 'filming': 11, 'nina': 1, 'troubled': 17, 'dancer': 11, 'dreamy': 1, 'prevent': 6, 'importation': 1, 'heroin': 1, 'summer': 19, 'unexpectedly': 3, 'discover': 19, 'provide': 6, 'mufasa': 5, 'blond': 5, 'wields': 2, 'hammer': 6, 'earned': 11, 'screenplay': 12, 'villagers': 1, 'hire': 7, 'somewhat': 4, 'boring': 5, 'disgusting': 1, 'diarrhea': 1, '47': 3, 'presented': 1, 'spends': 4, 'entire': 4, 'until': 15, 'stumbles': 9, 'colleague': 1, 'val': 10, 'kilmer': 10, 'dwarf': 6, 'ewok': 2, 'brave': 4, 'archer': 3, 'demi': 4, 'moore': 34, 'swayze': 6, 'pottery': 1, 'open': 8, 'shop': 5, 'sir': 13, 'conan': 1, 'doyle': 2, 'ritchie': 3, 'elwes': 6, 'buttercup': 4, 'elton': 5, 'sculptures': 1, 'almost': 12, 'location': 6, 'outside': 7, 'momma': 2, 'perspective': 4, 'wanting': 6, 'lose': 5, 'virginity': 4, 'audition': 1, 'read': 7, 'auditioning': 1, 'kenneth': 10, 'branagh': 7, 'ocd': 2, 'enchanted': 4, 'compassion': 1, 'paranormal': 5, 'footage': 13, 'abc': 2, 'network': 2, 'wears': 8, 'tie': 3, 'autobots': 6, 'decepticons': 5, 'popularized': 3, 'hello': 4, 'gorgeous': 3, 'peckinpah': 4, 'texas': 12, 'border': 2, 'nimoy': 1, 'shattner': 1, 'humpback': 1, 'compton': 1, 'casual': 2, 'elf': 2, 'dobby': 1, 'sadly': 1, 'lange': 3, 'girlfriends': 5, 'airline': 5, 'whip': 1, 'bum': 2, 'selected': 3, 'idiot': 7, 'angel': 13, 'business': 18, 'existed': 5, 'chewed': 1, 'scenery': 1, 'joan': 4, 'crawford': 2, 'wire': 2, 'coat': 3, 'hangers': 1, 'chooses': 2, 'pursue': 4, 'gritty': 8, 'hank': 3, 'azaria': 3, 'portal': 3, 'terrorizes': 10, 'hanger': 1, 'avoiding': 3, 'hunters': 6, 'target': 6, 'hiding': 6, 'carrie': 4, 'bradshaw': 1, 'retirement': 4, 'slim': 2, 'pickens': 2, 'airplane': 5, 'hacker': 7, 'fishburn': 1, 'morpheus': 1, '2013': 3, 'included': 8, 'seyfriend': 1, 'revolution': 6, 'utters': 2, 'gonna': 4, 'bigger': 4, 'attacks': 4, 'kept': 6, 'revolutionized': 1, 'force': 14, 'portrayal': 10, 'britain': 7, 'monarch': 3, 'spain': 7, 'painful': 2, 'stunts': 5, 'outer': 9, 'horcruxes': 3, 'existence': 5, 'objects': 3, 'wizarding': 2, 'patriotic': 2, 'clean': 9, 'poker': 1, 'robbery': 7, 'brainerd': 1, 'position': 1, 'gory': 2, 'assault': 5, 'omaha': 1, 'beach': 10, 'asleep': 4, 'distinction': 2, 'kenya': 1, 'introduces': 4, 'concepts': 1, 'xenomorphs': 1, 'facehuggers': 1, 'corporate': 7, 'aliases': 1, 'colors': 6, 'ie': 1, 'mr': 17, 'pink': 6, 'mostly': 9, 'later': 20, 'odessa': 2, 'steps': 3, 'sequence': 4, 'odysseus': 1, 'hick': 1, 'follow': 14, 'parent': 3, 'footprints': 1, 'porn': 4, 'overeating': 1, 'vince': 7, 'vaughn': 10, 'catches': 6, 'claim': 7, 'youngest': 1, 'traci': 2, 'officers': 8, 'munchkins': 3, 'gere': 10, 'don': 10, 'cheadle': 2, 'fuqua': 1, 'hulce': 5, 'demons': 4, 'creating': 4, 'masterpieces': 1, 'supervillian': 2, 'defeated': 3, 'call': 10, 'duty': 3, 'distributed': 2, 'number': 14, 'imdb': 2, '250': 1, 'maritla': 1, 'investigator': 14, 'publishing': 2, 'kesey': 2, 'questionable': 2, 'treatment': 3, 'unfit': 1, 'graduates': 1, 'spend': 4, 'cruising': 1, 'buddies': 6, 'heading': 1, 'dancers': 9, 'ogre': 14, 'hippies': 2, 'partied': 1, 'pattinson': 14, 'veterinarian': 3, 'circus': 13, 'split': 3, 'farrel': 1, 'smiling': 1, 'weimar': 1, 'republic': 4, 'gas': 2, 'aang': 3, 'overthrow': 4, 'peace': 3, 'fate': 9, 'manage': 3, 'p': 6, 'join': 6, 'forces': 23, 'smugglers': 2, 'conclusion': 3, 'novels': 13, 'tolkein': 4, 'rogers': 6, 'rowan': 4, 'atkinson': 4, 'pokes': 3, 'emile': 3, 'hirsch': 2, 'olivia': 7, 'thirlby': 2, 'gorak': 1, 'hoth': 2, 'frozen': 6, 'cave': 3, 'ceiling': 1, 'flee': 2, 'bi': 1, 'swoops': 1, 'sellers': 7, 'wibberley': 1, 'tramp': 2, 'yukon': 1, 'eating': 14, 'shoe': 2, 'attends': 1, 'party': 22, '2007': 15, 'gone': 17, '1976': 20, 'numerical': 5, 'spartan': 1, 'persians': 1, 'capital': 2, 'brainwashed': 3, 'gulf': 3, 'vice': 1, 'directing': 3, '1991': 32, 'rehearse': 1, 'lyrics': 2, 'hills': 2, 'alive': 13, 'kay': 2, 'hapless': 12, 'carnival': 3, 'performer': 4, 'caught': 11, 'tarentino': 5, 'nicknames': 2, 'pairs': 6, 'grifters': 2, 'caper': 2, 'helped': 13, 'chases': 3, 'contaminate': 1, 'fort': 1, 'knox': 1, 'contents': 1, 'preston': 2, 'minutes': 4, 'bluff': 1, 'spectacular': 2, 'mothership': 1, 'hear': 5, 'scream': 2, 'lovelorn': 1, 'monolith': 2, 'mandy': 8, 'krypton': 2, 'doc': 4, 'happens': 6, 'holed': 2, 'warehouse': 1, 'duvall': 4, 'confess': 1, 'spike': 20, 'direct': 4, 'rights': 8, 'iteration': 3, 'although': 8, 'hannibal': 7, 'cannibal': 2, 'thomas': 15, 'mcbride': 2, 'exodus': 3, 'cleese': 3, 'jewel': 2, 'filibustering': 2, 'hours': 5, 'wrongly': 5, 'nickelsen': 1, 'adultery': 2, 'mythological': 2, 'banished': 2, 'stuffed': 7, 'astronaut': 6, 'simians': 2, 'troll': 2, 'granter': 1, 'parrot': 1, 'quirky': 7, 'embroiled': 1, 'odd': 5, 'travis': 3, 'bickle': 2, 'imitated': 1, 'outcast': 4, 'jungle': 15, 'pull': 3, 'cons': 2, 'golightly': 5, 'socialite': 5, 'interested': 2, 'former': 46, 'assassin': 19, 'gunned': 2, 'youtube': 2, 'age': 20, 'we': 16, 'neighbor': 13, 'aired': 3, 'jumpsuit': 1, 'dons': 1, 'painted': 1, 'zero': 1, 'mostel': 1, 'springtime': 1, 'hitler': 5, 'germany': 7, 'chocolates': 3, 'wealthy': 14, 'france': 13, 'stuff': 3, 'hamilton': 7, 'ecstatic': 1, 'considering': 1, '1963': 6, 'escaping': 8, 'rooted': 1, 'writings': 4, '20': 9, 'romano': 2, 'leguizamo': 2, 'denis': 2, 'leary': 2, 'takeoff': 1, 'disaster': 10, 'screwball': 13, 'nba': 2, 'jabaar': 1, 'appearances': 2, 'aboard': 12, 'flight': 6, 'awry': 4, 'nerdy': 5, 'escaped': 11, 'commit': 5, 'ward': 8, 'speaking': 6, 'hinders': 1, 'politician': 12, 'raimi': 4, 'quintessential': 3, 'feral': 1, 'adulthood': 2, 'rejects': 2, 'civilized': 1, 'voorhes': 1, 'sexy': 3, 'venture': 2, 'amazing': 9, 'capra': 10, 'cendrillon': 1, 'perrault': 1, 'camera': 9, 'volunteers': 2, 'tearjerker': 3, 'delorean': 3, 'neve': 3, 'terrorized': 3, 'ghostface': 6, 'wore': 5, 'cape': 1, 'fly': 9, 'bowling': 1, 'smoking': 2, 'joints': 2, 'drinking': 3, 'russians': 3, 'dessert': 3, 'intentioned': 1, 'australian': 12, 'elite': 7, 'operative': 3, 'unbeknownst': 1, 'technologically': 1, 'advanced': 1, 'e': 19, 'survival': 7, 'practically': 1, 'moab': 1, 'desert': 13, 'ten': 5, 'memorable': 11, 'protects': 3, 'ancestors': 1, 'sasha': 7, 'baron': 11, 'cohen': 14, 'shines': 1, 'ousted': 1, 'eastern': 4, 'challenged': 2, 'ocean': 12, 'sinks': 10, 'rats': 2, 'commands': 1, 'bidding': 1, 'shoot': 11, 'pulled': 1, 'machine': 10, 'carpet': 3, 'inept': 5, 'burglars': 2, 'met': 4, 'match': 5, 'macaulay': 11, 'culkin': 11, 'sized': 1, 'helper': 1, 'satirical': 5, 'gun': 10, 'culture': 4, 'trade': 6, 'center': 4, 'pentagon': 1, 'albie': 1, 'nervous': 2, 'banks': 9, 'depth': 2, 'affects': 3, 'industrial': 3, 'usa': 5, 'occurrences': 1, 'tape': 2, 'appointed': 4, 'fill': 1, 'vacancy': 1, 'senate': 1, 'carl': 6, 'fredrickson': 1, '82': 1, 'nd': 5, 'segal': 5, 'predominantly': 1, 'ann': 3, 'arbor': 1, 'michigan': 3, 'university': 4, 'sights': 2, 'rwanda': 1, 'naturalist': 1, 'dian': 1, 'fossey': 2, 'trouble': 14, 'actions': 2, 'cause': 4, 'rumpelstiltskin': 3, 'kingdom': 8, 'dispatch': 1, 'warring': 4, 'apaches': 2, 'haruki': 1, 'miyzaki': 1, 'boar': 2, 'maggie': 6, 'bloodthirsty': 2, 'farmhouse': 2, 'catherine': 11, 'hicks': 1, 'developed': 5, 'status': 2, 'gwyneth': 10, 'midst': 4, 'racehorse': 8, 'slight': 1, 'limp': 2, 'superior': 3, 'han': 7, 'solo': 8, 'statham': 15, 'aid': 2, 'choose': 7, 'toys': 15, 'scarlet': 9, 'labute': 1, 'aaron': 6, 'eckhart': 3, 'nation': 7, 'christophe': 1, 'waltz': 3, 'foxx': 13, 'teamed': 2, 'wachowski': 5, 'laurence': 9, 'fishburne': 7, 'supporting': 20, 'acress': 1, 'olive': 1, 'guidance': 3, 'counselor': 4, 'extract': 3, 'organs': 4, 'paid': 3, 'share': 8, 'milla': 3, 'jolovich': 1, 'alice': 11, 'corporation': 4, 'supervisors': 1, 'san': 8, 'francisco': 6, 'idaho': 1, 'nerd': 3, 'immigrant': 7, 'pedro': 2, 'psychopaths': 1, 'paths': 1, 'mcguyver': 1, 'mcfadden': 1, 'slave': 15, 'subsequently': 2, 'emperor': 7, 'irving': 5, 'ruby': 2, 'slippers': 3, 'cousin': 3, 'freed': 3, 'nolan': 11, 'joseph': 27, 'condon': 3, 'shadows': 1, 'gotham': 9, 'bane': 8, 'reveal': 3, 'barrett': 3, 'cavilleri': 2, 'chemistry': 1, 'deny': 1, 'ignore': 1, 'tatou': 1, 'shy': 3, 'waitress': 8, 'nicer': 1, 'quention': 1, 'boomers': 1, 'meeting': 13, 'funeral': 5, 'spending': 6, 'weekend': 8, 'partying': 1, 'chloe': 2, 'griffiths': 1, 'examines': 4, 'distinct': 1, 'angles': 1, 'relate': 2, 'incident': 1, 'points': 5, 'fugitive': 4, 'false': 4, 'ids': 1, 'weapons': 6, 'apprentice': 11, 'mouse': 11, 'judge': 9, 'esteemed': 1, 'reviewing': 1, 'committing': 5, 'reorganizes': 1, 'convince': 7, 'superheros': 7, 'wallace': 10, 'scotland': 6, 'oppression': 1, 'costs': 1, 'principal': 3, 'hayao': 8, 'trek': 3, 'wally': 2, 'described': 3, 'dialogue': 4, 'mongrel': 1, 'labrador': 1, 'inferior': 1, 'monkeys': 5, 'brolin': 5, 'mindbending': 1, 'puerto': 2, 'bemused': 1, 'kemp': 1, 'tenma': 1, 'royale': 2, 'brutally': 4, 'backdrop': 4, 'vehicle': 4, 'recently': 13, 'collin': 1, 'farrell': 7, '1983': 16, 'debra': 4, 'winger': 4, 'sacrificed': 1, 'employed': 2, 'norwegian': 1, 'snowboard': 1, 'surprisingly': 1, 'non': 7, 'gives': 13, 'instruction': 1, 'randomly': 1, 'televised': 3, 'freakish': 1, 'firefly': 3, 'directorial': 5, 'bateman': 16, 'brian': 11, 'bale': 15, 'murderous': 5, 'wall': 8, 'brett': 1, 'easton': 1, 'ellis': 2, 'lea': 3, 'invading': 4, 'terminal': 2, 'lung': 2, 'disease': 8, 'befriended': 1, 'kindly': 1, 'protective': 2, 'klaatu': 4, 'gort': 1, 'farmer': 12, 'joins': 11, 'guerrilla': 4, 'shrek': 11, 'reminds': 1, 'zorro': 2, 'abbott': 1, 'senators': 1, 'pact': 2, 'pennant': 1, 'historic': 7, 'figure': 11, 'peaceful': 2, 'teachings': 1, 'moody': 2, 'olympic': 4, 'runners': 2, 'religious': 5, 'backgrounds': 1, 'sport': 11, 'failing': 2, 'ditch': 3, 'generated': 3, 'statistics': 1, 'unbeatable': 1, 'ties': 3, 'balloons': 9, 'lifelong': 4, 'south': 27, 'riding': 11, 'locke': 1, 'revolved': 3, 'bebe': 2, 'ebenezer': 5, 'scrooge': 8, 'december': 4, 'hybrid': 5, 'fanfare': 1, 'amour': 1, 'garry': 3, 'armed': 4, 'everywhere': 3, 'elliott': 1, 'plastic': 5, 'surgeon': 5, 'rely': 1, 'cover': 5, 'lie': 4, 'westley': 2, 'mcavoy': 10, 'working': 38, '19': 4, 'gregory': 17, 'peck': 17, 'entangled': 2, 'bitter': 3, 'feud': 1, 'internet': 4, 'ecological': 1, 'disappearing': 3, 'marvels': 1, 'aquatic': 3, 'maya': 11, 'rudolph': 10, 'newly': 9, 'sports': 37, 'shouts': 1, 'mediterranean': 1, 'near': 12, 'italy': 15, 'cars': 11, 'vin': 7, 'diesel': 7, 'dad': 11, 'buy': 2, 'grifter': 1, 'tout': 1, 'advantage': 2, 'fortuitous': 1, 'circumstances': 3, 'wrestling': 1, 'promoter': 2, 'kenau': 2, 'slowed': 1, 'moon': 7, 'scorsesi': 1, 'combined': 3, 'talents': 2, 'sorvino': 1, 'happy': 6, 'lucky': 5, 'honest': 3, 'interacts': 2, 'puts': 11, 'antarctica': 2, 'overdoses': 2, 'illegal': 5, 'squeal': 1, 'kaye': 3, 'inn': 1, 'vermont': 1, 'pill': 12, 'increases': 1, 'intelligence': 9, 'chernobyl': 1, 'amity': 7, 'forcing': 2, 'hates': 4, 'rescuing': 3, 'forte': 4, 'wolfgang': 2, 'amadeus': 2, 'mozart': 6, 'berridge': 2, 'indianna': 1, 'alter': 5, 'ego': 5, 'noam': 1, 'chomsky': 1, 'sides': 6, 'abortion': 1, 'debate': 1, 'psychologist': 3, 'cybertronian': 2, 'spacecraft': 2, 'reach': 1, 'daycare': 2, 'brink': 2, 'worlds': 2, 'safe': 2, 'affairs': 2, 'contract': 4, 'unable': 3, 'dates': 1, 'affair': 21, 'tasked': 10, 'scheming': 2, 'duping': 1, 'company': 29, 'cahoots': 1, '1944': 8, 'array': 1, 'designated': 1, '1945': 9, 'brooklyn': 15, 'tenement': 2, 'stuffing': 1, 'inside': 10, 'freezing': 1, 'astro': 1, 'modeled': 1, 'belle': 7, 'conner': 3, 'rookie': 3, 'veteran': 11, 'modus': 2, 'operandi': 2, 'vi': 6, 'details': 5, 'kings': 3, 'ordeal': 1, 'conquer': 2, 'incessant': 2, 'stammer': 1, 'quietly': 1, 'estranged': 4, 'decade': 2, 'braff': 1, '65': 2, 'merlin': 7, 'descendant': 3, 'krige': 1, 'unresisting': 2, 'reardon': 1, 'involvement': 3, 'kitty': 3, 'collins': 7, 'mazursky': 2, 'jill': 5, 'clayburgh': 3, 'use': 23, 'struggle': 10, 'freedom': 12, 'sunken': 2, '1956': 14, 'inhabit': 2, 'citizens': 3, 'bodies': 10, 'fraternity': 4, 'cukor': 4, 'remarriage': 1, 'alternate': 10, 'pertaining': 1, 'oldman': 2, 'mila': 15, 'kunis': 15, 'ridiculous': 2, 'pranks': 2, 'laughs': 5, 'expense': 1, 'pain': 1, 'bilbo': 6, 'baggins': 5, 'gandalf': 5, 'thorin': 1, 'defeat': 13, 'dragon': 7, 'smaug': 1, 'relive': 3, 'ensure': 2, 'jacob': 2, 'spaceship': 9, 'map': 4, 'discovered': 7, 'artifacts': 2, 'cultures': 1, 'stumble': 4, 'distant': 4, 'extinction': 1, 'walter': 11, 'brennan': 1, 'hillbilly': 2, 'sharpshooter': 2, 'sudeikis': 4, 'rid': 5, 'employers': 10, 'studded': 6, 'effectively': 1, 'warn': 1, 'destruction': 9, 'rapid': 2, 'arenas': 1, 'dinosaur': 5, 'themed': 7, 'pro': 4, 'piccolo': 1, 'nolte': 5, 'frankie': 1, 'madison': 1, 'expecting': 4, 'partner': 12, 'bootlegging': 2, 'prepare': 2, 'laugh': 2, 'robber': 4, 'feelings': 3, 'robbed': 1, 'instead': 16, 'scandal': 6, 'feel': 8, 'undersized': 2, 'toby': 2, 'maquire': 2, 'atlanta': 1, 'burns': 3, 'onslaught': 2, 'faces': 7, 'tremendous': 1, 'jedi': 5, 'durante': 1, 'mail': 1, 'narrator': 1, 'frees': 2, 'vintage': 2, 'highrise': 1, 'exciting': 3, 'front': 6, 'thirties': 1, 'capcom': 1, 'dressed': 5, 'monthly': 1, 'event': 3, 'fellow': 6, 'executives': 1, 'cusak': 2, 'brazilian': 5, 'serious': 7, 'inability': 1, 'speak': 7, 'victor': 5, 'hugo': 3, 'hugh': 11, 'jackman': 6, 'candlestick': 1, 'heroine': 4, 'bonds': 3, 'loyalty': 2, 'tested': 2, 'dario': 5, 'argento': 5, 'attending': 1, 'recite': 1, 'thy': 1, '40': 13, 'percent': 5, 'photo': 2, 'realistic': 2, 'meredith': 1, 'correspondent': 1, 'ernie': 1, 'pyle': 2, '18': 3, 'infantry': 1, 'shoes': 3, 'vacated': 1, 'petula': 1, 'collect': 1, 'mony': 1, 'python': 3, 'terry': 15, 'gilliam': 10, 'pilots': 4, 'communicate': 4, 'favreau': 7, 'region': 2, 'sorrowful': 1, 'laying': 2, 'wheelbarrow': 1, 'differently': 1, 'unique': 10, 'researchers': 2, 'secretary': 3, 'theif': 1, 'dropped': 2, 'buddy': 15, 'canyon': 2, 'humfrey': 1, 'supposed': 9, 'convenience': 4, 'staple': 2, 'daddy': 6, 'mist': 1, 'plotting': 3, 'humanoids': 2, 'harmony': 1, 'worship': 1, 'goddess': 1, 'eywa': 1, 'addresses': 1, 'social': 17, 'class': 22, 'focussing': 1, 'frenchman': 1, 'tribesman': 1, 'massacre': 1, 'beethoven': 2, 'section': 3, 'hate': 8, 'bigotry': 1, 'explodes': 2, 'overweight': 5, 'advice': 3, 'harvey': 10, 'embellished': 1, 'scar': 5, 'right': 17, 'burger': 1, 'dazzles': 1, 'hearts': 1, 'minds': 1, 'darth': 9, 'vader': 8, 'results': 3, 'stylized': 1, 'forgot': 1, 'cosner': 2, 'hears': 4, 'corn': 4, 'swap': 4, 'renolds': 1, 'kimble': 1, 'battling': 10, 'greater': 3, 'basis': 3, 'yul': 7, 'brynner': 4, 'blasts': 1, 'unrecognizable': 1, 'apes': 8, 'amusement': 4, 'blast': 1, '1934': 9, 'spanish': 4, 'inquisition': 1, 'coppola': 21, 'origins': 4, 'vito': 1, 'corleone': 3, 'sun': 2, 'cell': 1, 'phones': 1, 'cameras': 5, 'boys': 5, 'preparation': 2, 'spent': 6, 'bedtime': 1, 'amnesiac': 1, 'ludlum': 5, 'introducing': 2, 'daily': 3, 'eskimo': 1, 'screams': 1, 'used': 16, 'nonhumans': 1, 'fake': 8, 'highway': 1, '66': 3, 'effected': 1, 'problems': 9, 'angelina': 16, 'jolie': 17, 'transporting': 2, 'nitroglycerin': 1, 'outfit': 3, 'smurf': 1, 'cupcake': 1, 'east': 6, 'hits': 3, 'theseus': 2, 'zeus': 10, 'hyperion': 1, 'poe': 5, 'mafia': 11, 'beasley': 1, 'di': 12, 'caprio': 11, 'crashes': 5, 'retiring': 3, 'bag': 2, 'million': 6, 'dollars': 4, 'pirate': 11, 'cycling': 1, 'puppets': 3, 'kermit': 4, 'miss': 2, 'piggy': 2, 'concept': 2, 'considerably': 1, 'gruesome': 4, 'teeth': 1, 'flesh': 7, 'pus': 1, 'sores': 1, 'list': 4, 'skinned': 3, 'felines': 2, 'nevus': 1, 'mode': 2, 'transportation': 2, 'risk': 1, 'starvation': 1, 'terrible': 8, 'cannibalistic': 3, 'rural': 4, 'areas': 2, 'expose': 4, 'fianc': 2, 'pullman': 1, 'heroes': 11, 'autobiography': 4, 'ages': 4, 'present': 10, 'maturing': 1, 'pip': 2, 'christina': 8, 'aguilar': 1, 'chorus': 1, 'aguilera': 5, 'owned': 1, 'mississippi': 1, 'mixed': 11, 'minor': 12, 'telekinesis': 4, 'wiseguy': 1, 'pileggi': 1, 'tattoo': 3, 'assassinating': 1, 'largest': 3, 'ladd': 2, 'gunslinger': 4, 'tires': 1, 'range': 2, 'oda': 1, 'mae': 1, 'whoopi': 9, 'goldberg': 10, 'admirer': 1, 'hers': 2, 'despise': 1, 'jobs': 5, 'drummers': 3, 'johan': 2, 'venue': 1, 'concert': 15, 'slapstick': 3, '1958': 8, 'attributes': 1, 'na': 2, 'ruffalo': 3, 'clash': 2, 'penny': 2, 'grandson': 3, 'heaven': 3, 'touches': 1, 'inspires': 1, 'contact': 5, 'travelers': 2, 'monique': 1, 'abusive': 7, 'grossly': 1, 'obese': 2, 'illiterate': 3, 'arm': 9, 'surfing': 4, 'pfizer': 2, 'viagra': 3, 'challenges': 2, 'congress': 3, 'exhausting': 1, 'filibuster': 2, 'speech': 10, 'champ': 1, 'interferes': 1, 'stephenie': 2, 'meyer': 4, 'filmd': 1, 'assigned': 2, 'dozen': 2, 'murderers': 2, 'assassination': 6, 'shakespearean': 10, 'bickering': 1, 'hakuna': 4, 'matata': 4, 'tonight': 1, 'bio': 1, 'kissed': 2, 'underdog': 5, 'hawaii': 4, 'kiss': 6, 'rolling': 1, 'surf': 1, 'honolulu': 1, 'sinatra': 1, 'angela': 8, 'felissa': 1, 'shamalan': 2, 'psychiatrist': 2, 'haley': 7, 'osment': 4, 'flew': 1, 'thornhill': 1, 'chased': 8, 'scotsman': 3, 'independence': 2, 'pancakes': 1, 'fabian': 1, 'ask': 4, 'sandberg': 1, 'waste': 3, 'collecting': 2, 'embarks': 3, 'die': 13, 'solving': 2, 'project': 8, 'endangers': 1, 'tokyo': 9, 'rampaging': 2, 'psionic': 1, 'raucous': 1, 'bachelor': 7, 'asia': 2, 'shootout': 3, '78': 2, 'tying': 1, 'disappearance': 5, 'happenings': 6, 'plaguing': 1, 'suburban': 5, 'strict': 3, 'ninja': 2, 'woodie': 1, 'doll': 10, 'buzzlightyear': 1, 'longer': 5, 'currency': 1, 'eddy': 1, 'exotic': 4, 'superstar': 1, 'dwelling': 1, '1917': 1, 'photograph': 2, 'scientific': 1, 'evidence': 3, 'fairies': 2, 'danish': 1, 'marries': 6, 'piglet': 2, 'alec': 4, 'guiness': 1, 'dimensional': 1, 'ferguson': 4, 'kane': 2, 'hodder': 1, 'word': 9, 'cabin': 6, 'destinies': 1, 'tried': 4, 'romantically': 2, 'wesley': 4, 'snipes': 3, 'ib': 1, 'joy': 1, 'least': 6, 'wanted': 12, 'acadamy': 1, 'pizza': 7, 'delivery': 6, 'driver': 8, 'gyllenhal': 2, 'develop': 5, 'complex': 2, 'intimate': 1, 'incompatible': 1, 'abandoned': 4, 'ringing': 1, 'bells': 1, 'transforms': 4, 'shaggy': 1, 'jodie': 12, 'foster': 14, 'montagues': 1, 'capulets': 1, 'ticket': 5, 'sweet': 4, 'dances': 3, 'poseidon': 1, 'lightning': 6, 'bolt': 1, 'entry': 4, 'heroic': 5, 'la': 8, 'karaoke': 2, 'clownfish': 6, 'zoe': 7, 'saldana': 7, 'ren': 8, 'zellweger': 8, 'worker': 16, 'situation': 5, 'expected': 4, 'celebrated': 4, 'woos': 1, 'conceives': 1, 'treasury': 1, 'zachary': 5, 'levy': 1, 'volatile': 2, 'vickie': 1, 'novelist': 1, 'graham': 2, 'greene': 2, '1949': 7, '8': 7, 'poison': 2, 'charming': 10, 'relationships': 5, 'wholesale': 2, 'uncover': 8, 'buried': 4, 'raised': 9, 'siezes': 1, 'kung': 10, 'fu': 11, 'lansbury': 4, 'paige': 2, 'hara': 6, 'robby': 2, 'benson': 1, 'sacha': 4, 'tyrant': 3, 'rafting': 2, 'marked': 5, 'collaboration': 3, 'christoph': 2, 'raindrops': 1, 'dicken': 3, 'stingy': 2, 'fenway': 1, 'ugly': 4, 'unintelligent': 1, 'moments': 3, 'jenny': 2, 'eludes': 2, 'ago': 7, 'activities': 4, 'mobster': 8, 'associates': 1, 'tagline': 4, 'belong': 1, 'wishes': 9, 'navigate': 3, 'desolate': 3, 'grotesque': 1, 'trust': 2, 'famed': 10, 'activity': 2, 'debut': 11, 'names': 5, 'distress': 1, 'gabriel': 4, 'burne': 1, 'crossing': 2, 'assumable': 1, 'covers': 6, 'voyages': 1, 'explorers': 2, 'dispatched': 1, 'weyland': 1, 'stieg': 1, 'larsson': 1, 'disgraced': 1, 'mikael': 1, 'blomkvist': 1, 'investigates': 6, 'patriarch': 5, 'niece': 4, 'critics': 4, 'claimed': 2, 'uncomfortable': 1, 'overt': 1, 'homosexual': 4, 'themes': 2, 'piano': 4, 'revolving': 2, 'explores': 7, 'method': 1, 'chronicling': 3, 'naturally': 1, 'widely': 2, 'inaccurately': 1, 'depicted': 11, 'jumanji': 1, 'softer': 1, 'pale': 2, 'ricci': 1, 'headless': 1, 'horseman': 1, 'entrepreneur': 1, 'caribbean': 2, 'attraction': 1, 'floats': 2, 'ton': 2, 'attached': 1, 'bar': 2, 'morocco': 4, 'konami': 1, 'ahab': 2, 'crib': 1, 'macchio': 5, 'wax': 5, 'miyagi': 2, 'ebsen': 2, 'switched': 4, 'exit': 1, 'allergic': 1, 'reaction': 2, 'costume': 5, 'ginger': 1, 'cole': 2, 'porter': 1, 'comeback': 6, 'prisoner': 5, 'jailer': 1, 'conflict': 5, 'schizophrenia': 3, 'gigantic': 3, 'deaths': 3, 'warfare': 2, 'algiers': 1, 'algeria': 1, 'gettng': 1, 'beds': 1, 'scare': 5, 'addition': 3, 'mounted': 1, 'sequences': 1, 'burstyn': 1, 'every': 15, 'fitted': 2, 'prosthetics': 1, 'necks': 1, 'emaciated': 1, 'pound': 2, 'nine': 5, 'wigs': 1, 'snl': 3, 'kristin': 5, 'breakdown': 3, 'invention': 1, 'princeton': 1, 'spaghetti': 5, 'lean': 6, 'rousing': 2, 'toole': 9, 'arabia': 3, 'arab': 2, 'tribes': 2, 'turks': 1, 'fired': 3, 'starts': 15, 'dating': 4, 'cute': 4, 'chapter': 1, 'werewolf': 3, 'agreement': 1, 'wives': 6, 'committed': 9, 'bridal': 1, 'hans': 2, 'zimmer': 1, 'rza': 5, 'quiet': 3, 'drawn': 5, 'runaway': 7, 'christin': 1, 'saturday': 7, 'cigarettes': 1, 'unborn': 2, 'intense': 12, 'neighbors': 5, 'morpurgo': 1, 'curtiz': 5, 'technicolor': 4, 'nyc': 11, 'narcotics': 2, 'bureau': 2, 'onto': 5, 'katniss': 1, 'everdeen': 1, 'seth': 24, 'rogen': 10, 'screewriter': 1, 'reiser': 1, 'baruchel': 8, 'today': 2, 'upset': 2, 'piss': 1, 'parting': 1, 'austin': 1, 'waiting': 1, 'inspiring': 3, 'fab': 2, 'banned': 8, 'care': 12, 'type': 14, 'shreck': 1, 'antonia': 1, 'baderez': 1, 'wielding': 5, 'helena': 3, 'bonham': 2, 'carter': 5, 'insomnia': 1, 'maker': 5, 'jodi': 1, 'bombed': 2, 'missed': 1, 'crazies': 1, 'mtv': 2, 'knoxville': 2, 'tenth': 1, 'anniversary': 2, 'exacts': 1, 'prosecuting': 1, 'renowned': 3, 'rapper': 3, 'funky': 1, 'posing': 2, 'captures': 1, 'makers': 1, 'borat': 1, 'reclaim': 6, 'spot': 2, 'suess': 5, 'dredd': 1, 'bella': 3, 'coven': 1, 'engage': 5, 'madcap': 3, 'hijinks': 2, 'hall': 2, 'int': 1, 'whimsical': 3, 'balloon': 2, 'lifted': 1, 'earn': 4, '$': 1, '100': 10, 'kidnapper': 1, 'dunham': 1, 'ventriloquist': 1, 'debbie': 1, 'hilary': 5, 'swank': 8, 'accompanied': 1, 'zither': 1, 'dunbar': 2, 'intolerable': 1, 'aberration': 1, 'forth': 1, 'ny': 3, 'fran': 2, 'nic': 2, 'teutonic': 2, 'jarmusch': 3, 'zen': 1, 'saucer': 1, 'intent': 3, 'warning': 1, 'curb': 1, 'aggressions': 1, 'reveals': 2, 'jk': 3, 'persuade': 2, 'makeshift': 1, 'glory': 4, 'oddball': 1, 'psychopathic': 1, 'psychedelic': 2, 'escapades': 2, 'whats': 6, 'hypocritical': 1, 'swinging': 2, 'raising': 6, 'satiric': 1, 'elements': 5, 'bible': 8, 'quoting': 1, 'wash': 2, 'miguel': 1, 'arteta': 1, 'moscow': 2, 'via': 1, 'kidnaps': 3, 'torments': 2, 'philosophy': 3, 'arabian': 4, 'lamp': 3, 'yrs': 1, 'julianne': 9, 'glamorized': 1, 'unimpressionable': 1, 'frankly': 8, 'dear': 7, 'damn': 8, 'orphans': 2, 'mustafa': 1, 'selina': 1, 'gomez': 4, 'cassidy': 3, 'leighton': 3, 'meester': 3, 'summoned': 2, 'skeptic': 1, 'grad': 1, 'researching': 2, 'interment': 1, 'camps': 2, 'infected': 1, 'dashiell': 1, 'hammett': 1, 'eye': 14, 'spade': 3, 'lebowski': 1, 'slacker': 2, 'avid': 1, 'bowler': 1, 'philly': 1, 'apollo': 3, 'creed': 3, 'weight': 4, 'interconnected': 2, 'niven': 2, 'quinn': 3, 'defends': 7, 'raping': 2, 'montana': 4, 'aidan': 1, 'members': 10, 'ludlow': 1, 'thors': 1, 'weapon': 2, 'recovered': 1, 'lonely': 6, 'hidden': 7, 'sharni': 1, 'vincent': 3, 'hip': 6, 'hop': 6, 'competed': 1, 'natascha': 1, 'mcelhone': 1, 'asked': 6, 'pavel': 1, 'antipov': 1, 'pasha': 1, 'surprised': 3, 'offered': 2, '1955': 6, 'sergio': 11, 'leone': 10, 'robards': 1, 'cheyenne': 1, 'abilities': 9, 'higher': 1, 'standing': 5, 'frederic': 1, 'forrest': 2, 'warm': 1, 'olivier': 4, 'hawk': 1, 'adaptations': 1, 'encouraged': 2, 'vicious': 7, 'cartel': 8, 'partially': 3, 'letter': 11, 'victorian': 2, 'miser': 4, 'rebel': 5, 'suceed': 1, 'pursuits': 1, 'bodham': 1, 'pods': 3, 'fro': 1, 'sullivan': 2, 'keller': 3, 'bancroft': 4, 'patty': 2, 'duke': 3, 'jory': 1, 'caterpillars': 1, 'mischievous': 2, 'normal': 3, 'sentimental': 1, 'raft': 2, 'hungry': 3, 'procreate': 1, 'seventeen': 1, 'aristocrat': 2, 'claimant': 1, 'luxurious': 2, 'titanic': 3, 'brandishes': 2, 'mighty': 2, 'experiences': 7, 'unexplainable': 1, 'terror': 6, 'example': 2, 'voorhees': 2, 'figures': 3, 'bride': 8, 'waking': 1, 'coma': 5, 'bent': 2, 'sheepdogs': 2, 'herd': 1, 'hoggett': 2, 'rodriguez': 7, 'slashes': 1, 'corruption': 5, 'megan': 4, 'awarded': 2, 'cynically': 2, 'selfish': 1, 'bearing': 3, 'lessons': 8, 'delivered': 1, 'asks': 8, 'grown': 6, 'naked': 1, 'horribly': 5, 'established': 4, 'trademark': 1, 'idolize': 3, '1936': 5, 'tornados': 1, 'metro': 3, 'goldwyn': 4, 'mayer': 2, 'vera': 1, 'rosemary': 3, 'oh': 2, 'tap': 3, 'libyans': 1, 'flux': 1, 'capacitors': 1, 'loreans': 1, 'skateboards': 1, 'touching': 1, 'impossible': 2, 'bizarre': 4, 'llama': 2, 'joyce': 1, 'deed': 1, 'sentence': 5, 'dwayne': 8, 'johnson': 15, 'retells': 5, 'pranking': 1, 'ni': 2, 'celine': 1, 'dion': 1, 'childless': 3, 'bury': 1, 'backyard': 2, 'containing': 1, 'infant': 1, 'shortage': 1, 'budget': 11, 'arnie': 1, 'enjoys': 1, 'honey': 5, 'bailey': 7, 'rediscovers': 1, 'beauty': 7, 'already': 1, 'deer': 9, 'passing': 1, 'hitchhock': 1, 'unforgettable': 3, 'steals': 8, 'employer': 4, 'client': 3, 'cut': 5, 'relies': 1, 'rumor': 4, 'mill': 3, 'advance': 2, 'botched': 2, 'horrifically': 1, 'eyes': 3, 'blessed': 1, 'reel': 1, 'showman': 2, 'undergo': 2, 'procedure': 2, 'erase': 1, 'sour': 1, 'curry': 7, 'sarrandon': 1, 'lazy': 2, 'tired': 4, 'substance': 3, 'abuse': 4, 'drifter': 2, 'agrees': 4, 'stands': 2, 'inheritance': 2, 'lawn': 5, 'ornaments': 1, 'friendship': 12, 'knowing': 3, 'bikini': 3, 'extras': 2, 'graduation': 2, 'ceremony': 4, 'thirsty': 2, 'crawls': 1, 'promotion': 1, 'gathering': 2, 'purpose': 2, 'mocking': 1, 'whatever': 2, 'fools': 2, 'maniac': 2, 'superheroes': 7, 'asgardian': 2, 'levit': 1, 'dictator': 6, 'climbing': 3, 'eiffel': 2, 'twice': 3, 'quick': 2, 'overselling': 1, 'intentionally': 2, 'lesser': 2, 'segment': 2, 'walrus': 2, 'masks': 2, 'fur': 2, 'flowers': 2, 'cricket': 1, 'conscience': 3, 'prove': 4, 'worthy': 2, 'awaited': 2, 'within': 6, 'programs': 2, 'training': 8, '200': 2, 'puppies': 3, 'experiencing': 2, 'cheats': 2, 'cheerleader': 1, 'glimpse': 1, '1300': 1, 'necronomicon': 2, 'falsely': 3, 'marlin': 4, 'classicmovie': 1, 'impeccable': 1, 'capabilities': 1, 'craft': 3, 'promote': 1, 'contemporaries': 1, 'cocaine': 10, 'poking': 1, 'becca': 1, 'torn': 4, 'dramas': 1, 'prom': 2, 'sending': 1, 'waits': 1, 'mcconaughey': 6, 'democracy': 2, 'calls': 3, 'radio': 6, 'vents': 1, 'walks': 3, 'shoots': 2, 'servants': 2, 'putting': 1, 'beyond': 3, 'native': 4, 'ravaged': 1, 'technology': 3, 'replicating': 1, 'engaged': 3, 'area': 5, 'pay': 8, 'residence': 5, 'furter': 2, 'salesman': 3, 'worldwide': 1, 'producing': 1, 'aggressive': 2, 'praying': 1, 'mantis': 1, 'persons': 2, 'matter': 2, 'seuss': 10, 'recycles': 1, 'meyers': 1, 'prominent': 1, 'otherwise': 1, 'fantasia': 3, 'teaching': 4, 'skills': 8, 'reluctant': 3, 'otherworldly': 2, 'psychotic': 7, 'confronts': 2, 'tennis': 1, 'complete': 5, 'seems': 9, 'couples': 7, 'leave': 4, 'nightmarish': 2, 'biomechanical': 1, 'sexual': 7, 'unknown': 10, 'origin': 3, 'truckers': 1, 'nostromo': 1, 'surrealist': 1, 'hoping': 4, 'awesome': 5, 'imaginative': 2, 'since': 9, 'harder': 2, 'sergei': 3, 'eisenstein': 2, 'filmmaking': 3, 'sewed': 1, 'skins': 2, 'silly': 7, 'medieval': 4, 'bunny': 3, 'sure': 4, 'hokey': 1, 'cusack': 9, 'coddry': 1, 'madeup': 1, 'focusses': 1, 'feline': 4, 'mains': 1, 'rom': 2, 'com': 2, 'common': 13, 'tosses': 1, 'shield': 2, 'suspenseful': 2, 'frightening': 2, 'delighting': 1, 'generations': 1, 'eliminated': 1, 'thirty': 1, 'encounters': 6, 'proprietor': 1, 'domination': 2, 'witty': 2, 'defies': 2, 'norms': 2, 'farmboy': 1, 'cowhand': 1, 'talks': 9, 'trail': 8, 'hiring': 3, 'invents': 1, 'step': 4, 'hooks': 2, 'definitely': 1, 'profound': 3, 'destroying': 2, 'erich': 4, 'maria': 6, 'remarque': 3, 'witchcraft': 1, 'wizardry': 2, 'mandatory': 1, 'fitting': 1, 'supports': 1, 'courier': 1, 'service': 3, 'completely': 6, 'fit': 1, 'felt': 2, 'spinster': 1, 'problem': 12, 'accept': 2, 'witnessing': 8, 'meaningful': 1, 'reason': 2, 'statue': 3, 'bazooko': 1, 'yard': 3, 'reconnect': 1, 'proves': 1, 'surprise': 3, 'phillip': 7, 'why': 8, 'syrena': 1, 'chalices': 1, 'sparrow': 10, 'threats': 3, 'unlisted': 2, 'miami': 7, 'florida': 6, 'quickly': 1, 'establishes': 3, 'weary': 2, 'gunfighter': 4, 'homestead': 1, 'smoldering': 1, 'settler': 1, 'rancher': 1, 'guardians': 2, 'flic': 1, 'combines': 3, 'explosive': 2, 'underwent': 1, 'grueling': 1, 'except': 2, 'spared': 1, 'resent': 1, 'convey': 1, 'resentment': 1, 'idolized': 1, 'bynes': 1, 'announcing': 1, 'returned': 4, 'unite': 3, 'threatening': 3, 'strangely': 1, 'idealized': 1, 'frustrating': 2, 'northern': 1, 'overtaken': 1, 'feathered': 1, 'employeer': 1, 'knife': 2, 'stabbing': 1, 'animate': 1, 'specifically': 2, 'magically': 5, 'whenever': 3, 'packs': 2, 'groups': 5, 'arkansas': 2, 'housewife': 8, 'whisked': 3, 'barnabas': 4, 'ancestral': 1, 'dysfunctional': 5, 'descendants': 1, 'tops': 1, 'lists': 1, 'rep': 2, 'fraud': 1, 'arouses': 1, 'suspicions': 1, 'should': 3, 'racing': 7, 'orwellian': 1, 'pigs': 4, 'ran': 4, 'establishment': 2, 'insides': 1, 'skip': 1, 'heder': 4, 'alienated': 2, 'presidency': 2, 'mcclain': 3, 'fought': 7, 'terrorist': 7, 'caring': 3, 'miniature': 1, 'twenty': 5, 'wonders': 2, 'kendrick': 6, 'ladies': 3, 'cappella': 2, 'competition': 11, 'annasophia': 1, 'robb': 1, 'sleeve': 1, 'occur': 1, 'digitally': 1, 'removed': 1, 'directer': 1, 'alcohol': 3, 'antonio': 6, 'banderas': 4, 'furry': 4, 'happened': 7, 'implanted': 3, 'executes': 2, 'sinister': 3, 'sydney': 4, 'pollack': 1, 'views': 4, 'madeleine': 2, 'kahn': 2, 'bogdanovich': 1, 'hollywoods': 1, 'biggest': 4, 'flops': 1, 'beneath': 1, 'boulder': 2, 'faced': 4, 'decision': 5, 'titled': 4, 'hg': 1, 'wells': 6, 'martians': 2, 'matheson': 2, 'brashear': 1, 'trained': 6, 'acotr': 1, 'walsh': 1, 'bondsman': 1, 'moscone': 1, 'accountant': 1, 'jonathan': 10, 'mardukas': 1, 'grodin': 3, 'freemon': 1, 'sinful': 2, 'bradely': 1, 'allows': 4, 'allowing': 1, 'brain': 5, 'capacity': 2, 'breakthrough': 3, 'tracy': 7, 'rare': 4, 'card': 4, 'volkswagen': 1, 'beetle': 1, 'revised': 1, 'chevrolet': 1, 'camaro': 1, 'assailant': 2, 'reopen': 2, 'site': 4, 'drowning': 1, 'sanctuary': 2, 'regarding': 3, 'disorder': 2, 'forgotten': 1, 'chaos': 4, 'harassed': 1, 'locals': 2, 'leaving': 2, 'ensues': 5, 'deceased': 4, 'oz': 3, 'mocked': 1, 'stoltz': 1, 'excels': 2, 'despite': 8, 'facial': 1, 'deformity': 1, 'upside': 5, 'columbus': 2, 'chuckie': 3, 'dwarfs': 1, 'swordsmen': 2, 'sworn': 1, 'rapacious': 1, 'claudette': 5, 'colbert': 5, 'flashes': 1, 'hitchike': 1, 'heiress': 3, 'flags': 1, 'protege': 2, 'ailing': 1, 'recruiting': 2, 'handicapped': 1, 'galifiniakis': 1, 'crom': 1, 'driven': 7, 'thirst': 1, 'brutal': 5, 'slaughter': 1, 'gothic': 2, 'destined': 2, 'lynch': 5, 'lawnmower': 2, 'dax': 2, 'shepard': 2, 'june': 2, 'collects': 1, 'reaching': 2, 'waterboarded': 1, 'torture': 6, 'seconds': 1, 'per': 2, 'freaky': 2, 'grew': 2, 'starrin': 1, 'unfortunate': 2, 'arc': 1, 'praised': 1, 'posters': 1, 'immortal': 1, 'forever': 3, 'grim': 3, 'reaper': 1, 'obi': 2, 'wan': 2, 'kenobi': 2, 'tatooine': 1, 'meme': 1, 'dur': 1, 'mistakenly': 2, 'germans': 1, 'synonymous': 1, 'gale': 2, 'month': 2, 'georgine': 1, 'darcy': 2, 'torso': 1, 'relaxing': 2, 'marrying': 4, 'discrete': 1, 'months': 2, 'seduce': 2, 'mrs': 6, 'increasingly': 2, 'harassment': 1, 'debt': 1, 'fulfilling': 1, 'fantasies': 1, 'renegade': 2, 'elia': 4, 'kazan': 4, 'inge': 2, 'sultry': 1, 'scarlett': 11, 'none': 3, 'naomi': 1, 'watts': 1, 'orange': 6, 'indigo': 1, 'salieri': 2, 'nicolson': 2, 'nominations': 4, 'strongman': 1, 'bradbury': 3, 'architect': 2, 'infiltrates': 2, 'implants': 1, 'ideas': 4, 'sleeping': 2, 'subconscious': 1, 'ranked': 1, 'wohlberg': 1, 'attempting': 11, 'mini': 1, 'coopers': 1, 'website': 2, 'creator': 4, 'cunning': 1, 'charismatic': 4, 'drinks': 2, 'bourbon': 1, 'succeeding': 1, 'seventh': 1, 'christmastime': 1, 'slashers': 1, 'rack': 1, 'count': 3, 'globe': 6, 'frat': 1, 'reading': 3, 'conrad': 1, 'darkness': 4, 'loud': 2, 'keeper': 1, 'model': 1, 'silver': 3, 'nathan': 8, 'lane': 13, 'produce': 2, 'flop': 7, 'quote': 16, 'gena': 1, 'rowlands': 1, 'cassvetes': 1, 'ad': 2, 'libbed': 1, 'performs': 1, 'danced': 1, 'gershwin': 2, 'vicente': 1, 'fang': 1, 'goldfish': 2, 'unforgiving': 1, '45': 1, 'birthdays': 1, 'atticus': 3, 'finch': 5, 'penned': 1, 'hailed': 7, 'pays': 1, 'homage': 2, 'brawny': 1, 'colonel': 2, 'woodrow': 1, 'dolarhyde': 1, 'watched': 4, 'strippers': 4, 'heard': 4, 'posthumous': 1, 'cheating': 4, 'joad': 1, 'pursued': 4, 'minded': 1, 'bronson': 3, 'sinking': 6, 'liar': 2, 'priceless': 2, 'statuette': 1, 'situations': 4, 'berrymore': 1, 'golfing': 1, 'joke': 1, 'raved': 1, 'anymore': 4, 'hilariously': 4, 'burial': 1, 'nathalie': 1, 'benefits': 3, 'rode': 1, 'feared': 2, 'surfers': 1, 'wave': 3, 'kenji': 1, 'mizoguchi': 1, 'concubine': 1, 'stigma': 1, 'sold': 5, 'labor': 2, 'rightful': 1, 'acrophobia': 1, 'interaction': 1, 'laid': 2, 'talent': 3, 'sign': 2, 'packed': 14, 'stepmother': 4, 'aluminum': 1, 'diaz': 15, 'elementary': 2, 'simultaneously': 1, 'inmates': 7, 'banker': 7, 'raspy': 1, 'rogan': 5, 'yearns': 1, 'teach': 6, 'cubs': 2, 'wherein': 1, 'gnomes': 4, 'pretended': 1, 'suicide': 6, 'stared': 7, 'hamill': 2, 'schools': 3, 'confused': 2, 'jaded': 1, 'partake': 1, 'brawls': 1, 'napoleon': 2, 'lizard': 9, 'ralphie': 3, 'teachers': 2, 'types': 1, 'items': 4, 'mohr': 1, 'stalks': 3, 'removes': 1, 'clothing': 2, 'spunky': 1, 'hamster': 1, 'realize': 2, 'tinman': 1, 'lover': 10, 'searing': 1, 'heat': 2, 'pets': 5, 'ranch': 1, 'woodland': 5, 'royal': 5, 'sketch': 2, 'arthurian': 2, 'johansson': 5, 'entertaining': 1, 'dealer': 6, 'jung': 4, 'countries': 2, 'truly': 1, 'clashing': 1, 'ballerinas': 2, 'canyoneering': 1, 'tragically': 3, 'arguably': 2, 'performing': 5, 'muggles': 1, 'factions': 1, 'differences': 1, 'forward': 1, 'broke': 3, 'cera': 4, 'lloyd': 3, 'invent': 1, 'device': 2, 'elevator': 3, 'victorious': 1, '1865': 1, 'nicknamed': 4, 'chief': 4, 'broom': 2, 'roommates': 2, 'variety': 5, 'dreaming': 1, 'terrorize': 5, 'wiggles': 1, 'rehabilitative': 1, 'quintin': 2, 'tarintino': 1, 'greusome': 1, 'cuts': 3, 'pours': 1, 'engagement': 2, 'lasts': 1, 'pierce': 6, 'brosnan': 3, 'ewen': 1, 'mcgregor': 6, 'minister': 9, 'fullest': 2, 'physician': 1, 'fallen': 5, 'ranger': 1, 'bound': 3, 'cleaning': 5, 'therefore': 2, 'examples': 1, 'eradicate': 1, 'lycan': 1, 'uninhabited': 1, 'responsibility': 2, 'buscemi': 3, 'experiment': 6, 'robe': 3, 'dragons': 5, 'vikings': 1, 'marijuana': 1, 'blake': 8, 'lively': 6, 'label': 1, 'intern': 7, 'labels': 1, 'reunion': 2, 'individuals': 3, 'sidekicks': 1, 'traits': 1, 'cripple': 1, 'joss': 5, 'whedon': 5, 'dudley': 5, 'hayley': 2, 'message': 4, 'phoenix': 6, 'hustlers': 4, 'aunts': 1, 'insane': 10, 'destroys': 4, 'mordor': 3, 'ridden': 4, 'regular': 4, 'weatherman': 1, 'hunky': 1, 'jerry': 7, 'mcguire': 2, 'insecurity': 1, 'freeing': 1, 'risks': 3, 'proportions': 1, 'disowns': 1, 'noticeable': 1, 'pass': 2, 'framed': 2, 'attracted': 3, 'belushi': 3, 'toga': 2, 'parties': 1, 'el': 1, 'mariachi': 1, 'bandaras': 1, 'salma': 3, 'hayek': 3, 'mcdowell': 8, 'hosts': 1, 'reproduce': 1, 'pre': 5, 'subservient': 1, 'orangutans': 1, 'chimpanzees': 2, 'pierre': 4, 'boulle': 1, 'humankind': 2, 'catastrophic': 1, 'malevolent': 1, 'michelle': 10, 'pfeiffer': 4, 'journalists': 2, 'annette': 4, 'benning': 3, 'thora': 1, 'birch': 1, 'brenner': 1, 'sidekick': 11, 'wager': 1, 'believes': 6, 'commoner': 2, 'flighty': 1, 'brightens': 1, 'emerge': 1, 'positive': 1, 'strapped': 5, 'fairly': 1, 'bean': 1, 'limited': 3, 'universal': 4, 'afflicted': 1, 'lupine': 1, 'mildly': 1, 'retarded': 1, 'exploits': 4, 'dozens': 1, 'suzanne': 2, 'befriend': 5, 'miyaggi': 1, 'teaches': 10, 'blow': 2, 'hillary': 3, 'ruth': 2, 'citizen': 1, 'savant': 1, 'hershey': 3, 'urinate': 1, 'result': 2, 'hang': 2, 'ups': 4, 'downs': 3, 'wind': 3, 'skyscraper': 2, 'classics': 1, 'poem': 3, 'allan': 4, 'blown': 1, 'encounter': 9, 'amphibian': 2, 'frost': 4, 'dominant': 1, 'expires': 1, 'purchased': 3, 'beforehand': 1, 'hostile': 2, 'extra': 5, 'terrestrial': 4, 'stalk': 3, 'granny': 1, 'housing': 2, 'cfo': 1, 'ponzi': 3, 'crossdresses': 1, 'identities': 2, 'melanie': 2, 'griffith': 2, 'sells': 3, 'abducted': 1, 'molested': 1, 'pedophile': 1, 'sleeper': 2, 'visits': 2, 'reinvent': 1, 'purchasing': 1, 'villa': 1, 'killers': 4, 'inevitable': 1, 'defoe': 1, 'stationed': 1, 'forster': 2, 'gaiman': 1, 'viola': 1, 'historically': 1, 'luxury': 4, 'liner': 5, 'reduced': 2, 'miracle': 1, 'jesus': 1, 'cleaver': 1, 'inspector': 2, 'juvenile': 1, 'punks': 1, 'displays': 1, 'rather': 7, 'cecile': 2, 'b': 8, 'demille': 3, 'burrough': 1, 'undying': 1, 'loving': 2, 'sadistic': 2, 'temper': 4, 'poitier': 1, 'originated': 1, 'virgil': 2, 'tibbs': 1, 'sperm': 6, 'donor': 5, 'actively': 1, 'blossoms': 1, 'infamously': 1, 'passenger': 4, 'witnessed': 1, 'jane': 9, 'newlyweds': 1, 'valjean': 1, 'ntozake': 2, 'shange': 2, 'feminist': 1, 'phylicia': 1, 'rashad': 1, 'janet': 10, 'wildness': 1, 'mate': 4, 'mary': 10, 'connor': 11, 'guillermo': 8, 'faun': 3, 'exorcisms': 3, 'exorcism': 4, 'renee': 3, 'harmless': 1, 'multi': 2, 'talented': 5, 'effective': 1, 'purportedly': 1, 'parker': 9, 'interrupted': 1, 'derailment': 1, 'errors': 1, 'online': 2, 'anythig': 1, 'particualr': 1, 'sorry': 1, 'eh': 1, 'anybody': 1, 'stupid': 2, 'negative': 1, 'nancy': 2, 'sit': 2, 'watching': 5, 'jump': 3, 'fear': 6, 'fillini': 1, 'ruins': 2, 'quinten': 3, 'terentino': 1, 'delfonics': 1, 'motorcycles': 1, 'liked': 2, 'asunder': 1, 'mom': 12, 'suggestions': 1, 'plant': 4, 'parodied': 2, 'usually': 2, 'guess': 2, 'mann': 5, 'stresses': 1, 'hitting': 1, 'milestone': 2, 'identify': 1, 'receive': 2, 'shia': 7, 'lebouf': 2, 'swarzenegger': 1, 'menace': 2, 'warheads': 1, 'bed': 1, 'depends': 2, 'timeless': 9, 'include': 5, 'imagine': 2, 'bloody': 3, 'misnomer': 1, '24': 2, 'enlists': 8, 'mutants': 5, 'superhuman': 6, 'malicious': 1, 'iii': 2, 'security': 2, 'parkinson': 6, 'sena': 2, 'dystopic': 1, 'detroit': 2, 'terminally': 2, 'wounded': 3, 'submerged': 1, 'haunting': 3, 'slang': 1, 'term': 1, 'squeeze': 1, 'yoda': 2, 'rescues': 2, 'perhaps': 4, 'staff': 2, 'interviewed': 1, 'tuskegee': 3, 'airmen': 1, 'access': 1, 'logbooks': 1, 'siam': 3, 'esteem': 1, 'inaccuracies': 2, 'perceived': 2, 'disrespect': 2, 'monarchy': 2, 'muppets': 1, 'needed': 2, 'closing': 1, 'muppet': 1, 'tobacco': 3, 'smoke': 1, 'pumped': 1, 'tube': 1, 'slid': 1, 'pant': 1, 'leg': 5, 'shirt': 1, 'dinklage': 1, 'achondroplasic': 1, 'auditioned': 1, 'washes': 1, 'ashore': 1, 'attitudes': 2, 'seal': 1, 'rocks': 2, 'neptune': 1, 'islands': 1, 'landis': 3, 'tourists': 5, 'admit': 1, 'exists': 1, 'instant': 1, 'straight': 3, 'enforcer': 4, 'clans': 1, 'tradition': 5, 'slaying': 1, 'adder': 1, 'mcinnerny': 1, 'reported': 1, 'accompanies': 1, 'haddock': 2, 'tenessee': 1, 'stella': 1, 'broken': 3, 'anxiety': 1, 'prepares': 1, 'hospitalized': 2, 'disability': 1, 'scuba': 1, 'diver': 1, 'concoct': 1, 'headed': 7, 'fuehrer': 1, 'enraged': 1, 'rant': 2, 'memes': 1, 'fail': 1, 'smash': 7, 'ariel': 1, 'dweller': 1, 'finale': 4, 'reaches': 2, 'billionaire': 4, 'storm': 2, 'knocks': 1, 'atop': 1, 'veers': 1, 'admiral': 1, 'ozzel': 1, 'incompetence': 1, 'loner': 2, 'stepmom': 1, 'infested': 2, 'foe': 4, 'busy': 3, 'crate': 1, 'adorable': 4, 'succumbing': 2, 'greed': 4, 'gay': 9, 'addict': 2, 'suffering': 4, 'professionals': 2, 'record': 10, 'investigation': 2, 'suspicious': 1, 'package': 4, 'irish': 6, 'sharing': 2, 'lifestyles': 3, 'apparently': 3, 'closes': 1, 'interact': 5, 'repent': 1, 'riotous': 1, 'critter': 1, 'golfers': 1, 'programed': 1, 'detest': 1, 'plastique': 1, 'item': 2, 'larry': 5, 'cable': 3, 'tow': 1, 'spinning': 6, 'cogs': 1, 'cope': 4, 'possible': 5, 'dwarves': 3, 'plunder': 1, 'eras': 1, 'tribute': 2, 'jj': 1, 'aprams': 1, 'presents': 1, 'schoolkids': 1, 'sophisticated': 2, 'mcmurphy': 2, 'voluntarily': 1, 'lesley': 1, 'request': 1, 'unleash': 1, 'revealed': 3, 'mcdonald': 1, 'mac': 1, 'rodridguez': 1, 'stahl': 1, 'bastard': 1, 'refer': 1, 'dastan': 1, 'safeguard': 1, 'dagger': 4, 'hippy': 2, 'fbi': 5, 'cool': 1, 'amy': 13, 'adams': 15, 'pfieffer': 1, 'recreate': 1, 'zookeeper': 1, 'occupying': 1, 'lifeboat': 1, 'bengal': 1, 'yippy': 1, 'ki': 3, 'ya': 2, 'expletive': 1, 'resort': 4, 'switching': 2, 'transferred': 1, 'canine': 1, 'millions': 3, 'slimer': 1, 'halls': 1, 'prestigious': 2, 'momsen': 1, 'cindy': 1, 'lou': 2, 'homicidal': 2, 'xd': 1, 'electricity': 1, 'utter': 2, 'knightly': 1, 'jordan': 4, 'parallax': 1, 'threatened': 2, 'newcomer': 2, 'fabulous': 1, 'taste': 1, 'corral': 1, 'purebreds': 1, 'harness': 1, 'storybook': 1, 'liberty': 1, 'tongue': 1, 'pirates': 2, 'birds': 6, 'genres': 4, 'collide': 1, 'regain': 4, 'tactics': 1, 'lips': 3, 'miraculous': 2, 'landing': 2, 'marlowe': 1, 'blackmail': 1, 'sleeps': 1, 'credit': 4, 'foil': 1, 'thwart': 1, 'antagonist': 5, 'rodents': 3, 'size': 2, 'brute': 1, 'norway': 1, 'rothstein': 1, 'chucky': 1, 'entity': 4, 'satan': 2, 'mega': 2, 'desire': 3, 'mcfarlane': 3, 'fredo': 1, 'lay': 1, 'gyllenhall': 2, 'amount': 1, 'worthington': 6, 'kraken': 2, 'diagnosed': 3, 'stalone': 1, 'write': 4, 'insists': 1, 'crying': 1, 'encourage': 1, 'veterans': 4, 'irreparably': 1, 'roofer': 1, 'caused': 7, 'potions': 2, 'headmaster': 1, 'maguical': 1, 'conspire': 2, 'pension': 2, 'fund': 2, 'doldrums': 1, 'tiny': 11, 'keyhole': 1, 'rocky': 4, 'decker': 6, 'sing': 2, 'rapes': 1, 'primates': 4, 'caution': 1, 'fasten': 1, 'seatbelts': 1, 'bumpy': 1, 'clarice': 3, 'lecter': 3, 'instantly': 1, 'matchsticks': 1, 'repeatedly': 2, 'announce': 1, 'excellent': 2, 'autistic': 4, 'xavier': 1, 'recruit': 2, 'cyclops': 1, 'iceman': 1, 'thrown': 3, 'groupie': 1, 'carolina': 2, 'propels': 1, 'mainly': 2, 'plain': 1, 'whisper': 1, 'rosebud': 3, 'plagued': 4, 'insanity': 3, 'toxin': 2, 'contaminated': 6, 'gylenhaal': 3, 'injured': 3, 'reporters': 1, 'decipher': 1, 'raunchy': 3, 'geeks': 4, 'singapore': 1, 'refuge': 2, 'lame': 1, 'binoculars': 1, 'chapman': 3, 'britons': 3, 'grateful': 1, 'pod': 2, 'doors': 2, 'transvestite': 3, 'transsylvania': 1, 'decapitated': 1, 'preview': 2, 'suffer': 2, 'cloned': 2, 'exhibits': 1, 'amok': 1, 'bridesmaid': 1, 'threaten': 2, 'upend': 1, 'pastry': 1, 'chef': 2, 'claustrophobic': 1, 'boredom': 1, 'filth': 1, 'sheer': 1, 'wings': 4, 'onstage': 1, 'random': 4, 'shopping': 2, 'surrounded': 2, 'annoying': 2, 'rehab': 1, 'alcoholism': 1, 'surfer': 3, 'obstacles': 3, 'losing': 4, 'mathematical': 1, 'janitor': 2, 'protaganist': 1, 'protech': 1, 'malfunction': 1, 'jacuzzi': 1, 'portrayle': 1, 'frances': 6, 'megatron': 1, 'starscream': 1, 'deceptions': 1, 'sofia': 5, 'johansen': 1, '23': 3, 'lorraine': 1, 'insisted': 2, 'studio': 7, 'drop': 1, 'upper': 4, 'picks': 3, 'rigs': 1, 'squalid': 1, 'succession': 1, 'bars': 2, 'motels': 1, 'pumpkin': 4, 'sitting': 1, 'knocking': 1, 'pittsburgh': 2, 'iran': 6, 'amidst': 2, 'crumbling': 1, 'cagney': 6, 'hoodlum': 2, 'ranks': 2, 'underworld': 10, 'cohan': 2, 'warner': 3, 'bros': 4, 'wonder': 2, 'mastrantonio': 1, 'easy': 5, 'trains': 2, 'rebelled': 1, 'donna': 3, 'reed': 4, 'elwood': 1, 'dowd': 1, 'lindbergh': 2, 'creepiest': 1, 'gaynor': 1, 'brien': 4, '1927': 4, 'ghibli': 7, 'relentless': 1, 'gabin': 1, 'stroheim': 1, 'lundegaard': 1, 'henchmen': 3, 'bungling': 1, 'persistent': 2, 'marge': 2, 'gunderson': 1, 'ingenuity': 1, 'bravery': 1, 'aviation': 1, 'jodelle': 1, 'ferland': 1, 'roasted': 1, 'bonfire': 1, 'roast': 1, 'oven': 1, 'iris': 1, 'merrick': 1, 'sympathetic': 1, 'lautner': 6, 'hatter': 3, 'johny': 1, 'lends': 1, 'bought': 1, 'spans': 2, 'sales': 3, 'fozzy': 1, 'lieutenant': 2, 'outposts': 1, 'exist': 5, 'bullies': 2, 'reappear': 1, 'impending': 1, 'disagrees': 1, 'domestic': 1, 'nearly': 4, 'decaprio': 2, 'peoples': 2, 'thoughts': 1, 'zeta': 8, 'tormented': 2, 'studying': 1, 'persecution': 2, 'goof': 1, 'screened': 2, 'bbc': 3, 'april': 1, 'root': 1, 'theatrically': 1, 'marines': 3, 'pendleton': 2, 'educating': 1, 'marine': 12, 'thank': 1, 'sneak': 1, 'march': 2, 'teddy': 13, 'kurtz': 1, 'buck': 2, 'watson': 4, 'misfit': 3, 'apocalyotic': 1, '2020': 1, 'glen': 2, 'homicide': 3, 'melodramatic': 1, 'let': 7, 'allegedly': 2, '15': 2, 'pounds': 1, 'fassbender': 7, 'cannes': 2, 'festival': 2, 'generation': 1, 'penelope': 4, 'cruz': 4, '30': 8, 'artificially': 1, 'inseminated': 1, 'beau': 1, 'involve': 3, 'micheal': 5, 'towns': 1, 'goods': 2, 'entrepreneurs': 2, 'wide': 1, 'ne': 1, 'yo': 1, 'rival': 13, 'skeleton': 3, 'software': 2, 'writers': 2, 'account': 3, 'clicking': 1, 'heels': 3, 'pal': 2, 'transport': 2, 'hazardous': 1, 'materials': 2, 'proper': 1, 'handling': 2, 'turban': 1, 'ashby': 1, 'paramount': 2, 'interesting': 5, 'lineman': 1, 'repairman': 1, 'thousand': 2, 'polish': 6, 'refugees': 2, 'holocost': 2, 'surface': 2, 'lurking': 1, 'bedroom': 1, 'closets': 1, 'specific': 3, 'staham': 1, 'steak': 1, 'lunch': 2, 'chic': 1, 'gowns': 1, 'poisoning': 4, 'strikes': 4, 'unladylike': 1, 'behavior': 1, 'scouting': 1, 'assassins': 6, 'await': 1, 'terroist': 1, 'voicing': 4, 'wrangling': 1, 'chauffeur': 1, 'makaws': 2, 'taker': 1, 'childs': 1, 'pillow': 1, 'unbelievable': 1, 'boots': 2, 'sassy': 1, 'mccully': 1, 'culcan': 1, 'rowlings': 3, 'trash': 4, 'organizing': 1, 'louie': 3, 'caped': 1, 'crusader': 2, 'faithful': 1, 'caine': 3, 'comer': 1, 'carriage': 2, 'slipper': 5, 'godmother': 1, 'happily': 2, '1863': 2, 'caan': 1, 'toll': 2, 'booth': 1, 'breaks': 8, 'retrain': 1, 'nemesis': 2, 'guitar': 1, 'sherif': 1, 'reboot': 7, 'climb': 1, 'walls': 4, 'webs': 1, 'villan': 1, 'insert': 1, 'wished': 1, 'ein': 1, 'joker': 3, 'steamy': 1, 'kathleen': 4, 'turner': 5, 'sources': 1, 'ernest': 3, 'borgnine': 1, 'prey': 1, 'albert': 3, 'cazale': 1, 'klaus': 1, 'kinski': 1, 'guinness': 2, 'merchant': 1, 'ivory': 1, 'pkd': 1, 'news': 5, 'paddled': 1, 'canoe': 2, 'prior': 2, 'ironic': 2, 'cinema': 3, 'bankruptcy': 1, 'jokingly': 2, 'osca': 1, 'drink': 3, 'nicole': 3, 'kidman': 3, 'satine': 1, 'baz': 2, 'luhrmann': 1, 'nigel': 2, 'camelot': 2, 'cherie': 1, 'lunghi': 1, 'guenevere': 1, 'refereed': 1, 'cdc': 1, 'okay': 1, 'wartime': 1, 'overwhelmed': 1, 'cousins': 1, 'villains': 2, 'splash': 1, 'transition': 3, 'fritz': 3, 'lang': 5, 'silverware': 1, 'hasbro': 2, 'sixteenth': 1, 'rewrites': 1, 'gambling': 3, 'scam': 2, 'idealist': 3, 'barges': 2, 'realm': 4, 'copy': 1, 'ares': 1, 'hades': 5, 'lorre': 5, 'libbing': 1, 'annoyance': 1, 'ostrum': 1, 'bucket': 2, 'weller': 3, 'reassembled': 1, 'indestructible': 1, 'verhoeven': 2, 'photographer': 4, 'kincaid': 2, 'wanders': 4, 'francesca': 1, 'cliint': 1, 'please': 2, 'dubbed': 1, 'racecar': 1, 'campaign': 4, 'succumbs': 1, 'lust': 3, 'ermey': 2, 'personally': 4, 'supervised': 2, 'recreation': 1, 'parris': 1, 'upstart': 1, 'producer': 7, 'accepts': 5, 'challenge': 2, 'reviving': 1, 'daytime': 1, 'boom': 3, 'mancini': 1, 'remind': 1, 'academic': 1, 'probation': 1, 'preparatory': 1, 'fielding': 1, 'rex': 6, 'eliza': 1, 'doolittle': 1, 'anouilh': 1, 'faux': 1, 'banging': 1, 'shearer': 1, 'mckean': 1, 'derek': 3, 'jnr': 1, 'moriaty': 1, 'clothes': 1, 'natural': 1, 'lincoln': 4, 'probably': 1, 'notable': 1, 'serling': 1, 'memoirs': 4, 'jeopardy': 1, 'streets': 7, 'bernstein': 1, 'livingston': 4, 'lampoons': 1, 'wise': 7, 'cracking': 2, 'accolades': 1, 'math': 2, 'hoard': 1, 'sell': 1, 'scandalous': 1, 'arena': 3, 'london': 6, 'cleef': 1, 'eli': 3, 'wallach': 2, 'mcgovern': 1, 'turturro': 3, 'sherwood': 1, 'surrounds': 1, 'merry': 2, 'cabrini': 1, 'projects': 1, 'traverse': 1, 'creators': 3, 'overcoming': 2, 'adversity': 2, 'cameo': 1, 'eleven': 3, 'belongs': 2, 'woke': 1, 'explore': 1, 'stampede': 1, 'sissy': 3, 'spacek': 3, 'spencer': 6, 'lawyers': 2, 'opposing': 2, 'imagery': 1, 'environment': 4, 'monoliths': 1, 'poked': 1, 'comedians': 2, 'preparations': 1, 'display': 1, '1938': 6, 'carrel': 5, 'fanning': 2, 'lukewarm': 1, 'reviews': 2, 'hersheys': 1, 'strongly': 1, 'affinity': 1, 'awful': 2, 'eclectic': 1, 'hays': 1, 'jabbar': 1, 'staute': 1, 'thrills': 6, 'charley': 2, 'brewster': 2, 'guesses': 1, 'dandrige': 1, 'string': 2, 'profits': 1, 'morph': 1, 'miner': 1, 'extraterrestrial': 5, 'passed': 3, 'sidney': 5, 'prescott': 2, 'thanks': 2, 'visited': 9, 'painless': 1, 'coincide': 2, 'choices': 2, 'accidental': 3, 'mix': 4, 'identical': 3, 'plaid': 1, 'overnight': 2, 'bags': 2, 'journal': 3, 'milne': 1, 'hundred': 1, 'acre': 1, '16': 6, 'slayer': 1, 'monstrosities': 1, 'organized': 3, 'dynasty': 2, 'transfers': 2, 'clandestine': 1, 'aptly': 1, 'realizing': 1, 'faked': 1, 'victory': 3, 'magazine': 1, 'columnist': 1, 'gladys': 2, 'taber': 1, 'stillmeadow': 1, 'charlestown': 1, 'chiefs': 1, 'coached': 2, 'reggie': 2, 'dunlop': 1, 'kubreck': 1, 'clock': 1, 'borrowing': 1, 'malcom': 1, 'commits': 2, 'doorknob': 1, 'comback': 1, 'speaks': 5, 'yiddish': 1, 'blacks': 1, 'deceiving': 1, 'economist': 1, 'delusions': 2, 'detailed': 1, 'diner': 3, 'lykans': 1, 'haired': 2, 'stumbled': 1, 'mayor': 1, 'sleigh': 2, 'wonderland': 2, 'vs': 3, 'document': 2, 'bike': 5, 'zuckerberg': 4, 'dorm': 3, 'boyfriend': 6, 'imagines': 1, 'whether': 1, 'penis': 1, 'kissing': 1, 'staying': 2, 'carried': 2, 'peacock': 3, 'conquering': 2, 'prejudice': 1, 'jurors': 1, 'slowly': 4, 'trial': 3, 'leap': 2, 'guard': 6, 'arriving': 1, 'forman': 3, 'enough': 3, 'greatly': 1, 'punching': 1, 'hammil': 1, 'continue': 1, 'sweeping': 1, 'unlinked': 1, 'fend': 2, 'oversees': 1, 'haunts': 2, 'hot': 6, 'inspiration': 3, 'prolific': 2, 'stupidest': 1, 'commercials': 1, 'rapeable': 1, 'orlando': 2, 'bloom': 2, 'viggo': 2, 'mortensen': 2, 'serie': 2, 'umbrellas': 1, 'puddle': 1, 'hopping': 1, 'herb': 1, 'colorful': 7, 'impostor': 1, 'mechanical': 4, 'ruining': 1, 'seals': 5, 'labeouf': 3, 'curtain': 1, 'voted': 5, 'premiere': 1, 'dolittle': 1, 'raid': 1, 'badges': 6, 'ai': 2, 'stinking': 1, '36': 1, 'consecutive': 1, 'pidgeon': 1, 'tempest': 1, 'laurents': 1, 'gangs': 15, 'disillusioned': 1, 'palme': 1, 'palm': 1, '227': 1, 'lennon': 2, 'chrysanthemum': 1, 'opened': 3, 'franklin': 1, 'schaffner': 1, 'bleak': 1, 'tis': 1, 'scratch': 1, 'galahad': 1, 'lancelot': 4, 'bedevere': 1, 'warlord': 4, 'montand': 1, 'marcel': 1, 'pagnol': 1, 'collaborations': 1, 'starling': 1, 'lector': 3, 'credits': 1, 'offering': 1, 'launched': 7, 'depiction': 2, 'harsh': 3, 'realities': 3, 'dueling': 2, 'hooded': 2, 'jonze': 1, 'averts': 1, 'jazz': 3, 'latifa': 1, 'nabbing': 1, 'aileen': 2, 'wuornos': 1, 'daytona': 1, 'crossover': 1, 'rumored': 1, 'published': 5, 'chronological': 1, 'poverty': 1, 'tsa': 1, 'employee': 7, 'somehow': 1, 'contemporary': 2, 'vik': 2, 'muniz': 2, 'boost': 1, 'malkovich': 5, 'timothy': 4, 'olyphant': 2, 'iowan': 1, 'violently': 1, 'levi': 2, 'labeauf': 1, 'portis': 2, 'disasters': 1, 'greece': 1, 'remaining': 1, 'maids': 3, 'pegg': 2, 'gabriele': 2, 'muccino': 2, 'coaching': 1, 'rap': 1, 'wu': 2, 'tang': 1, 'clan': 2, '11': 6, 'reunites': 1, 'worked': 2, 'beetlejuice': 1, 'karl': 2, 'reconstruct': 1, 'duo': 2, 'vow': 1, 'philandering': 2, 'ufc': 1, 'wookies': 1, 'tan': 2, 'cels': 1, 'misses': 3, 'replica': 1, 'departure': 1, 'orphaned': 5, 'kobe': 1, 'bombing': 1, 'enforcement': 2, 'boundaries': 1, 'troublesome': 1, 'gehrig': 1, 'stoically': 1, 'finishes': 1, 'namesake': 1, 'visual': 2, 'afterlife': 2, 'mathmetician': 1, 'adjusted': 1, 'inflation': 1, 'siegel': 1, 'abandons': 3, 'blatty': 2, 'entering': 4, 'flailing': 1, 'claus': 7, 'deemed': 2, 'arguing': 1, 'indeed': 1, 'motley': 1, 'active': 2, 'discontentment': 1, 'opens': 2, 'shattering': 1, 'idealistic': 1, 'utilizing': 1, 'parliamentary': 1, 'fued': 1, 'orders': 3, 'strike': 1, 'soviet': 3, 'avoid': 1, 'apocalypse': 3, 'rodney': 3, 'dangerfield': 3, 'gophers': 1, 'cleavon': 2, 'acceptance': 2, 'wadiya': 1, 'discussion': 1, 'hayden': 2, 'panettiere': 2, 'fargo': 1, 'feels': 3, 'herman': 3, 'detailing': 2, 'whaling': 1, 'vessel': 1, 'respective': 1, 'il': 3, 'buono': 1, 'brutto': 1, 'cattivo': 1, 'orphanage': 1, 'picnic': 1, 'baskets': 1, 'hanna': 1, 'britsh': 1, 'renamed': 1, 'uk': 2, 'traces': 3, 'roots': 2, 'lester': 1, 'ringo': 1, 'starr': 1, 'aspects': 2, 'technically': 1, 'podracing': 1, 'tournament': 5, 'decoy': 1, 'introduction': 1, 'jar': 3, 'binks': 1, 'deliver': 2, 'packages': 2, 'galifiankis': 2, 'invite': 1, 'losers': 1, 'deployed': 1, 'themepark': 1, 'beggers': 1, 'shakespearian': 2, 'decorations': 1, 'hitchock': 1, 'arthouse': 1, 'wong': 2, 'kar': 2, 'wai': 2, 'spouses': 2, 'clarence': 1, 'faith': 4, 'getchell': 1, 'rutger': 3, 'hauer': 3, 'rhi': 1, 'halmi': 1, 'junior': 1, 'weathers': 1, 'daryl': 4, 'hannah': 4, 'marker': 1, 'stowe': 1, 'plummer': 2, 'morse': 1, 'ants': 2, 'grasshopper': 1, 'satirizing': 1, 'typical': 2, 'unkrich': 1, 'gilroy': 1, 'potter': 2, 'watkins': 1, 'stoller': 1, 'strained': 1, 'continually': 2, 'extended': 2, 'thoroughbred': 1, 'rescued': 2, 'nurtured': 1, 'shipwreck': 1, 'pajamas': 2, 'wear': 4, 'macaw': 4, 'bird': 11, 'blu': 3, 'madly': 1, 'suburbs': 2, 'israeli': 1, 'lewi': 1, 'scarred': 3, 'avenger': 1, 'recruits': 4, 'lack': 1, 'widow': 6, 'obtain': 1, 'element': 2, 'janus': 1, 'glittering': 1, 'cruel': 1, 'michel': 3, 'gondry': 2, 'sharecroppers': 1, 'vivid': 1, 'whovians': 1, 'successfully': 1, 'creative': 2, 'everyday': 2, 'keitel': 5, 'pimp': 2, 'shared': 4, 'wy': 1, 'tone': 1, 'greeting': 2, 'precious': 2, 'toe': 2, 'octopus': 1, 'arms': 1, 'incarnation': 1, 'norse': 4, 'appeal': 1, 'dna': 3, 'wad': 1, 'resin': 1, 'worse': 1, 'biel': 6, 'cairn': 1, 'terrier': 1, 'dropout': 3, 'investment': 4, 'firm': 4, 'legitimate': 4, 'sounds': 3, 'trucks': 1, 'ambush': 1, 'constantly': 4, 'constant': 2, 'hyped': 2, 'bone': 1, 'steller': 1, 'realistically': 1, 'propose': 1, 'outlawed': 2, 'delinquent': 1, 'therapy': 1, 'effort': 5, 'orge': 2, 'lower': 3, 'steamliner': 1, 'hosted': 1, 'celebrates': 1, 'idiotic': 1, 'extraordinarily': 1, 'manipulative': 4, 'roguish': 2, 'reconstruction': 2, 'elemental': 1, '1897': 1, 'optics': 1, 'reverse': 1, 'damage': 1, 'reign': 1, 'israel': 1, 'loss': 3, 'wil': 2, 'fixing': 1, 'automobiles': 2, 'concentration': 4, 'precocious': 2, 'prancing': 1, 'snapping': 1, 'wailing': 1, 'divided': 1, 'album': 3, 'sessions': 1, 'bloomfield': 1, 'kooper': 1, 'stills': 1, 'educate': 1, 'walberg': 1, 'numbskull': 1, 'hysterical': 2, 'mistakes': 2, 'hitters': 2, 'potential': 1, 'explored': 3, 'research': 3, 'chemist': 1, 'dig': 1, 'parish': 2, 'sites': 1, 'dreyfus': 2, 'mclain': 1, 'aurora': 2, 'havasham': 1, 'describing': 1, 'successes': 1, 'blunders': 1, 'drummer': 1, 'confrontation': 2, 'natives': 2, 'topped': 1, 'avatar': 2, 'phillp': 1, 'jrr': 3, 'thats': 1, 'hanson': 4, 'curious': 2, 'maintain': 1, 'demme': 1, 'previous': 4, 'terrestrials': 1, 'risque': 1, 'rodgers': 1, 'hammerstein': 4, 'stansilaw': 1, 'lem': 1, 'avengers': 2, 'courtney': 1, 'terrors': 1, 'fears': 1, 'accepting': 2, 'throats': 1, 'weddings': 3, 'delroy': 1, 'lindo': 1, 'joint': 2, 'bedford': 2, 'stuyvesant': 1, 'felon': 2, 'fashionable': 1, 'manner': 2, 'aims': 1, 'kubrik': 1, 'defined': 3, 'loan': 2, 'collector': 2, 'whoville': 3, 'glowing': 2, 'user': 1, 'utilize': 1, 'toad': 1, 'smooches': 1, 'newest': 2, 'wildly': 2, 'billed': 2, 'v': 6, 'instinct': 1, 'guide': 4, 'perilous': 1, 'miles': 4, 'backpackers': 1, 'slovak': 1, 'promises': 2, 'hedonistic': 1, 'expectations': 2, 'gentlemen': 2, 'captive': 2, 'nemo': 1, 'narrates': 2, 'tribe': 6, 'unusually': 1, 'sharp': 2, 'celebrate': 4, 'pondering': 1, 'imitate': 1, 'stint': 1, 'stated': 1, 'paradise': 3, 'venezuela': 1, 'helium': 1, 'bikers': 3, 'housekeeping': 1, 'cliff': 1, 'forget': 2, 'phil': 2, 'stu': 4, 'doug': 2, 'stephanie': 5, 'plum': 1, 'assignment': 1, 'brooding': 1, 'vincente': 3, 'romanticized': 1, 'madea': 1, 'wehat': 1, 'gore': 1, 'verbinski': 1, 'peasant': 1, 'bandit': 1, 'tensions': 3, 'hottest': 1, 'jumping': 2, 'pheonix': 1, 'maybe': 2, 'discharged': 1, 'alps': 1, 'energy': 1, 'elizabethan': 1, 'versus': 1, 'weak': 2, 'myself': 1, 'luckiest': 1, 'investigated': 1, 'plains': 1, 'dust': 4, 'bowl': 3, 'bogie': 1, 'bacall': 2, 'hurricane': 2, 'wrapped': 1, 'antagonists': 1, 'addiction': 2, 'forms': 4, 'leisurely': 1, 'loopy': 1, 'shift': 1, 'shape': 4, 'liquid': 1, 'amazon': 2, 'infraction': 1, 'misfits': 1, 'joining': 2, 'fireman': 1, 'manages': 1, 'stoppard': 1, 'consumed': 2, 'jealousy': 3, 'elaborate': 2, 'trap': 1, 'homer': 1, 'odyssey': 2, 'pennsylvania': 1, 'owns': 3, 'rachmaninoff': 1, 'background': 2, 'visions': 1, 'mesa': 1, 'sw': 1, 'mild': 1, 'mannered': 1, 'cartoons': 2, 'degree': 2, 'ruin': 1, 'hustler': 1, 'matches': 1, 'gleason': 2, 'yeats': 1, 'cormac': 2, 'education': 1, 'flower': 6, 'shaw': 4, 'pygmalion': 3, 'integration': 1, 'defeating': 2, 'valueless': 1, 'speck': 2, 'participates': 1, 'austen': 1, 'nonexistent': 1, 'kaiser': 1, 'sosa': 1, 'buttons': 1, 'sown': 1, 'preform': 1, 'float': 2, 'authorities': 1, 'pseudo': 4, 'bottom': 3, 'eighties': 2, 'sweeps': 1, 'combat': 2, 'sort': 4, 'weakling': 1, 'moral': 3, 'undergoes': 2, 'possess': 2, 'strength': 1, 'recorded': 1, 'visionary': 1, 'contends': 1, 'uncontrolled': 2, 'farming': 4, 'pare': 1, 'lorenz': 2, 'agriculture': 1, 'grimm': 2, 'wooden': 3, 'editor': 3, 'rosalind': 2, 'omler': 1, 'barber': 1, 'carraclough': 1, 'reluctantly': 2, 'wyman': 1, 'resolved': 1, 'humphey': 1, 'prospectors': 2, 'digging': 1, 'lauren': 1, 'tropical': 4, 'lamppost': 1, 'artistic': 2, 'paranoid': 4, 'wanna': 1, 'graziano': 2, 'mcmurray': 1, 'opus': 1, 'paparazzo': 2, 'scantily': 1, '007': 1, 'convent': 1, 'governess': 7, 'naval': 3, 'michelangelo': 2, 'pope': 2, 'julius': 2, 'plight': 2, 'bridgewater': 1, 'criminally': 2, 'price': 2, 'casting': 1, 'vastness': 1, 'telly': 1, 'savalas': 1, 'medium': 2, 'jaws': 1, 'gilliman': 1, 'unstable': 3, 'dramatization': 4, 'pacific': 2, 'cimino': 1, 'roulette': 2, 'stormin': 1, 'mawwage': 1, 'eliot': 3, 'karen': 2, 'fondness': 1, 'nosed': 3, 'griswold': 4, 'flashbacks': 2, 'winnebago': 1, 'jareth': 1, 'connelly': 5, 'wright': 2, 'patinkin': 2, 'corps': 1, 'drill': 1, 'instructor': 1, 'katsuhiro': 1, 'otomo': 1, 'hoskins': 2, 'intermingled': 1, 'solitary': 1, 'outpost': 3, 'bassett': 2, 'mick': 1, 'nichols': 6, 'cabaret': 1, 'homespun': 1, 'baldwin': 1, 'hogan': 2, 'elijah': 6, 'holm': 1, 'study': 3, 'mansion': 4, 'camping': 1, 'terrifying': 1, 'continued': 1, 'wuxia': 3, 'prodigy': 1, 'bush': 1, 'administration': 1, 'emperors': 1, 'seymor': 2, 'recounting': 1, 'washed': 5, 'braddock': 4, 'amongst': 2, 'polley': 1, 'engineers': 3, 'witherspoon': 11, 'malfeasance': 1, 'ho': 1, 'balthazar': 1, 'inherit': 2, 'alamos': 1, 'lise': 1, 'friedman': 2, 'ceil': 1, 'physical': 8, 'terapist': 1, 'tanner': 1, 'respectively': 4, 'mirren': 2, 'ops': 5, 'realization': 3, 'bobby': 1, 'floor': 5, 'payments': 2, 'secrecy': 2, 'heigl': 2, 'antin': 1, 'carrying': 2, 'curator': 1, 'anniston': 6, 'switches': 1, 'samples': 1, 'scifi': 1, 'greg': 6, 'strause': 2, 'abduction': 2, 'poems': 1, 'diving': 1, 'zooey': 1, 'deschanel': 1, 'mortimer': 1, 'chu': 1, 'international': 3, 'premier': 1, 'occurs': 1, 'suspects': 3, 'brewer': 1, 'largely': 1, 'lurie': 1, 'alexander': 3, 'skarsg': 1, 'ginnifer': 6, 'goodwin': 6, 'hudson': 5, 'cutcher': 1, 'reprising': 2, 'blaze': 1, 'morphs': 1, 'incarnations': 2, 'bethany': 2, 'keeping': 2, 'irvine': 2, 'thewlis': 3, 'extraordinary': 2, 'grants': 2, 'desires': 1, 'messenger': 2, 'survives': 2, 'walters': 2, 'brenda': 1, 'dea': 2, 'iron': 6, 'thor': 3, 'indecisive': 1, 'pot': 3, 'kingpin': 3, 'unleashed': 1, 'collegiate': 2, 'freshman': 1, 'grandparents': 1, 'granchildren': 1, 'ban': 2, 'mothers': 2, 'inner': 3, 'underachieving': 1, 'elisabeth': 5, 'shue': 5, 'dollar': 1, 'kitsch': 2, 'zemekis': 1, 'commercial': 2, 'transformed': 7, 'antebellum': 1, 'charter': 2, 'broadcast': 2, 'representative': 1, 'li': 6, 'dolph': 1, 'lundgren': 1, 'randy': 1, 'cotor': 1, 'crews': 2, 'horatio': 1, 'mistake': 2, 'hitchcok': 1, 'contained': 3, 'stabbed': 1, 'hiromasa': 1, 'yonebayashi': 1, 'borrowers': 1, 'polluted': 1, 'bookworm': 1, 'slinky': 1, 'potato': 1, 'gained': 3, 'billing': 1, 'portion': 1, 'repunzal': 1, 'adrian': 3, 'hold': 5, 'celebrities': 1, 'pug': 1, 'allegations': 1, 'capsized': 1, 'dazzling': 2, 'biological': 1, 'jersey': 1, 'category': 1, 'delightful': 1, 'enjoy': 1, 'lever': 1, 'bb': 3, 'angry': 3, 'hawks': 4, 'paired': 3, 'conneley': 1, 'bowe': 1, 'artifact': 3, 'retake': 1, 'litter': 2, 'hough': 2, 'rufus': 2, 'iowa': 4, 'cornfield': 1, 'cushing': 1, 'helsing': 1, 'bloodsucker': 1, 'lam': 1, 'lone': 4, 'safety': 3, 'feig': 1, 'alumni': 1, 'nbc': 1, 'caesar': 2, 'jared': 1, 'hess': 1, 'eponymous': 2, 'meatloaf': 2, 'fishnet': 1, 'stockings': 1, 'culturally': 1, 'significant': 1, 'alert': 1, 'current': 4, 'cronenberg': 1, 'goldbloom': 1, 'teleportation': 1, 'mamet': 1, 'professionally': 1, 'hideous': 2, 'absence': 1, 'timon': 1, 'pumba': 1, 'offspring': 1, 'difficulties': 3, 'vignettes': 2, 'groundbreaking': 2, 'gylenhal': 2, 'doo': 1, 'wop': 1, 'newton': 2, 'preppy': 2, 'rogue': 5, 'replicants': 3, 'roth': 1, 'charecterized': 1, 'reliving': 1, 'youths': 1, 'catching': 2, 'wait': 2, 'roundtree': 1, 'isaac': 1, 'hayes': 2, 'rubber': 1, 've': 6, 'began': 2, 'climax': 1, 'mutant': 2, 'beginnings': 2, 'solver': 1, 'skit': 1, 'chastain': 1, 'clueless': 1, 'octavia': 1, 'guiding': 1, 'strathairn': 1, 'specially': 1, 'antiquated': 1, 'sheik': 1, 'infatuated': 1, 'englishwoman': 1, 'abducts': 1, 'saharan': 1, 'scimitars': 1, 'acrobatics': 1, 'vulgar': 1, 'tragedies': 2, 'passangers': 1, 'redgrave': 1, 'glenda': 1, 'harrelson': 1, 'mcdonagh': 2, 'bruges': 2, 'rallies': 1, 'oppressive': 1, 'ratched': 1, 'deranged': 3, 'anchor': 4, 'ravings': 1, 'revelations': 1, 'media': 3, 'profit': 1, 'generate': 1, 'scaring': 1, 'sorkin': 1, 'legal': 3, 'recourse': 1, 'lavish': 1, 'gasoline': 1, 'grammy': 1, 'hefley': 2, 'concentrates': 1, 'lesson': 2, 'walken': 2, 'torro': 1, 'snakes': 1, 'itzhak': 1, 'stern': 2, 'national': 1, 'crop': 2, 'dusting': 1, 'toungue': 1, 'ebeneezer': 1, 'gunner': 2, 'sociopath': 1, 'gein': 1, 'pursues': 1, 'checked': 1, 'reid': 2, 'angelos': 2, 'benecio': 2, 'modernization': 1, 'carol': 6, 'donner': 3, 'cinematography': 2, 'winstead': 2, 'lowe': 1, 'lacey': 1, 'chabert': 1, 'trachtenberg': 1, 'cloke': 1, 'andrea': 1, 'bomback': 1, '1921': 1, 'lasky': 1, 'melford': 1, 'valentino': 1, 'agnes': 1, 'ayres': 1, 'adolphe': 1, 'menjou': 1, '1932': 5, 'sideshow': 1, 'performers': 2, 'composed': 1, 'rko': 1, 'tailed': 1, 'cotten': 1, 'alida': 1, 'valli': 1, 'trevor': 1, 'particularly': 1, 'atmospheric': 1, 'entre': 1, 'les': 1, 'morts': 1, 'boileau': 3, 'narcejac': 2, 'jerome': 1, 'lerner': 1, 'loewe': 1, 'yakuza': 2, 'seijun': 1, 'suzuki': 1, 'tetsuya': 1, 'watari': 1, 'reformed': 2, 'tetsu': 1, 'roam': 1, 'execution': 1, '1913': 1, 'lardner': 1, 'hooker': 1, 'mash': 1, 'doctors': 3, 'boorman': 2, 'wolfe': 1, 'speed': 2, 'aeronautical': 1, 'edwards': 2, 'mercury': 1, 'manned': 1, 'spaceflight': 1, 'burlinson': 1, 'williamson': 1, 'ballard': 1, 'autobiographical': 2, 'puyi': 1, 'peploe': 1, 'zwick': 1, 'volume': 1, 'laura': 2, 'hillenbrand': 1, 'commentator': 1, 'cressida': 1, 'cowell': 1, 'mechner': 1, 'boaz': 2, 'yakin': 2, 'miro': 1, 'carlo': 1, 'newell': 1, 'bruckheimer': 1, 'edited': 2, 'stamm': 1, 'thekla': 1, 'reuten': 1, 'violante': 1, 'placido': 1, 'irina': 1, 'bj': 1, 'rklund': 1, 'paolo': 1, 'bonacelli': 1, 'jo': 2, 'jardim': 2, 'harley': 1, 'cooperation': 1, 'scavengers': 1, 'recyclables': 1, 'gramacho': 1, 'landfills': 1, 'serving': 4, 'metropolis': 1, 'rio': 5, 'janeiro': 2, 'rockwell': 2, 'kenny': 1, 'florian': 1, 'henckel': 1, 'donnersmarck': 1, 'yuh': 1, 'nelson': 4, 'whitesell': 1, 'fogel': 1, 'rhymer': 1, 'dystopia': 1, 'bettany': 4, 'wilde': 2, 'rosenberg': 1, 'prosthetic': 1, 'fogelman': 1, 'picasso': 1, 'wenk': 2, 'kaufman': 2, 'agosto': 1, 'payday': 1, 'bibi': 1, 'andersson': 1, 'liv': 1, 'ullmann': 1, 'gig': 1, 'excorcist': 1, 'locking': 1, 'hoards': 1, 'nathanson': 1, 'potentially': 1, 'filthy': 1, 'thirteen': 1, 'placed': 3, 'avi': 1, 'gibsons': 1, 'raging': 3, 'bull': 2, 'schrader': 1, 'mardik': 1, 'memoir': 2, '2029': 1, 'biehn': 2, 'palma': 3, 'leung': 2, 'ephron': 2, 'malick': 1, 'exceptional': 1, 'corbett': 1, 'miller': 6, 'clay': 2, 'bana': 2, 'toni': 2, 'collette': 2, 'nickelodeon': 1, 'disabled': 3, 'bennett': 1, 'beane': 3, 'derailing': 1, 'subsequent': 1, 'trank': 1, 'telekinetic': 1, 'hires': 6, 'homeless': 3, 'breasts': 1, 'hilarity': 2, 'nell': 1, 'materialistic': 1, 'boob': 1, 'baking': 1, 'warming': 2, 'arch': 3, 'ewan': 8, 'macgregor': 3, 'quinlan': 1, 'civilization': 1, 'stays': 4, 'willie': 3, 'hopeful': 1, 'scandals': 1, 'cultivates': 1, 'contracts': 1, 'scenario': 1, 'brains': 1, 'possibly': 4, 'magneto': 4, 'druglords': 1, 'cyborgs': 1, 'notice': 1, 'mma': 4, 'sturges': 1, 'residents': 1, 'fact': 2, 'cheung': 1, 'expressionist': 1, 'brigitte': 1, 'helm': 1, 'fancy': 2, 'donation': 1, 'museum': 3, 'leopard': 1, 'onset': 2, 'segments': 2, 'confronted': 1, 'hobo': 3, 'commander': 1, 'tempted': 2, 'railway': 2, 'mime': 1, 'thwarted': 1, 'entertainer': 3, 'declares': 2, 'occupied': 3, 'territory': 2, 'placement': 1, 'uprising': 1, 'bligh': 1, 'sailing': 1, 'tahiti': 1, 'benjamin': 3, 'mia': 1, 'farrow': 2, 'troupe': 3, 'pryce': 2, 'bureaucrat': 2, 'bureaucracy': 1, '96': 1, 'malls': 1, 'confront': 1, 'satirizes': 2, 'bands': 1, 'swimming': 5, 'cocoons': 1, 'youthful': 1, 'chapionship': 1, 'resurrected': 1, 'patric': 1, 'kiefer': 1, 'sutherland': 2, 'battles': 8, 'showcasing': 1, 'randall': 3, 'dale': 2, 'sentenced': 1, 'willian': 1, 'munny': 1, 'patricia': 2, 'arquette': 1, 'slackers': 1, 'linear': 1, 'stores': 1, 'randal': 1, 'liberate': 1, 'pretend': 1, 'recounts': 1, 'landings': 2, 'morre': 1, 'polansky': 1, 'smuggler': 3, 'luc': 4, 'besson': 4, 'mercenary': 1, 'zellwegger': 1, 'gould': 1, 'benches': 1, 'mvp': 1, 'grades': 1, 'infernal': 1, 'ivan': 3, 'emilie': 2, 'ravin': 3, 'markle': 1, 'freight': 1, 'loretta': 2, 'devine': 1, 'goode': 1, 'ireland': 2, 'craigh': 1, 'kosinski': 1, 'redform': 1, 'surratt': 1, 'robyn': 1, 'doherty': 1, 'hiccup': 1, 'desertto': 1, 'conceived': 1, 'abe': 1, 'abstract': 1, 'leonoardo': 1, 'undefeated': 2, 'warrant': 1, 'millar': 1, 'tracey': 4, 'edmonds': 1, 'misfortunes': 2, 'serve': 1, 'laz': 1, 'alonso': 1, 'folk': 1, 'chandler': 2, 'exploration': 1, 'janiero': 1, '51': 1, 'connick': 3, 'ashley': 3, 'sturgess': 3, 'peeing': 2, 'greta': 2, 'gerwig': 1, 'navy': 9, 'briefcase': 2, 'replacing': 1, 'treks': 1, 'sorta': 1, 'magee': 1, 'suraj': 1, 'sharma': 1, 'afterwards': 1, 'jonny': 3, 'rhianna': 1, 'fleet': 1, 'ships': 3, 'squadron': 1, 'timur': 1, 'bekmambetov': 1, 'hutcherson': 2, 'lycans': 1, 'mcg': 2, 'vying': 1, 'judges': 1, 'mather': 1, 'leger': 1, 'radcliff': 2, 'lookout': 1, 'slayed': 1, 'voldemort': 2, 'ordeals': 1, 'submariners': 1, 'morgana': 1, 'le': 1, 'darkest': 1, 'sorceress': 2, 'popeye': 1, 'rough': 2, 'irresponsible': 1, 'sail': 2, 'wardrobe': 4, 'tippi': 1, 'hedren': 1, 'woodsboro': 1, 'birthday': 4, 'practice': 1, 'kungfu': 1, 'struck': 3, 'houses': 2, 'borrow': 1, 'irons': 1, 'repunzel': 1, 'cost': 1, 'siblings': 3, 'bears': 2, 'selena': 3, 'curls': 1, 'jessie': 2, 'statistical': 1, 'analysis': 2, 'larger': 2, 'luhrman': 1, 'misadventures': 5, 'spec': 1, 'acne': 1, 'tautou': 1, 'shut': 2, 'notoriously': 1, 'likes': 8, 'safari': 1, 'midgets': 1, 'bookish': 2, 'eventual': 1, 'light': 4, 'sabers': 1, 'animalistic': 1, 'bah': 3, 'humbug': 3, 'grouchy': 2, 'wakes': 7, 'shakepeare': 1, 'intricate': 1, 'grind': 1, 'stops': 1, 'promoted': 1, 'efficiency': 1, 'experts': 1, 'interrupts': 2, 'impersonates': 1, 'dutch': 2, 'mo': 1, 'channel': 1, 'cycle': 3, 'ears': 3, 'fawn': 1, 'vil': 1, 'jonathon': 1, 'warthog': 1, 'stepfamily': 1, 'royalty': 2, 'voodoo': 2, 'shaman': 1, 'gains': 1, 'appreciation': 1, 'furniture': 1, 'dishes': 1, 'heralded': 1, 'ascension': 2, 'kingship': 1, 'silverman': 1, 'surround': 1, 'minimal': 1, 'spreads': 1, 'epidemically': 1, 'laso': 1, 'monty': 2, 'jouney': 1, 'hairy': 1, 'feet': 1, 'rakish': 1, 'rhett': 1, 'picked': 1, 'schoolmates': 1, 'verge': 1, 'tools': 1, 'austria': 1, 'swimmers': 2, 'poland': 3, 'drives': 1, 'batmobile': 1, 'arrive': 1, 'goth': 1, 'sake': 1, 'pressure': 2, 'rife': 1, 'opts': 1, 'erased': 2, 'wo': 3, 'locales': 1, 'invade': 2, 'colored': 1, 'slaves': 3, 'homes': 3, 'luther': 1, 'judi': 1, 'undertone': 1, 'boulders': 1, 'inmate': 4, 'clues': 2, 'disappearances': 1, 'flash': 1, 'mobs': 1, 'handheld': 1, 'surveillance': 3, 'passage': 1, 'virtual': 5, 'complains': 1, 'asian': 3, 'stoners': 1, 'holloween': 1, 'incredibly': 1, 'enormous': 2, 'preys': 2, 'inspirational': 6, 'spirituality': 1, 'connects': 1, 'directorhood': 1, 'insured': 1, 'chronic': 1, 'tub': 1, 'kasdan': 3, 'seduced': 1, 'tricked': 2, 'object': 1, 'rejuvenate': 1, 'rivals': 4, 'heros': 1, 'racist': 1, 'townfolk': 1, 'financials': 1, 'clue': 5, 'disappeared': 2, 'chocolate': 1, 'painting': 1, 'latin': 1, 'suspension': 2, 'babysits': 3, 'revolt': 4, 'executed': 1, 'creacher': 1, 'pushed': 2, 'limits': 2, 'gamble': 1, 'underneath': 1, 'freelance': 3, 'kicking': 1, 'sparkled': 1, 'clear': 6, 'enamoured': 1, 'hep': 1, 'defenseless': 1, 'fumble': 1, 'petty': 1, 'hopes': 3, 'safely': 1, 'marshal': 5, 'rope': 1, 'playground': 1, 'creek': 1, 'ideal': 2, 'biography': 2, 'abused': 2, 'beagle': 1, 'amateur': 1, 'file': 1, 'evolved': 2, 'chimps': 1, 'dominate': 1, 'sniper': 1, 'fields': 2, 'beth': 1, 'britt': 3, 'teaming': 2, 'cow': 1, 'boobs': 1, 'mitch': 2, 'gulag': 2, 'escapees': 2, '4000': 1, 'overland': 1, 'dressing': 2, 'twisp': 1, 'sight': 2, 'sheeni': 1, 'fouth': 1, 'braving': 1, 'quint': 2, 'lago': 1, 'othello': 1, 'manipulate': 2, 'explorer': 3, 'mushrooms': 1, 'enter': 2, 'tied': 3, 'highschool': 1, 'sweethearts': 2, 'ma': 1, 'greedy': 4, 'monologue': 2, 'makeup': 2, 'mcfarland': 1, 'harms': 1, 'suspected': 2, 'holds': 4, 'meetings': 1, 'buffalo': 1, 'ewoks': 2, 'chests': 1, 'demigod': 1, 'persius': 1, '1074': 1, 'jury': 1, 'executioner': 1, 'authority': 1, 'hauling': 1, 'cargo': 1, 'represents': 2, 'conversation': 3, 'sandwiches': 1, 'breathing': 1, 'forming': 4, 'attachment': 1, 'quintuplets': 1, 'solid': 1, 'constructs': 1, 'skerritt': 1, 'incubates': 1, 'meal': 2, 'stupedest': 1, 'weirdest': 1, 'idiots': 1, 'crippled': 1, 'workforce': 1, 'routines': 1, 'explains': 1, 'lackluster': 1, 'diary': 1, 'controlled': 1, 'chavez': 1, 'nations': 2, 'attic': 1, 'incompetent': 1, 'recovers': 1, 'repair': 1, 'chimp': 1, 'troubles': 2, 'georgia': 2, 'solved': 1, 'conspiracy': 4, 'endanger': 1, 'micky': 2, 'progresses': 1, 'trusted': 1, 'shifter': 1, 'tables': 2, 'roughnecks': 1, 'imperials': 1, 'occupation': 3, 'abs': 1, 'sara': 4, 'cent': 1, 'carel': 1, 'advocating': 1, 'hats': 1, 'requirements': 1, 'themsleves': 1, 'shady': 1, 'overenthusiastic': 1, 'racoon': 1, 'atlantic': 1, '1919': 1, 'amnesia': 2, 'obsesses': 1, 'minka': 1, 'reserved': 1, 'seat': 2, 'villian': 2, 'sailors': 2, 'presentable': 1, 'minions': 4, 'adopting': 2, 'bugs': 1, 'underwood': 1, 'cheetahs': 1, 'middler': 1, 'maze': 3, 'else': 3, 'limb': 1, 'sideways': 1, 'puft': 1, 'cabinet': 1, 'hooked': 1, 'inswct': 1, 'jimmie': 1, 'colossal': 1, 'shine': 1, 'cody': 1, 'jarrett': 1, 'saks': 2, 'slater': 1, '1001': 1, 'nights': 2, 'virtuous': 1, 'lords': 3, 'frame': 1, 'whoiver': 1, 'carry': 4, 'turbulent': 1, 'montage': 1, 'baxter': 1, 'dalmations': 1, 'meat': 2, 'bicycle': 2, 'pheiffer': 1, 'critical': 3, 'darling': 1, 'cultural': 1, 'viewed': 1, 'yell': 1, 'bitch': 1, 'thumper': 2, 'restoring': 1, 'cumberbatch': 1, 'hiddleston': 2, 'hauted': 1, 'pleasure': 1, 'holding': 3, 'orgasm': 1, 'scharwtzman': 1, 'gideon': 2, 'directions': 1, 'intro': 1, 'mercilessly': 1, 'cassette': 1, 'raider': 1, 'hulk': 2, 'colony': 2, 'anthropologist': 1, 'gorillas': 2, 'spying': 1, 'sticks': 1, 'microscopic': 1, 'quebec': 1, 'places': 5, 'medical': 5, 'womanizing': 1, 'solves': 2, 'mysteries': 1, 'competes': 1, 'minute': 3, 'chariot': 1, 'delivering': 2, 'rendition': 2, 'roaming': 1, 'tremblay': 1, 'patterned': 1, 'kick': 1, 'sharpen': 1, 'invetrofertilization': 1, 'integrate': 1, 'gruen': 1, 'presumed': 1, 'primarily': 4, 'oregon': 1, 'lied': 1, 'anakin': 2, 'tattooine': 2, 'galaxy': 4, 'waddums': 1, 'skynet': 1, 'resistence': 1, 'byrne': 3, 'advisor': 1, 'singers': 4, 'rockatansky': 1, 'bridget': 1, 'cedric': 1, 'damme': 2, 'norris': 3, 'icons': 1, 'swing': 1, 'gecko': 1, 'dock': 1, 'unions': 1, 'pits': 3, 'odin': 2, 'englishman': 1, 'nph': 1, 'approval': 1, 'vary': 2, 'behave': 1, 'wendelin': 1, 'draanen': 1, 'magento': 2, 'conservative': 4, 'ally': 3, 'bruiser': 1, 'vadar': 2, 'insight': 1, 'lakeside': 1, 'retreat': 3, 'mater': 3, 'grandpa': 1, 'gentile': 1, 'strives': 1, 'feeds': 1, 'reece': 1, 'motion': 9, 'embraces': 1, 'arrogant': 2, 'onscreen': 1, 'shelter': 1, 'regulate': 1, 'lon': 1, 'chaney': 1, 'bubba': 1, 'primary': 1, 'protectors': 1, 'punishment': 1, 'overgrown': 1, 'wreaks': 3, 'carney': 1, 'airborne': 2, 'graves': 1, 'reitman': 2, 'aykroyd': 3, 'harold': 5, 'ramis': 4, 'inbred': 1, 'cannibals': 1, 'pursuit': 3, 'salim': 1, 'akil': 1, 'nineteen': 2, 'stripping': 2, 'guided': 1, 'farmers': 2, 'diseased': 1, 'grips': 1, 'embarking': 1, 'wake': 4, 'spectre': 1, 'inheriting': 1, 'bandidos': 1, 'stinkin': 1, 'ricki': 1, 'clevon': 1, 'egg': 2, 'failure': 3, 'precrime': 1, 'lenny': 1, 'mystics': 1, 'goers': 1, 'scariest': 1, 'oppose': 1, 'kowalski': 1, 'thru': 1, 'compromise': 1, 'lying': 1, 'skulpt': 1, 'fir': 1, 'growers': 1, 'chon': 1, 'lineup': 1, 'exposes': 1, 'shifting': 2, 'toon': 1, 'innocence': 1, 'profile': 2, 'dont': 1, 'pixie': 1, 'tinkerbell': 1, 'cards': 1, 'sought': 1, 'global': 2, 'outbreak': 1, 'blend': 2, 'obedient': 1, 'defeats': 1, 'duels': 1, 'backwards': 1, 'stairs': 1, 'suicidal': 1, 'valuable': 1, 'unbalanced': 1, 'vivian': 1, 'unforeseen': 2, 'rebuilding': 1, 'reinventing': 1, 'enslaved': 1, 'captors': 2, 'formed': 1, 'solely': 1, 'gin': 2, 'swilling': 1, 'riverboat': 1, 'persuaded': 1, 'strait': 1, 'laced': 2, 'wisdom': 2, 'blooded': 1, 'liver': 1, 'fava': 1, 'beans': 1, 'nice': 2, 'chianti': 1, 'widower': 1, 'disguising': 1, 'confide': 1, 'traditional': 1, 'spiteful': 1, 'rigged': 1, 'brandon': 2, 'routh': 2, 'proposed': 1, 'handbook': 1, 'mystical': 2, 'shepherdess': 1, 'fakes': 1, 'oakland': 3, 'athletics': 1, 'stan': 2, 'mogul': 1, 'randolph': 1, 'hearst': 1, 'suiters': 1, 'arnett': 1, 'shifu': 1, 'marial': 1, 'matthau': 5, 'hiatus': 1, 'wrecked': 1, 'lex': 2, 'luthor': 2, 'hovering': 1, 'skateboard': 1, 'taye': 1, 'diggs': 1, 'woes': 1, 'roderick': 1, 'whiterspoon': 1, 'wit': 1, 'catcher': 3, 'illness': 1, 'angelica': 2, 'bening': 1, 'artists': 4, 'marv': 1, 'relase': 1, 'nursery': 1, 'appearing': 1, 'directd': 1, 'upton': 1, 'sinclair': 1, 'sound': 3, 'editing': 3, 'farragut': 2, 'propaganda': 2, 'brussels': 1, 'slavers': 1, 'banding': 1, 'halle': 3, 'berry': 3, 'progress': 1, 'error': 6, 'destiny': 2, 'appears': 3, 'impression': 3, 'woe': 1, 'feuding': 2, 'classes': 1, 'jacobi': 2, 'sox': 1, 'bribes': 1, 'supplement': 1, 'measly': 1, 'milk': 1, 'hoax': 1, 'usher': 1, 'stylish': 1, 'constance': 1, 'towers': 1, 'blondes': 1, 'branaugh': 1, 'planning': 3, 'christians': 1, 'partial': 1, 'quotes': 2, 'stacked': 1, 'shit': 2, 'slimy': 1, 'scumbag': 1, 'puke': 1, 'wheelchair': 3, 'kirk': 2, 'gladiators': 1, 'taxidermy': 1, 'integrated': 1, 'switch': 3, 'grenades': 1, 'helmets': 1, 'mckellen': 2, 'nordic': 1, 'shane': 2, 'lotion': 1, 'potion': 1, 'hermione': 1, 'emerged': 1, 'getaway': 2, 'bigelow': 2, 'targeted': 2, 'loathing': 2, 'bandages': 1, 'influenced': 1, 'cg': 1, 'cheif': 1, 'asa': 2, 'runner': 1, 'refused': 1, 'sabbath': 1, 'hewitt': 1, 'mates': 1, 'karate': 1, 'ozarks': 1, 'scenerio': 1, 'seizes': 1, 'opportunity': 3, 'pretending': 1, 'kato': 2, 'vandalize': 1, 'closely': 1, 'resemble': 1, 'eduardo': 2, 'dilution': 1, 'percentage': 1, 'article': 2, 'chicken': 2, 'cannibalism': 1, 'inanimate': 2, 'playmates': 1, 'arrested': 1, 'switzerland': 1, 'decisions': 1, 'finished': 1, 'swiss': 1, 'arrest': 1, 'remained': 1, 'productions': 1, 'cecil': 2, 'dramatized': 1, 'hebrew': 2, 'egyptian': 1, 'deliverer': 1, 'natchitoches': 1, 'blends': 1, 'debute': 1, 'jenniffer': 1, 'oilman': 2, 'wealth': 1, 'centuries': 2, 'invitation': 1, 'managers': 2, 'humor': 1, 'afghanistan': 1, 'sebastian': 2, 'junger': 1, 'photojournalist': 2, 'hetherington': 1, 'pettyfer': 4, 'gwen': 1, 'stacy': 2, 'unicorn': 1, 'horn': 1, 'subject': 4, 'testing': 1, 'radiation': 1, 'simulation': 2, 'galifinakis': 2, 'ventures': 1, 'aws': 1, 'attacking': 2, 'beachgoers': 1, 'demonically': 1, 'ironically': 1, 'influence': 2, 'draws': 1, 'ligers': 1, 'thick': 2, 'fernando': 1, 'meirelles': 1, 'hid': 1, 'ventura': 1, 'commandos': 2, 'imagining': 1, 'midwest': 2, 'modified': 1, 'aragorn': 2, 'draw': 1, 'gaze': 1, 'approach': 2, 'images': 2, 'seagal': 1, 'cook': 1, 'battleship': 2, 'aided': 2, 'seize': 1, 'whom': 2, 'planned': 2, 'weaponized': 1, 'repossessed': 1, 'giallo': 1, 'gallery': 1, 'moviegoers': 1, 'hugely': 1, 'keir': 2, 'dullea': 2, 'overly': 2, 'middleweight': 1, 'paramedic': 2, 'edge': 2, 'severed': 2, 'consisted': 1, 'consists': 1, 'reference': 3, 'slavery': 2, 'alcatraz': 1, 'razor': 1, 'knives': 2, 'fingers': 2, 'nfl': 1, 'quarterback': 2, 'mechanic': 2, 'riot': 1, 'stunt': 1, 'retro': 2, 'unleashes': 2, 'tyranny': 1, 'epidemic': 1, 'secluded': 1, 'prix': 1, 'lt': 1, 'shed': 1, 'relative': 1, 'journeys': 1, 'wishing': 2, 'stallion': 3, 'participate': 2, 'babysitting': 1, 'unaware': 1, 'ahead': 1, 'gladiator': 1, 'trainer': 2, 'cavalry': 3, 'wonderful': 3, 'promise': 1, 'respect': 1, 'heffley': 2, 'paratrooper': 1, 'peppard': 1, 'truman': 2, 'capote': 2, 'feeling': 2, 'kevon': 1, 'macaws': 1, 'cockatoo': 1, 'smuggle': 1, 'beaten': 1, 'stiff': 1, 'lures': 1, 'rebellion': 1, 'rave': 1, 'theodore': 1, 'unoccupied': 2, 'hopkons': 1, 'primitive': 1, 'perfection': 2, 'vengeful': 4, 'retribution': 1, 'combining': 1, 'soundtracked': 1, 'steeler': 1, 'wheel': 3, 'rehearsing': 1, 'andr': 2, 'accent': 2, 'prevented': 1, 'understood': 1, 'remedy': 1, 'slapped': 1, 'concentrate': 1, 'earlier': 1, 'afred': 1, 'moe': 2, 'insist': 1, 'possibility': 2, 'heather': 1, 'instructed': 1, 'sawyer': 1, 'yippee': 2, 'yay': 2, 'towering': 1, 'mentioned': 1, 'catchphrase': 2, 'whoever': 1, 'wont': 1, 'kinney': 1, 'illustrated': 1, 'sixth': 2, 'grade': 4, 'muhammad': 1, 'zaire': 1, 'affluent': 1, 'griswald': 1, 'nursing': 2, 'policemen': 1, 'denmark': 4, 'ant': 2, 'grasshoppers': 1, 'dealt': 1, 'injury': 1, 'marauder': 1, 'reassembles': 3, 'collar': 3, 'stubborn': 1, 'graders': 2, 'solider': 3, 'singles': 2, 'dumped': 1, 'sugar': 3, 'arguments': 1, 'pitch': 1, 'launches': 1, 'musicians': 2, 'insinuate': 1, 'whiting': 1, 'hussey': 1, 'allocation': 1, 'lifter': 1, 'initials': 1, 'apted': 1, 'edmund': 2, 'pevensie': 4, 'eustace': 1, 'caspian': 1, 'presidents': 1, 'haughty': 1, 'attitude': 3, 'adapting': 1, 'utilizes': 1, 'acquire': 2, 'athletic': 1, 'ira': 2, 'defector': 2, 'accuses': 1, 'galvanizing': 1, 'offensive': 1, 'laziness': 1, 'marvelous': 1, 'rips': 1, 'participation': 1, 'devoted': 1, 'druggie': 1, 'unsinkable': 1, 'slay': 1, 'organize': 2, 'militia': 1, 'flees': 2, 'exile': 2, 'invades': 1, 'lawless': 2, 'desperate': 3, 'gravely': 1, 'pea': 2, 'soup': 2, 'vomit': 2, 'worth': 1, 'tutelage': 1, 'handy': 1, 'surprising': 1, 'flow': 1, 'sensation': 1, 'humbling': 1, 'counterespionage': 1, 'acquainted': 1, 'knot': 2, 'unreliable': 1, 'photographs': 1, 'assignments': 1, 'faster': 1, 'speeding': 1, 'bullet': 1, 'needle': 2, 'arabic': 1, 'peninsula': 1, 'saudi': 1, 'bearded': 1, 'bam': 1, 'margera': 1, 'communists': 2, 'mammal': 1, 'hydrgens': 1, 'oxygen': 1, 'prepared': 1, 'rainforest': 1, 'documentarians': 1, 'detonating': 1, 'grief': 2, 'uncovering': 1, 'closure': 1, 'unresolved': 1, 'cases': 1, 'unreciprocated': 1, 'haunt': 1, 'songwriter': 1, 'donald': 3, 'spoon': 1, 'medicine': 2, 'flair': 1, 'copier': 1, 'emitted': 1, 'clouds': 1, 'signalling': 1, 'visitor': 1, 'arrives': 2, 'urgent': 2, 'barada': 1, 'nikto': 1, 'kent': 3, 'handle': 1, 'suited': 1, 'tai': 1, 'janice': 1, 'fische': 1, 'jeremias': 1, 'evangelical': 2, 'requires': 2, 'swim': 1, 'mermaids': 1, 'subway': 1, 'platform': 1, 'wherever': 1, 'worldly': 1, 'insect': 2, 'burrows': 1, 'ripe': 1, 'fruit': 1, 'heathcliffe': 1, 'earnshaw': 1, 'chemotherapy': 1, 'moderate': 1, 'knack': 1, 'repairing': 1, 'cash': 6, 'thrust': 1, 'extends': 1, 'superpowers': 3, 'developing': 1, 'zany': 2, 'converse': 1, 'succeed': 1, 'pen': 3, 'loyal': 1, 'pose': 1, 'careless': 1, 'palahniuk': 1, 'mugging': 1, 'enables': 1, 'trainspotters': 1, 'sprinting': 1, 'whistle': 1, 'robs': 2, 'syrup': 2, 'guns': 2, 'eager': 1, 'verona': 1, 'induces': 1, 'endearing': 1, 'assorted': 1, 'chateau': 1, 'endangered': 1, 'preceding': 1, 'courteney': 2, 'troop': 1, 'landscape': 1, 'barracus': 1, 'murdoch': 1, 'uplifting': 2, 'chocolatier': 1, 'fezzik': 1, 'niccol': 1, 'elise': 1, 'deliberately': 1, 'crosses': 2, 'swamp': 1, 'brittany': 3, 'barden': 1, 'heavily': 1, 'mistreated': 1, 'scraping': 1, 'chigurh': 1, 'homemade': 1, 'rifle': 1, 'cultist': 1, 'motor': 2, 'strictly': 1, 'stockton': 1, 'greenwich': 1, 'bookstore': 1, 'hephburn': 1, 'ava': 2, 'gardner': 2, 'seduces': 1, 'lancaster': 4, 'wyler': 1, 'regards': 1, 'chapeaus': 1, 'cussler': 1, 'cleary': 1, 'deposed': 1, 'convincing': 1, 'enchanting': 1, 'cranston': 1, 'diplomatic': 1, 'zane': 1, 'caledon': 1, 'cal': 1, 'hockley': 1, 'geldof': 1, 'isolation': 1, 'cogan': 1, 'protected': 1, 'insomniac': 1, 'mlb': 1, 'peak': 1, 'steep': 1, 'offerred': 1, 'expand': 1, 'developer': 1, 'subdivision': 1, 'protesters': 1, 'disturbed': 1, 'assailants': 1, 'surprises': 2, 'corpse': 1, 'sewn': 1, 'stapled': 1, 'bolted': 1, 'settlers': 1, 'pushing': 1, 'dreyer': 1, '1928': 1, 'endeavors': 1, 'legion': 2, 'emblem': 1, 'mat': 1, 'meltdown': 1, 'spectacle': 1, 'bryner': 2, 'rameses': 1, 'yeoh': 2, 'ziyi': 2, 'zhang': 2, 'fantastic': 1, 'lock': 2, 'touch': 1, 'sum': 1, 'seas': 3, 'viciously': 1, 'sink': 3, 'difficulty': 1, 'manipulates': 1, 'grande': 1, 'desk': 1, 'jockeys': 2, 'gillespie': 1, 'holland': 1, 'elfman': 2, 'zanuck': 1, 'severely': 1, 'sacred': 1, 'text': 2, 'shorter': 1, 'denzil': 1, 'whitta': 1, 'accomplished': 1, 'physicist': 1, 'avowed': 1, 'radium': 1, 'mixes': 3, 'cloony': 1, 'merly': 1, 'flamboyant': 3, 'guilt': 1, 'bread': 1, 'alleged': 1, 'maximum': 1, 'interviews': 2, 'brandt': 2, 'gielgud': 2, 'jolly': 1, 'punched': 1, 'occasionally': 1, '00': 2, 'imaginary': 1, 'distressed': 1, 'hollowness': 1, 'stole': 1, 'investing': 1, 'careful': 1, 'effie': 1, 'trinket': 1, 'attach': 1, 'kerry': 1, 'bradford': 1, 'estelle': 1, 'parsons': 1, 'barrow': 1, 'everybody': 1, 'freudian': 1, 'jugian': 1, 'latter': 1, 'lasseter': 2, 'dubbing': 1, 'dialog': 1, 'mouth': 1, 'movements': 1, 'ruled': 3, 'explosion': 1, 'epa': 1, 'shuts': 1, 'ois': 1, 'coping': 1, 'myriad': 1, 'crises': 1, 'friendships': 1, 'dex': 1, 'gabourey': 3, 'sidibe': 3, 'harlem': 1, 'enroll': 1, 'alternative': 1, 'drafted': 1, 'gawky': 1, 'nerds': 3, 'joined': 1, 'heartless': 1, 'tycoons': 1, 'longtime': 1, 'stoick': 1, 'vast': 1, 'aspires': 1, 'bannion': 2, 'politically': 4, 'syndicate': 3, 'kinnear': 1, 'finals': 1, 'pageant': 1, 'vw': 1, 'katy': 1, 'michell': 1, 'eliminates': 1, 'replicas': 1, 'henri': 1, 'georges': 2, 'clouzot': 1, 'mistress': 1, 'winkler': 1, 'majors': 1, 'researcher': 2, 'welded': 1, 'exploded': 1, 'promoting': 1, 'chicks': 2, 'aware': 1, 'explosions': 1, 'dobbs': 2, 'mine': 2, 'sierra': 1, 'madre': 1, 'mountains': 1, 'traven': 1, 'racism': 2, 'antartica': 1, 'medlock': 1, 'terrifyingly': 1, 'wed': 2, 'chihiro': 1, 'bathe': 1, 'bathhouse': 1, 'wrecking': 1, 'cranky': 1, 'populous': 1, 'georgian': 1, 'joads': 1, 'oklahomah': 1, 'disadvantaged': 1, 'plainview': 1, 'deserted': 1, 'sentient': 3, 'hamburger': 1, 'rihanna': 2, 'effron': 1, 'forgets': 2, 'kim': 4, 'krickitt': 1, 'thurnan': 1, 'packing': 1, 'handicap': 1, 'zoro': 1, 'trafficking': 1, 'morrocco': 1, 'fitzgerald': 2, 'linearly': 1, 'attended': 2, 'liste': 1, 'americas': 1, 'loonies': 1, 'skivvies': 1, 'interweaves': 2, 'pals': 2, 'milo': 1, 'til': 2, 'forays': 1, 'bree': 1, 'daniels': 2, 'shanghai': 1, 'killings': 1, 'weaving': 1, 'attacus': 1, 'handyman': 1, 'suspended': 1, 'comedian': 3, 'whoopie': 1, 'shouting': 2, 'beer': 1, 'hippos': 1, 'scharzeneiger': 1, 'horrid': 1, 'enthusiasts': 2, 'spotting': 1, 'thinkin': 1, 'bishop': 1, 'cathedral': 1, 'prays': 1, 'carson': 2, 'traumatic': 1, 'merciless': 1, 'nimr': 1, 'antal': 1, 'nien': 1, 'jen': 1, 'elaine': 1, 'jin': 1, 'issei': 1, 'ogata': 1, 'polar': 1, 'caps': 1, 'begun': 2, 'melt': 1, 'conflicting': 1, 'opponents': 1, 'bursting': 1, 'hugging': 1, 'kristi': 2, 'resides': 1, 'cratchit': 1, 'gonzo': 1, 'moss': 1, 'extravaganza': 1, 'rebooted': 2, 'darker': 1, 'grittier': 1, 'tyson': 1, 'decrepit': 1, 'nitroglycerine': 1, 'shipment': 1, 'equipment': 1, 'alaska': 3, 'greenpeace': 1, 'volunteer': 1, 'rapidly': 1, 'fascist': 1, 'stepdaughter': 1, 'eerie': 4, 'captivating': 1, 'ferrara': 1, 'finzi': 1, 'contini': 1, 'aristocratic': 1, 'urbane': 1, 'abducting': 1, 'transforming': 2, 'maidens': 1, 'nearby': 1, 'trapp': 2, 'unfortunately': 1, 'annual': 4, 'mortal': 3, 'likely': 1, 'expulsion': 1, 'ghostly': 2, 'terminator': 1, 'conceive': 1, 'impress': 1, 'guardian': 1, 'climactic': 1, 'awoken': 1, 'strugges': 1, 'morals': 1, 'seniors': 2, 'throw': 2, 'midnight': 1, 'vulnerable': 1, 'dinero': 2, 'defended': 1, 'pranksters': 1, 'shock': 1, 'barrel': 1, 'sandy': 2, 'claws': 1, 'heal': 1, 'disarm': 1, 'describe': 1, 'reckoning': 1, 'waxing': 1, 'cratchet': 1, 'mcallister': 1, 'bashful': 1, 'sleepy': 1, 'sneezy': 1, 'worry': 1, 'suitors': 1, 'capable': 1, 'unlocking': 1, 'hansen': 1, 'skin': 1, 'rip': 1, 'replace': 1, 'dawes': 1, 'wyatt': 2, 'earp': 2, 'holliday': 1, 'southwestern': 1, 'recieve': 1, 'wrapper': 1, 'demise': 1, 'covert': 2, 'expedition': 1, 'dicky': 1, 'eklund': 1, 'yourself': 1, 'federal': 3, 'rampant': 1, 'haller': 1, 'maccormack': 1, 'motto': 1, 'brightest': 1, 'blackest': 1, 'shall': 1, 'dent': 1, 'chopsticks': 1, 'accomplish': 1, 'blurb': 1, 'increase': 1, 'aviator': 2, 'burst': 1, 'victoria': 1, 'lightening': 1, 'ra': 1, 'ghul': 1, 'irene': 2, 'dunne': 2, 'divorcing': 1, 'undermine': 1, 'isao': 1, 'takahata': 1, 'akiyuki': 1, 'nosaka': 1, 'mookie': 1, 'compared': 1, 'cuddly': 1, 'tabloid': 1, 'liquor': 1, 'soaked': 1, 'barbossa': 1, 'elusive': 1, 'levant': 2, 'sal': 1, 'mineo': 1, 'intentions': 1, 'explain': 1, 'finance': 2, 'design': 1, 'deputy': 1, 'marlene': 1, 'dietrich': 1, 'saloon': 2, 'gal': 1, 'frenchie': 1, 'nomiated': 1, 'momoa': 3, 'determination': 1, 'sexually': 2, 'harrasses': 1, 'batemen': 1, 'disgruntled': 3, 'inherits': 1, 'gelfling': 1, 'goblins': 1, 'mcteigue': 1, 'neville': 1, 'idle': 1, 'fanciful': 1, 'ambassador': 2, 'divine': 2, 'boozer': 1, 'willed': 1, 'copies': 1, 'overtones': 1, 'coaxed': 1, 'walked': 3, 'enthralled': 1, 'leagues': 1, 'basinger': 1, 'lynn': 1, 'bracken': 1, 'interracial': 1, 'bare': 1, 'huntsman': 2, 'leterrier': 1, 'edmond': 1, 'foolish': 1, 'bend': 1, 'arrow': 1, 'fuzzy': 1, 'lollipop': 1, 'mario': 3, 'andreacchio': 1, 'puppy': 1, 'puzo': 2, 'gripping': 1, 'genuine': 1, 'bogus': 1, 'kellogg': 1, 'macfadyen': 1, 'stevenson': 2, 'swashbuckling': 1, 'marisa': 1, 'tomei': 1, 'seller': 1, 'sleazy': 1, 'foolproof': 2, 'metallurgy': 2, 'plutonium': 1, 'processing': 1, 'purposefully': 1, 'psychologically': 1, 'exposing': 1, 'blatant': 1, 'violations': 1, 'rebellious': 2, 'accordance': 1, 'aquinas': 1, 'playback': 1, 'fed': 2, 'mundane': 2, 'bobs': 1, 'celebrity': 2, 'haircut': 1, 'stems': 1, 'lawsuits': 1, 'plump': 1, 'whi': 1, 'ch': 1, 'aimlessly': 1, 'colonial': 1, 'extraterrestrials': 2, 'fading': 1, 'mysteriously': 1, 'viewers': 2, 'belt': 1, 'wlaberg': 1, 'vocal': 1, 'hemisphere': 1, 'profession': 2, 'brook': 2, 'granddaughter': 1, 'noah': 1, 'baumbach': 1, 'uninterested': 1, 'bolshevik': 1, 'instrumental': 1, 'titles': 2, 'russ': 1, 'tamblyn': 1, 'acrobat': 1, 'richards': 1, 'rarely': 1, 'falstaff': 1, 'roistering': 1, 'blending': 1, 'stroll': 1, 'pat': 4, 'morita': 2, 'mentors': 1, 'jaden': 1, 'disobeys': 1, 'giamatti': 4, 'haden': 2, 'disappointment': 1, 'anchorman': 1, 'chenery': 1, 'tweedy': 1, 'colleagues': 1, 'unbeaten': 1, 'heavens': 1, 'hedges': 2, 'ahmet': 1, 'zappa': 1, 'bargained': 1, 'manhunt': 1, 'lowly': 1, 'gardener': 1, 'utterances': 1, 'phoebe': 1, 'cates': 1, 'stoned': 1, 'spicoli': 1, 'therman': 1, 'hood': 1, 'hoods': 1, 'examine': 1, 'sudden': 1, 'hansel': 1, 'gretel': 1, 'operatives': 2, 'wage': 2, 'meteorite': 1, 'bump': 1, 'bourgeois': 1, 'misogynistic': 1, 'snobbish': 2, 'phonetics': 1, 'stud': 1, 'judith': 1, 'rossner': 1, 'portland': 1, 'dector': 1, 'obnoxious': 1, 'flaherty': 1, 'doren': 1, 'shelton': 1, 'bankrupt': 1, 'freedonia': 1, 'sylvania': 1, 'teasdale': 1, 'penniless': 1, 'bud': 1, 'cort': 1, 'napping': 1, 'treacherous': 1, 'pedal': 1, 'reddy': 1, 'breadwinner': 1, 'allison': 1, 'pearson': 1, 'critters': 1, '1825': 2, '1884': 1, 'sudan': 1, 'mason': 2, 'resigns': 1, 'regiment': 1, 'rebels': 2, 'cheer': 1, 'keach': 2, 'lockwood': 1, 'visitors': 1, 'posse': 1, 'attempted': 1, 'persona': 1, 'steed': 1, 'ronald': 1, 'reagan': 1, 'nun': 2, 'prejean': 1, 'bellas': 1, 'campus': 2, 'deborah': 3, 'kerr': 3, 'rushing': 1, 'zinnemann': 3, 'focus': 3, 'amir': 1, 'lev': 1, 'spark': 1, 'rush': 1, 'berle': 1, 'deteriorate': 1, 'clinton': 1, '44': 2, 'magnums': 1, 'fiery': 1, 'archery': 1, 'slap': 1, 'achieve': 1, 'humiliating': 1, 'stuffy': 1, 'dimensions': 2, 'marty': 2, 'mcfly': 1, 'wrecks': 1, 'evolve': 1, 'missile': 1, 'sic': 1, 'shaken': 1, 'contribution': 1, 'illinois': 1, 'governor': 1, 'blagojevich': 1, 'referee': 1, 'mathilda': 1, 'treebeard': 1, 'anthropomorphic': 2, 'stark': 2, 'avant': 1, 'garde': 1, 'conflicted': 1, 'loyalties': 1, 'aint': 1, 'administers': 1, 'disturbing': 1, 'lethal': 1, 'tests': 1, 'unwilling': 1, 'subjects': 1, 'klondike': 1, '1925': 1, 'defied': 1, 'adversary': 1, 'amputated': 1, 'laboratory': 1, 'mgm': 1, 'garbo': 1, 'dockworker': 1, 'orchestrated': 1, 'condensed': 1, '360': 1, 'igor': 1, 'disastrous': 1, 'vengeance': 2, 'empty': 1, 'bluth': 1, 'robery': 1, 'championed': 2, 'clsssic': 1, 'popularize': 1, 'maritime': 1, 'berg': 2, 'sank': 1, '1912': 1, 'backstabbed': 1, 'reversal': 1, 'transylvania': 2, 'attain': 1, 'sprague': 1, 'grayden': 1, 'molly': 1, 'ephraim': 1, 'sends': 2, 'cameos': 1, 'rajon': 1, 'rondon': 1, 'dwight': 1, 'dusty': 1, 'butterfield': 1, 'cinematographer': 1, 'weirdness': 1, 'unravel': 1, 'polynesians': 1, 'settled': 1, 'macfarlane': 4, 'mustached': 1, 'hooper': 1, 'renounce': 1, '942': 1, 'aaaron': 1, 'moretz': 1, 'shooter': 1, 'fighters': 1, 'narratives': 1, 'mayhem': 1, 'bile': 1, 'duct': 1, 'streak': 1, 'continuing': 1, 'sweep': 2, 'puttin': 1, 'ritz': 1, 'unscathed': 1, 'delarge': 1, 'droogs': 1, 'workings': 1, 'schrek': 1, 'jackosn': 1, 'competitive': 1, 'capella': 1, 'inflates': 1, 'absurd': 1, 'contest': 1, 'reservation': 1, 'timid': 2, 'villainous': 1, 'likeable': 1, 'gru': 1, 'sylvain': 1, 'pricks': 1, 'finger': 2, 'slumber': 1, 'sticking': 1, 'armored': 1, 'letting': 1, 'founds': 1, 'dissolves': 1, 'closest': 1, 'slain': 1, 'feminine': 1, 'delema': 1, 'honcho': 1, 'despises': 1, 'mashing': 1, 'westerners': 1, 'unworldly': 1, 'heady': 1, 'rian': 1, 'represent': 2, 'exact': 1, 'deplorable': 1, 'nightmares': 1, 'signature': 1, 'bladed': 1, 'carbonite': 1, 'humphry': 2, 'extensive': 1, 'surgery': 1, 'adele': 1, 'shattered': 1, 'telepathy': 1, 'assembles': 1, 'catwoman': 1, 'besieged': 1, 'crook': 1, 'reaves': 1, 'aziz': 2, 'ansari': 2, 'orcs': 1, 'commonly': 1, 'scientifically': 1, 'inaccurate': 1, 'recreations': 1, 'extinct': 1, 'clare': 1, 'alot': 1, 'bordering': 1, 'gravity': 1, 'hereo': 1, 'katara': 1, 'waterbender': 1, 'disobeying': 1, 'overprotective': 1, 'starters': 1, 'slum': 1, 'development': 1, 'districts': 1, 'fables': 1, 'dread': 1, 'rockumentary': 1, 'improve': 1, 'enslavement': 1, 'cobwebs': 1, 'ideals': 1, 'obelisk': 1, 'accelerates': 1, 'evolution': 1, 'selick': 1, 'spread': 1, 'cirus': 1, 'romp': 1, 'outing': 1, 'below': 1, 'review': 1, 'rotten': 1, 'tomatoes': 1, 'arrived': 1, 'chenoweth': 1, 'cloris': 1, 'leachman': 1, 'meagan': 1, 'holder': 1, 'gamma': 1, 'phi': 1, 'beta': 1, 'sorority': 1, 'contacted': 1, 'playthings': 1, 'flanagan': 1, 'harasses': 1, 'musante': 1, 'acquaintances': 1, 'level': 1, 'minus': 1, 'baggage': 1, 'sully': 1, 'encountered': 1, 'patterson': 1, 'uptight': 1, 'relax': 1, 'werner': 1, 'herzog': 1, 'treadwell': 1, 'amie': 1, 'huguenard': 1, 'activists': 1, 'hallucinations': 1, 'pyramid': 1, 'bolton': 1, 'driving': 4, 'hour': 1, 'crowd': 2, 'stunning': 2, 'newer': 1, 'creepy': 2, 'dolls': 1, 'button': 1, 'overrated': 1, 'interplanetary': 1, 'jedis': 1, 'gostling': 1, 'policia': 1, 'warp': 1, 'backstabbing': 1, 'material': 1, 'sled': 1, 'zimba': 1, 'believing': 1, 'moor': 1, 'venice': 1, 'doubts': 1, 'virtue': 1, 'mod': 1, 'confuse': 1, 'goodfellas': 1, 'roadside': 1, 'heck': 1, 'moralistic': 1, 'cryptography': 1, 'bankers': 1, 'raiding': 1, 'coulda': 1, 'hunts': 1, 'plants': 1, 'laser': 1, 'countless': 1, 'deviant': 1, 'investigators': 1, 'curly': 2, 'matrix': 1, 'mixing': 1, 'anarchist': 1, 'trippy': 1, 'distractions': 1, 'prolonged': 1, 'attracts': 2, 'goslin': 1, 'sorts': 1, 'nasty': 2, 'ernst': 1, 'lubitsch': 1, 'jeanette': 1, 'macdonald': 1, 'horizon': 1, 'vigo': 1, 'chico': 1, 'harpo': 1, 'carlisle': 1, 'ruggles': 1, 'barry': 2, 'ace': 1, 'astor': 1, 'mankiewicz': 3, 'prospecting': 1, 'scoring': 1, 'total': 1, 'despatch': 1, 'schaefer': 1, 'pierson': 1, 'confronting': 2, 'emptiness': 1, 'coldness': 1, 'vidor': 1, 'tumultuous': 1, 'relations': 1, 'auclair': 1, 'scottie': 2, 'novak': 1, 'bel': 1, 'geddes': 1, 'franju': 1, 'controvery': 1, 'rossen': 1, 'piper': 1, 'laurie': 2, 'provine': 1, 'pollard': 1, 'vivacious': 1, 'harmonica': 1, 'hopper': 1, 'topol': 1, 'norma': 1, 'frey': 1, 'martino': 1, 'roscoe': 1, 'browns': 1, 'dern': 2, 'colleen': 1, 'dewhurst': 1, 'schneider': 1, 'jacqueline': 1, 'bisset': 1, 'leaud': 1, 'valentina': 1, 'cortese': 1, 'frankenstein': 2, 'fim': 1, 'gardenia': 1, 'madeline': 2, 'kringle': 1, 'heater': 1, 'marin': 1, 'rene': 1, 'daalder': 1, 'resorts': 1, 'remick': 1, 'dignity': 1, 'fisher': 2, 'somewhere': 1, 'adventured': 1, 'mashed': 1, 'potatoes': 1, 'margot': 1, 'kidder': 1, 'raises': 1, 'mining': 1, 'fransisco': 1, 'whil': 1, 'togeather': 1, 'adoptive': 1, 'upbringing': 1, 'eikenberry': 1, 'helmond': 1, 'siamese': 1, 'terminate': 2, 'olmos': 1, 'contracted': 2, 'colorado': 1, 'alliance': 1, 'militarily': 1, 'wolverines': 1, 'oust': 1, 'occupiers': 1, 'barash': 1, 'levinson': 1, 'nowhere': 1, 'mentor': 3, 'cambodia': 1, 'annie': 2, 'potts': 1, 'unicorns': 1, 'becker': 1, 'depalma': 2, 'hunger': 2, 'fingered': 1, 'henchman': 1, 'inigo': 1, 'montoya': 1, 'bootcamp': 1, 'toontown': 1, 'debutante': 1, 'cynical': 1, 'enduring': 1, 'superstars': 1, 'lucchese': 1, 'conway': 1, 'jeannot': 1, 'szwarc': 1, 'hierarchy': 1, 'hairstylist': 1, 'hedge': 1, 'trimmer': 1, 'ninny': 1, 'threadgoode': 1, 'contestants': 1, 'paquin': 2, 'campion': 2, 'lusted': 1, 'neill': 2, 'wang': 1, 'histories': 1, 'keys': 1, 'jan': 1, 'bont': 1, 'morrow': 1, 'bradd': 1, 'manga': 1, 'masamune': 1, 'shirow': 1, 'zuker': 2, 'ormond': 1, 'deception': 1, 'showcased': 1, 'ricks': 1, 'nephew': 1, 'exclaiming': 1, 'technique': 1, 'sizemore': 1, 'loaf': 1, 'ambulance': 1, 'sanity': 1, 'taymor': 1, 'diedrich': 1, 'bader': 1, 'sear': 1, 'communicated': 1, 'shuffling': 1, 'cate': 2, 'blanchett': 2, 'mangold': 2, 'dominique': 1, 'bauby': 1, 'syndrome': 1, 'jonas': 1, 'shimit': 1, 'amin': 1, 'brunhilde': 1, '2019': 1, 'strode': 1, 'trailer': 1, 'grindhouse': 1, 'coulter': 1, 'tennant': 1, 'gio': 1, 'perez': 1, 'amber': 1, 'valletta': 1, 'crowley': 1, 'sequal': 1, 'carlsbad': 1, 'agreeing': 1, 'szostak': 1, 'vogel': 1, 'turteltaub': 1, 'molina': 1, 'teresa': 2, 'palmer': 2, 'jorma': 1, 'taccone': 1, 'phillippe': 2, 'balfor': 2, 'faison': 1, 'johnston': 1, 'jovovich': 2, 'weir': 2, 'siberian': 1, 'totally': 1, 'eighth': 1, 'luketic': 1, 'selleck': 1, 'wilkinson': 1, 'schwentke': 1, 'caregivers': 1, 'mutual': 1, 'murdock': 1, 'ferrera': 1, 'carlos': 3, 'mencia': 1, 'lance': 1, 'regina': 1, 'dalton': 1, 'berkoff': 1, 'designer': 2, 'coined': 2, 'murderess': 1, 'faulty': 2, 'rosario': 1, 'dawson': 1, 'marriages': 2, 'holy': 1, 'behalf': 1, 'totalitarion': 1, 'organization': 2, 'crashed': 1, 'rhode': 1, 'fogler': 2, 'isla': 1, 'abigail': 1, 'breslin': 1, 'saldanha': 2, 'convention': 1, 'roulet': 1, 'dugan': 2, 'perlman': 1, 'foy': 1, 'nicholls': 1, 'mckendry': 1, 'purcell': 1, 'seminary': 1, 'olsen': 1, 'levine': 1, 'greenfield': 1, 'egglesfield': 1, 'krasinski': 1, 'mylod': 1, 'scorses': 1, 'dowse': 1, 'structure': 1, 'morra': 1, 'sarsgaard': 1, 'mullan': 1, 'striped': 1, 'exiled': 1, 'tail': 1, 'valerie': 1, 'racers': 1, 'gluck': 1, 'jenna': 1, 'clarkson': 1, 'fuller': 1, 'dominik': 1, 'jenkins': 1, 'gandolfini': 1, 'biology': 1, 'throwing': 1, 'ayoade': 1, 'rosemarie': 1, 'dewitt': 1, 'evanovich': 1, 'evade': 1, 'brent': 2, 'lorenzo': 1, 'bonaventura': 1, 'sucsy': 1, 'garner': 2, 'tours': 1, 'sienna': 1, 'guillory': 1, 'skarsgard': 1, 'lauter': 1, 'targets': 2, 'blunderbuss': 1, 'midwestern': 1, 'famke': 1, 'janssen': 1, 'julianna': 1, 'guilll': 1, 'felton': 1, 'thandie': 1, 'tarsem': 1, 'singh': 1, 'cavill': 1, 'freida': 1, 'pinto': 1, 'phaedra': 1, 'callahand': 1, 'magnum': 1, 'revolver': 1, 'cinderella': 1, 'pathetic': 1, 'fresh': 2, 'adventrues': 1, 'penguin': 3, 'mumble': 1, 'homeland': 1, 'tenaciously': 1, 'napped': 1, 'hyenas': 1, 'stroker': 1, 'unimaginable': 1, 'nieson': 1, 'stopping': 1, 'snowy': 1, 'centric': 1, 'legions': 1, 'workout': 1, 'narcotic': 1, 'addicted': 2, 'dickey': 1, 'meerkat': 1, 'barrier': 1, 'reef': 1, 'seinfeld': 3, 'fiber': 1, 'druggies': 1, 'sin': 1, 'aide': 1, 'hakkuna': 1, 'mattata': 1, 'emotionless': 1, 'fails': 1, 'casket': 1, 'biting': 1, 'projectile': 1, 'vomiting': 1, 'rituals': 1, 'awareness': 1, 'fracking': 1, 'unravels': 1, 'folks': 1, 'marquand': 1, 'connect': 1, 'less': 1, 'draped': 1, 'coward': 1, 'interacting': 1, 'separate': 3, 'dedicated': 1, 'coachroach': 1, 'slime': 1, 'interactions': 1, 'cabra': 1, 'chaffrey': 1, 'patton': 2, 'catapulted': 1, 'corddry': 1, 'giffin': 1, 'mccoy': 1, 'waugh': 1, 'alexis': 1, 'knapp': 1, 'fatt': 1, 'vergara': 1, 'attend': 1, 'galaxies': 1, 'moriarty': 1, 'israelites': 1, 'receiving': 1, 'tablets': 1, 'haples': 1, 'womanizer': 1, 'severe': 1, 'platonic': 1, 'lapd': 1, 'juliane': 1, 'separates': 1, 'wing': 1, 'inception': 1, 'ratings': 1, 'hillside': 1, 'ving': 1, 'rhames': 1, 'endure': 1, 'hurdles': 1, 'platoon': 1, 'impressive': 1, 'firearm': 2, 'moeller': 1, 'boles': 2, 'lionel': 1, 'atwill': 1, 'riccardo': 1, 'baroni': 1, 'lassparri': 1, 'federic': 1, 'carven': 1, 'teapot': 1, 'webb': 1, 'tyrannical': 1, 'jade': 1, 'whiz': 1, 'chained': 1, 'bathroom': 2, 'jessical': 1, '22': 1, 'barbera': 1, 'daft': 1, 'reuben': 1, 'heals': 1, 'letterman': 1, 'tilda': 1, 'swinton': 1, 'berkeley': 1, 'breathed': 1, 'crusades': 1, 'saoirse': 1, 'ronan': 1, 'lesster': 1, 'defenders': 1, 'cactus': 1, 'flinn': 1, 'banded': 1, 'supervillain': 1, 'britian': 1, 'cocky': 1, 'birthmark': 1, 'standardized': 1, 'scores': 1, 'asner': 1, 'newhart': 1, 'woodstock': 1, 'helpful': 1, 'biz': 1, 'claymation': 1, 'encampment': 1, 'alda': 1, 'comidy': 1, 'theives': 1, 'copycats': 1, 'illusions': 1, 'awakened': 1, 'lifelike': 1, 'profitable': 1, 'health': 1, 'dungeon': 1, '54': 1, 'resistance': 1, 'bonnet': 1, 'miscreant': 1, 'incapable': 1, 'bilk': 1, 'investors': 1, 'prestige': 1, 'exorcised': 1, 'rouge': 1, 'barbarian': 1, 'mcgowan': 1, 'thurmeier': 1, 'allcott': 1, 'carnahan': 1, 'eyed': 2, 'documented': 1, 'dunns': 1, 'weston': 1, 'shakespheare': 1, 'brannaugh': 1, 'thorton': 1, 'specialist': 1, 'conversations': 1, 'intertwining': 1, 'busey': 1, 'knew': 1, 'tawanda': 1, 'vehicular': 1, 'paralyzed': 1, 'docks': 1, 'peeta': 1, 'mellark': 1, 'messed': 2, 'carrol': 1, 'lamborgini': 1, 'hijacked': 1, 'seaton': 1, 'keri': 1, 'tractor': 1, 'unpopular': 1, 'reputations': 1, 'jazmine': 1, 'lease': 1, 'letters': 1, 'fascism': 1, 'extreme': 1, 'reactor': 1, 'leveled': 1, 'lilliputuans': 1, '1050': 1, 'kinds': 1, 'mya': 1, 'lil': 1, 'collaborated': 1, 'weakness': 1, 'kryptonite': 1, 'hideously': 1, 'correct': 1, 'administrative': 1, 'carradine': 1, 'envelope': 1, 'concieve': 1, 'ash': 1, 'kooky': 1, 'schwarzeneggar': 1, 'veterinary': 1, 'studies': 1, 'blended': 1, 'galactic': 1, 'armbands': 1, 'typewriter': 1, 'perennial': 1, 'visitations': 1, 'lear': 1, 'rent': 2, 'knock': 1, 'poppins': 1, 'whipping': 1, 'sorcery': 1, 'marcus': 1, 'nispel': 1, 'dementors': 1, 'silverbacks': 1, 'helpers': 1, 'teri': 1, 'wet': 1, 'horrendous': 1, 'acient': 1, 'mention': 1, 'hatchet': 1, 'defraud': 1, 'bending': 1, 'segorney': 1, '39': 1, 'gilbert': 1, 'consciousness': 1, 'dicks': 1, 'lauded': 1, 'yann': 1, 'martel': 1, 'miramax': 1, 'janero': 1, 'amplifier': 1, 'sparkly': 1, 'reforms': 1, 'discuss': 1, 'outsiders': 1, 'absurdities': 1, 'contradictions': 1, 'hackford': 1, 'cadet': 1, 'originate': 1, 'miserly': 1, 'thrift': 1, 'sloppy': 1, 'formulaic': 1, 'flips': 1, 'ax': 1, 'coffee': 1, 'regardless': 1, 'resulting': 1, 'voilent': 1, 'wraith': 1, 'laguna': 1, 'grisly': 1, 'baja': 1, 'tapes': 1, 'olds': 1, 'filling': 1, 'lovably': 1, 'disinherited': 1, 'favreu': 1, 'nineteenth': 1, 'rumplestiltskin': 1, 'dug': 1, 'designed': 1, 'apeocolypse': 1, 'gaming': 1, '2074': 1, 'masters': 1, 'gothem': 1, 'fending': 1, 'blizzards': 1, 'acheron': 1, 'expensive': 1, 'queequeg': 1, 'ishmael': 1, 'shaolin': 1, 'axe': 1, 'chop': 1, 'wifes': 1, 'simulated': 1, 'thaqt': 1, 'blinded': 1, 'courted': 1, 'supper': 1, 'menaced': 1, 'munched': 1, 'july': 1, 'compiled': 1, 'clips': 1, 'ineffective': 1, 'landmark': 1, 'stroke': 1, 'firearms': 1, 'initial': 1, 'menacing': 1, 'casino': 1, 'matters': 1, 'calling': 1, 'scorpio': 1, 'menaces': 1, 'nails': 1, 'callahan': 1, 'ferret': 1, 'unexpected': 1, 'harrowing': 1, 'fic': 1, 'strands': 1, 'epiphany': 1, 'expressing': 1, 'athlete': 1, 'sabermetrics': 1, 'dehaan': 1, 'iconically': 1, 'gainesville': 1, 'ripper': 1, 'thrice': 1, 'scratched': 1, 'marsellus': 1, 'mccallister': 1, 'labeled': 1, 'tolstoy': 1, 'foray': 1, 'crafted': 1, 'sicily': 1, 'confined': 1, 'chair': 1, 'window': 1, 'martindale': 1, 'uttering': 1, 'praise': 1, 'narnia': 1, 'infiltrate': 1, 'personal': 1, 'idiodic': 1, 'obscene': 1, 'notoriety': 1, 'glows': 1, 'heche': 1, 'lobbyist': 1, 'defection': 1, 'ascending': 1, 'gotten': 1, 'listed': 1, 'allens': 1, 'composers': 1, 'bombs': 1, 'varying': 1, 'landed': 1, 'whos': 1, 'pacifists': 1, 'easily': 1, 'vegetarians': 1, 'smitth': 1, 'nsa': 1, 'goons': 1, 'key': 1, 'motivated': 1, 'mckay': 1, 'dieter': 1, 'cunth': 1, 'heartedly': 1, 'devious': 1, 'northwest': 1, 'koreans': 1, 'prefects': 1, 'bath': 1, 'mull': 1, 'regularly': 1, 'scheduled': 1, 'traveler': 1, 'recreated': 1}\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "d7pIJmZCgRMA" + }, + "source": [ + "### 1.2.4 Cumulative token frequency" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "1mi4Vw5ABFXg", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 529 + }, + "outputId": "0e421a91-75a4-4b2d-b5cb-8fb1bd2bac98" + }, + "source": [ + "# Plot the cumulative distribution of token frequency\n", + "def cumulative_token_frequency(series, limit=20):\n", + " '''\n", + " Input:\n", + " series - pd.Series of words\n", + " Output:\n", + " [plot] - cumulative distribution of token frequency\n", + " '''\n", + " corpus=[word for word in series]\n", + " counter=Counter(corpus)\n", + " tokens_count = dict(counter).items()\n", + "\n", + " prop_list = []\n", + " print(\"Vocabulary Size: \", len(tokens_count))\n", + " for i in range(limit):\n", + " tokens_filtered = len(list(filter(lambda x: x[1]<=i, tokens_count)))\n", + " prop_list.append(round(tokens_filtered*100/len(tokens_count),2))\n", + " a4_dims = (11.7, 8.27)\n", + " fig, ax = plt.subplots(figsize=a4_dims)\n", + " plt.plot(prop_list)\n", + " plt.grid()\n", + " plt.xlabel(\"Counts\")\n", + " plt.ylabel(\"Proportion of Vocabulary (%)\")\n", + " # print(\"Proportion of unique words less than\",limit,\": \", round(tokens_filtered*100/len(tokens_dict),2),\"%\")\n", + "\n", + "cumulative_token_frequency(df[\"Word\"])" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Vocabulary Size: 10987\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "KE4VV7X4fUwj" + }, + "source": [ + "### 1.2.5 Entity Frequency" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "dB_9MvjOAoqq", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 553 + }, + "outputId": "4f8faa61-0504-4701-de7f-d825b78dec00" + }, + "source": [ + "tag_counter = plot_top_non_stopwords_barchart(df[\"Tag\"], top=25, word=False)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "There are 25 distinct tags in dataset\n", + "{'B-Actor': 5010, 'I-Actor': 6121, 'O': 55895, 'B-Plot': 6468, 'I-Plot': 62107, 'B-Opinion': 810, 'I-Opinion': 539, 'B-Award': 309, 'I-Award': 719, 'B-Year': 2702, 'B-Genre': 3384, 'B-Origin': 779, 'I-Origin': 3340, 'B-Director': 1787, 'I-Director': 1653, 'I-Genre': 2283, 'I-Year': 195, 'B-Soundtrack': 50, 'I-Soundtrack': 158, 'B-Relationship': 580, 'I-Relationship': 1206, 'B-Character_Name': 1025, 'I-Character_Name': 760, 'B-Quote': 126, 'I-Quote': 817}\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "YPCRkkm768sx" + }, + "source": [ + "\n", + "## Conclusion after analysis" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "BhDBB8847Dev" + }, + "source": [ + "### Vocab\n", + "1. The vocab size is 10987 (excluding stopwords). This is quite a large dataset.\n", + "2. Our dataset it comprised of long sentences, with average length of 20 and median length of 19, and the length ranges from 1 to 71. Only a small amount of sentences have the length greater than 40 => set the max length equal to 40 => need a lot of padding tokens.\n", + "3. 45% of the vocabulary only occur once but they could be person's names so let's keep them.\n", + "4. Year: can be replaced by a common\n", + "5. Number in text: can be replaced by a common\n", + "6. Lemmatization: *films* to *film*\n", + "7. All words are in lowercase.\n", + "8. No punctuations.\n", + "9. No informal text.\n", + "10. Lots of abbreviation (like *i'm, i'll, can't, 2morrow*)\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "fRaEnZIe8UWy" + }, + "source": [ + "### Tags\n", + "1. Most of tags are short, with average length of 2 words and median length of 4 words. There're some long tag and the longest tag has 40 words. This could be a plot description.\n", + "2. Most sentences are about plots.\n", + "3. There are 25 classes of entities, divided to 3 categories: B Tags (Beginning of an entity), I Tag (Intermediate Entity), Or None Tag (O). Proportions of B,I,O are about 14.50%, 50.31%, 35.19% respectively (section 2.7 Check the dataset imbalance)\n", + "4. Most of the tags are in the minority and O is the most common entity => need to over-sample the tags from the minority groups.\n", + "5. As under the section 2.7 Check the dataset imbalance, the percentage of sentences that only contain O tags -> 0.02% => It's small amount so we don't need to delete those sentences.\n", + "6. ALso as under the section 2.7 Check the dataset imbalance, the percentage of OOV tokens in test set -> 3.34%" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "Dtr2P3ZE3sE6" + }, + "source": [ + "\n", + "# Part 2: Pre-process the data\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "MRofcZfAUd9g" + }, + "source": [ + "\n", + "## 2.1 Stemming" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "1J6BsedjUeOf" + }, + "source": [ + "def stem_sentence(sentence):\n", + " sentence = sentence.split(' ')\n", + " stemmer = PorterStemmer()\n", + " result = [stemmer.stem(word) for word in sentence]\n", + " stemmed_sentence = ' '.join(result)\n", + " return stemmed_sentence" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "4p6nLSAePE9U" + }, + "source": [ + "\n", + "## 2.2 Lemmatization" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "reh2ym9rO2PM", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "8aeeca0a-9089-4146-db88-b7ba43d9f6b5" + }, + "source": [ + "nltk.download('punkt')\n", + "nltk.download('wordnet')\n", + "def lemmatize_sentence(sentence):\n", + " tokenization = nltk.word_tokenize(sentence)\n", + " lemmatizer = WordNetLemmatizer()\n", + " result = [lemmatizer.lemmatize(word) for word in tokenization]\n", + " lemmatized_sentence = ' '.join(result)\n", + " return lemmatized_sentence" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "[nltk_data] Downloading package punkt to /root/nltk_data...\n", + "[nltk_data] Unzipping tokenizers/punkt.zip.\n", + "[nltk_data] Downloading package wordnet to /root/nltk_data...\n", + "[nltk_data] Unzipping corpora/wordnet.zip.\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "Hm9UnKBqUeZA" + }, + "source": [ + "\n", + "## 2.3 Replacement" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "EZzhc1zBUeiv" + }, + "source": [ + "import re\n", + "\n", + "def replace(sentence, to_replace, replace_by):\n", + " replaced = sentence.replace(to_replace, replace_by)\n", + " return replaced\n", + "\n", + "def replace_num(sentence):\n", + " replaced = re.sub(r'^\\d{1,2}$', \"NUM\", sentence) # replace 1, 2 digits\n", + " replaced = re.sub(r'^\\d{4} s$', \"YEAR\", replaced) # replace year\n", + " replaced = re.sub(r'^\\d{4}$', \"YEAR\", replaced) # replace year\n", + " return replaced\n" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "vfswr-bAXwl5" + }, + "source": [ + "\n", + "## 2.4 Pre-processing pipeline" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "uvGcbsdOXw3N" + }, + "source": [ + "def apply_preproc(data_generator):\n", + " data_generator = list(map(lambda x: replace(x, \"ca n t\",\"cannot\"), data_generator))\n", + " data_generator = list(map(lambda x: replace(x, \"ll\",\"will\"), data_generator))\n", + " data_generator = list(map(lambda x: replace_num(x), data_generator))\n", + " data_generator = list(map(lambda x: lemmatize_sentence(x), data_generator))\n", + " return data_generator\n", + "\n", + "processed_sentences = apply_preproc(sentences)\n", + "processed_test_sentences = apply_preproc(test_sentences)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "WMK0gaxoOi97" + }, + "source": [ + "\n", + "## 2.5 Split to train/val datasets" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "yjPJLq1I0_CX" + }, + "source": [ + "split_ratio = 0.8\n", + "\n", + "def train_val_split(data, label, ratio, shuffle=True, random_seed=33):\n", + " length = len(data)\n", + " lines_index = [*range(length)] \n", + " # shuffle the indexes if shuffle is set to True\n", + " rnd.seed(random_seed)\n", + " if shuffle:\n", + " rnd.shuffle(lines_index)\n", + " split_point = int(length * ratio)\n", + "\n", + " train_data = []\n", + " train_label = []\n", + " val_data = []\n", + " val_label = []\n", + " for i in range(length):\n", + " if i <= split_point:\n", + " train_data.append(data[lines_index[i]])\n", + " train_label.append(label[lines_index[i]])\n", + " else:\n", + " val_data.append(data[lines_index[i]])\n", + " val_label.append(label[lines_index[i]])\n", + " return train_data, train_label, val_data, val_label\n", + "\n", + "\n", + "train_sentences, train_tags, val_sentences, val_tags = \\\n", + " train_val_split(processed_sentences, tags, split_ratio)\n" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "3Sot8kCJltNu" + }, + "source": [ + "\n", + "## 2.6 Tokenization and Padding" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "GDfsVDq9KxgV" + }, + "source": [ + "oov_tok = \"\"\n", + "trunc_type='post'\n", + "pad_type='post'\n", + "max_length = 71\n", + "\n", + "vocab_tokenizer = Tokenizer(oov_token=oov_tok)\n", + "vocab_tokenizer.fit_on_texts(train_sentences)\n", + "\n", + "vocab = vocab_tokenizer.word_index\n", + "reverse_vocab = dict([(value, key) for (key, value) in vocab.items()])\n", + "vocab_size = len(vocab)\n", + "\n", + "\n", + "train_sequences = vocab_tokenizer.texts_to_sequences(train_sentences)\n", + "val_sequences = vocab_tokenizer.texts_to_sequences(val_sentences)\n", + "test_sequences = vocab_tokenizer.texts_to_sequences(processed_test_sentences)\n", + "\n", + "train_padded_sequences = pad_sequences(train_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "val_padded_sequences = pad_sequences(val_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "test_padded_sequences = pad_sequences(test_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "CP_DmX44K24X", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "2d2cf840-66e6-495e-f3b6-ccd62dcbd3f9" + }, + "source": [ + "tag_tokenizer = Tokenizer(filters=\".\", lower=False, oov_token=oov_tok)\n", + "tag_tokenizer.fit_on_texts(train_tags)\n", + "\n", + "tag_map = tag_tokenizer.word_index\n", + "reverse_tag_map = dict([(value, key) for (key, value) in tag_map.items()])\n", + "tag_size = len(tag_map)\n", + "\n", + "train_tag_sequences = tag_tokenizer.texts_to_sequences(train_tags)\n", + "val_tag_sequences = tag_tokenizer.texts_to_sequences(val_tags)\n", + "test_tag_sequences = tag_tokenizer.texts_to_sequences(test_tags)\n", + "\n", + "\n", + "train_padded_tags = pad_sequences(train_tag_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "val_padded_tags = pad_sequences(val_tag_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "test_padded_tags = pad_sequences(test_tag_sequences,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "print(\"\\nExample of a a sentence and its tokenized, padded version\")\n", + "print(train_sentences[0])\n", + "print(train_padded_sequences[0])\n", + "print(\"\\nExample of a list of tags in a sentence and its tokenized, padded version\")\n", + "print(train_tags[0])\n", + "print(train_padded_tags[0])\n" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "\n", + "Example of a a sentence and its tokenized, padded version\n", + "what 2011 animated movie feature the voice of seth green joan cusack and dan fogler\n", + "[ 4 52 40 7 37 3 200 5 627 628 3033 1735 6 2199\n", + " 3760 0 0 0 0 0 0 0 0 0 0 0 0 0\n", + " 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n", + " 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n", + " 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n", + " 0]\n", + "\n", + "Example of a list of tags in a sentence and its tokenized, padded version\n", + "O B-Year B-Genre O O O O O B-Actor I-Actor I-Actor I-Actor O B-Actor I-Actor\n", + "[3 9 7 3 3 3 3 3 6 5 5 5 3 6 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n", + " 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "VmNBER6vh8fZ" + }, + "source": [ + "\n", + "## 2.7 Check the Imbalance in train/test dataset" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "1JkHts8fDElj", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "21199e96-faf4-4ff5-b376-683f6f91db13" + }, + "source": [ + "#Percentage of B, I and O Tags in train dataset\n", + "def get_tag_proportion(series_tags):\n", + " '''\n", + " Input:\n", + " series_tags - pd.Series of tags\n", + " Output:\n", + " [print] - B, I and O tags' proportion\n", + " '''\n", + " tags_list=[tag for tag in series_tags]\n", + " counter=dict(Counter(tags_list))\n", + " beg = 0\n", + " inter = 0\n", + " out = 0\n", + " for key, value in counter.items():\n", + " if key.startswith(\"B\"):\n", + " beg += value\n", + " elif key.startswith(\"I\"):\n", + " inter += value\n", + " else:\n", + " out += value\n", + " total = len(tags_list)\n", + " print(\"B tags proportion = {0:.2%}\".format(round(beg/total,4)))\n", + " print(\"I tags proportion = {0:.2%}\".format(round(inter/total,4)))\n", + " print(\"O tags proportion = {0:.2%}\".format(round(out/total,4)))\n", + "\n", + "get_tag_proportion(df[\"Tag\"])\n", + "\n", + "# Percentage of sentences that only contain O tags\n", + "# If this percentage > 50% => the dataset is imbalanced => drop empty sentences\n", + "def get_empty_tag_sentence_proportion(list_tag_sequence):\n", + " '''\n", + " Input:\n", + " list_tag_sequence - list of tag sequences in train/test set\n", + " Output:\n", + " [print] - Percentage of sentences that only contain O tags\n", + " '''\n", + " count = 0\n", + " for seq in list_tag_sequence:\n", + " if sum(seq) == 2 * len(seq): # if seq contains only 2 (token for O tag)\n", + " count += 1\n", + "\n", + " \n", + " print(\"\\nPercentage of sentences that only contain O tags -> {0:.2%}\".\\\n", + " format(round(count/len(list_tag_sequence),4)))\n", + " \n", + "get_empty_tag_sentence_proportion(train_tag_sequences)\n", + "\n", + "def get_OOV_density(list_token_sequence):\n", + " '''\n", + " Input:\n", + " list_token_sequence - list of token sequences in test set\n", + " Output:\n", + " [print] - Percentage of OOV token in the test set\n", + " '''\n", + " list_token_sequence = [token for seq in list_token_sequence for token in seq]\n", + " counter=dict(Counter(list_token_sequence))\n", + " print(\"\\nPercentage of OOV tokens in test set -> {0:.2%}\".\\\n", + " format(round(counter[1]/len(list_token_sequence),4)))\n", + "\n", + "get_OOV_density(test_sequences)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "B tags proportion = 14.50%\n", + "I tags proportion = 50.31%\n", + "O tags proportion = 35.19%\n", + "\n", + "Percentage of sentences that only contain O tags -> 0.02%\n", + "\n", + "Percentage of OOV tokens in test set -> 3.36%\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "0OZsbuw2iDAp" + }, + "source": [ + "\n", + "## 2.8 One-hot encoding" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "qdsnDiIZwGaC" + }, + "source": [ + "# Ont hot encoding\n", + "train_padded_tags = np.array([to_categorical(tags, num_classes = tag_size+1) \\\n", + " for tags in train_padded_tags])\n", + "val_padded_tags = np.array([to_categorical(tags, num_classes = tag_size+1) \\\n", + " for tags in val_padded_tags])\n", + "test_padded_tags = np.array([to_categorical(tags, num_classes = tag_size+1) \\\n", + " for tags in test_padded_tags])\n", + "\n" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "84RwGepu37jO" + }, + "source": [ + "\n", + "# Part 3: Building the model\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "yX9PMBUtuJmJ" + }, + "source": [ + "\n", + "## 3.1 Glove Embedding" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "3QAO1_GehxQ1", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "e91e9cf9-a8b1-42d6-b836-5ec6a2a9ae74" + }, + "source": [ + "!mkdir -p /glove_embedding\n", + "# Download data\n", + "!wget --no-check-certificate \\\n", + "http://nlp.stanford.edu/data/glove.6B.zip -O /glove_embedding/glove.6B.zip\n", + "!unzip /glove_embedding/glove.6B.zip -d /glove_embedding\n", + "\n" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "--2021-06-02 17:10:13-- http://nlp.stanford.edu/data/glove.6B.zip\n", + "Resolving nlp.stanford.edu (nlp.stanford.edu)... 171.64.67.140\n", + "Connecting to nlp.stanford.edu (nlp.stanford.edu)|171.64.67.140|:80... connected.\n", + "HTTP request sent, awaiting response... 302 Found\n", + "Location: https://nlp.stanford.edu/data/glove.6B.zip [following]\n", + "--2021-06-02 17:10:13-- https://nlp.stanford.edu/data/glove.6B.zip\n", + "Connecting to nlp.stanford.edu (nlp.stanford.edu)|171.64.67.140|:443... connected.\n", + "HTTP request sent, awaiting response... 301 Moved Permanently\n", + "Location: http://downloads.cs.stanford.edu/nlp/data/glove.6B.zip [following]\n", + "--2021-06-02 17:10:13-- http://downloads.cs.stanford.edu/nlp/data/glove.6B.zip\n", + "Resolving downloads.cs.stanford.edu (downloads.cs.stanford.edu)... 171.64.64.22\n", + "Connecting to downloads.cs.stanford.edu (downloads.cs.stanford.edu)|171.64.64.22|:80... connected.\n", + "HTTP request sent, awaiting response... 200 OK\n", + "Length: 862182613 (822M) [application/zip]\n", + "Saving to: ‘/glove_embedding/glove.6B.zip’\n", + "\n", + "/glove_embedding/gl 100%[===================>] 822.24M 5.08MB/s in 2m 40s \n", + "\n", + "2021-06-02 17:12:53 (5.15 MB/s) - ‘/glove_embedding/glove.6B.zip’ saved [862182613/862182613]\n", + "\n", + "Archive: /glove_embedding/glove.6B.zip\n", + " inflating: /glove_embedding/glove.6B.50d.txt \n", + " inflating: /glove_embedding/glove.6B.100d.txt \n", + " inflating: /glove_embedding/glove.6B.200d.txt \n", + " inflating: /glove_embedding/glove.6B.300d.txt \n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "u2SIlF-NeR_o", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "2b6137ac-2988-4387-dc70-41fb7b13f552" + }, + "source": [ + "GLOVE_DIR = \"/glove_embedding\"\n", + "embedding_dim = 300\n", + "hits = 0\n", + "misses = 0\n", + "embeddings_index = {}\n", + "\n", + "with open(os.path.join(GLOVE_DIR, 'glove.6B.300d.txt')) as f:\n", + " for line in f:\n", + " values = line.split()\n", + " word = values[0]\n", + " coefs = np.asarray(values[1:], dtype='float32')\n", + " embeddings_index[word] = coefs\n", + "\n", + "print('Found %s word vectors.' % len(embeddings_index))\n", + "\n", + "embedding_matrix = np.zeros((len(vocab) + 1, embedding_dim))\n", + "for word, i in vocab.items():\n", + " embedding_vector = embeddings_index.get(word)\n", + " if embedding_vector is not None:\n", + " # words not found in embedding index will be all-zeros.\n", + " embedding_matrix[i] = embedding_vector\n", + " hits += 1\n", + " else:\n", + " misses += 1\n", + "print(\"Converted %d words (%d misses)\" % (hits, misses))\n", + "\n", + " " + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Found 400000 word vectors.\n", + "Converted 8448 words (664 misses)\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "3h-sT21kjV2X" + }, + "source": [ + "\n", + "## 3.2 Define the model " + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "tshH4jK03oWM", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "f6018b1a-5900-48f7-e5c0-5d21647288a9" + }, + "source": [ + "!pip install tensorflow_addons\n", + "import tensorflow_addons as tfa\n", + "\n", + "\n", + "# Model architecture\n", + "batch_size = 32\n", + "embedding_dim = 300\n", + "max_length = 71\n", + "\n", + "def BiLSTM(vocab_size=vocab_size, tag_size=tag_size, hidden_size = 32, \n", + " embedding_dim=embedding_dim):\n", + " sequence_input = Input(shape = (max_length,))\n", + "\n", + " model = Embedding(input_dim = vocab_size+1, \n", + " output_dim = embedding_dim, \n", + " input_length = max_length, \n", + " embeddings_initializer=Constant(embedding_matrix),\n", + " trainable=False,\n", + " mask_zero = False)(sequence_input)\n", + " \n", + " model = Bidirectional(LSTM(units = hidden_size,return_sequences=True,\n", + " recurrent_dropout=0.1))(model)\n", + "\n", + " model = TimeDistributed(Dense(hidden_size, activation=\"relu\"))(model)\n", + " outputs = Dense(tag_size+1, activation='softmax')(model)\n", + " #crf = tfa.layers.CRF(tag_size+1) # CRF layer\n", + " #outputs = crf(model)\n", + "\n", + " model = Model(inputs=sequence_input, outputs=outputs)\n", + "\n", + " model.compile(optimizer=\"RMSprop\", \n", + " loss = tf.keras.losses.categorical_crossentropy, \n", + " metrics=['accuracy'])\n", + " #loss=crf.loss_function, metrics=[crf.accuracy])\n", + " return model\n", + "\n", + "\n", + "model = BiLSTM(vocab_size=vocab_size, tag_size=tag_size, hidden_size = 32, \\\n", + " embedding_dim=embedding_dim)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Requirement already satisfied: tensorflow_addons in /usr/local/lib/python3.7/dist-packages (0.13.0)\n", + "Requirement already satisfied: typeguard>=2.7 in /usr/local/lib/python3.7/dist-packages (from tensorflow_addons) (2.7.1)\n", + "WARNING:tensorflow:Layer lstm_3 will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n", + "WARNING:tensorflow:Layer lstm_3 will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n", + "WARNING:tensorflow:Layer lstm_3 will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "sXWo98J3jgLQ" + }, + "source": [ + "\n", + "## 3.3 Callbacks" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "d2hDC-w7DGvu" + }, + "source": [ + "# Callback\n", + "class myCallback(tf.keras.callbacks.Callback):\n", + " def on_epoch_end(self, epoch, logs={}):\n", + " if(logs.get('val_accuracy')>0.95):\n", + " print(\"\\nReached 95% accuracy so cancelling training!\")\n", + " self.model.stop_training = True\n", + "\n", + "checkpointer = ModelCheckpoint(filepath = 'NER_BiLSTM.h5',\n", + " verbose = 0,\n", + " mode = 'auto',\n", + " save_best_only = True,\n", + " monitor='val_loss')\n", + "\n", + "earlystopper = EarlyStopping(monitor='val_loss', min_delta=0, patience=3, \n", + " verbose=0, mode='auto', \n", + " baseline=None, restore_best_weights=True)\n", + "\n", + "initial_learning_rate = 0.001\n", + "epochs = 15\n", + "decay = initial_learning_rate / epochs\n", + "def lr_time_based_decay(epoch, lr):\n", + " return lr * 1 / (1 + decay * epoch)\n", + "\n", + "lr_scheduler = tf.keras.callbacks.LearningRateScheduler(lr_time_based_decay, verbose=1)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "gKnkkiSS4EwI" + }, + "source": [ + "\n", + "# Part 4: Train the Model \n" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "vQ7uIYmgX3SC", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "faa0e4c9-2a76-4ada-eab4-8f2f85460b38" + }, + "source": [ + "num_epochs = 15\n", + "history = model.fit(train_padded_sequences, train_padded_tags, \n", + " batch_size=batch_size, epochs=num_epochs, \n", + " validation_data= (val_padded_sequences, val_padded_tags),\n", + " callbacks=[checkpointer, earlystopper, lr_scheduler])\n", + "\n", + "model.summary()" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Epoch 1/15\n", + "\n", + "Epoch 00001: LearningRateScheduler reducing learning rate to 0.0010000000474974513.\n", + "196/196 [==============================] - 88s 428ms/step - loss: 0.4487 - accuracy: 0.9060 - val_loss: 0.2207 - val_accuracy: 0.9352\n", + "Epoch 2/15\n", + "\n", + "Epoch 00002: LearningRateScheduler reducing learning rate to 0.0009999333852717665.\n", + "196/196 [==============================] - 81s 415ms/step - loss: 0.1669 - accuracy: 0.9501 - val_loss: 0.1606 - val_accuracy: 0.9510\n", + "Epoch 3/15\n", + "\n", + "Epoch 00003: LearningRateScheduler reducing learning rate to 0.0009998000348467315.\n", + "196/196 [==============================] - 81s 414ms/step - loss: 0.1343 - accuracy: 0.9582 - val_loss: 0.1579 - val_accuracy: 0.9514\n", + "Epoch 4/15\n", + "\n", + "Epoch 00004: LearningRateScheduler reducing learning rate to 0.0009996001259493627.\n", + "196/196 [==============================] - 80s 410ms/step - loss: 0.1163 - accuracy: 0.9629 - val_loss: 0.1405 - val_accuracy: 0.9560\n", + "Epoch 5/15\n", + "\n", + "Epoch 00005: LearningRateScheduler reducing learning rate to 0.0009993336718878056.\n", + "196/196 [==============================] - 81s 413ms/step - loss: 0.1035 - accuracy: 0.9662 - val_loss: 0.1327 - val_accuracy: 0.9579\n", + "Epoch 6/15\n", + "\n", + "Epoch 00006: LearningRateScheduler reducing learning rate to 0.0009990006859666584.\n", + "196/196 [==============================] - 81s 414ms/step - loss: 0.0936 - accuracy: 0.9692 - val_loss: 0.1316 - val_accuracy: 0.9579\n", + "Epoch 7/15\n", + "\n", + "Epoch 00007: LearningRateScheduler reducing learning rate to 0.0009986012978557466.\n", + "196/196 [==============================] - 80s 410ms/step - loss: 0.0856 - accuracy: 0.9716 - val_loss: 0.1377 - val_accuracy: 0.9581\n", + "Epoch 8/15\n", + "\n", + "Epoch 00008: LearningRateScheduler reducing learning rate to 0.0009981355208293137.\n", + "196/196 [==============================] - 81s 414ms/step - loss: 0.0780 - accuracy: 0.9744 - val_loss: 0.1378 - val_accuracy: 0.9582\n", + "Epoch 9/15\n", + "\n", + "Epoch 00009: LearningRateScheduler reducing learning rate to 0.0009976034845113318.\n", + "196/196 [==============================] - 81s 415ms/step - loss: 0.0709 - accuracy: 0.9762 - val_loss: 0.1359 - val_accuracy: 0.9582\n", + "Model: \"model_3\"\n", + "_________________________________________________________________\n", + "Layer (type) Output Shape Param # \n", + "=================================================================\n", + "input_4 (InputLayer) [(None, 71)] 0 \n", + "_________________________________________________________________\n", + "embedding_3 (Embedding) (None, 71, 300) 2733900 \n", + "_________________________________________________________________\n", + "bidirectional_3 (Bidirection (None, 71, 64) 85248 \n", + "_________________________________________________________________\n", + "time_distributed_3 (TimeDist (None, 71, 32) 2080 \n", + "_________________________________________________________________\n", + "dense_7 (Dense) (None, 71, 27) 891 \n", + "=================================================================\n", + "Total params: 2,822,119\n", + "Trainable params: 2,822,119\n", + "Non-trainable params: 0\n", + "_________________________________________________________________\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "M3vHe1vaC-fG", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "3b226742-9c0d-4451-e357-e8816e877ab9" + }, + "source": [ + "history.history.keys()" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy', 'lr'])" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 59 + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "K-LuVEjYZ1qg", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 517 + }, + "outputId": "74ec1ae3-0c31-486f-804c-4037a65ca371" + }, + "source": [ + "acc = history.history['accuracy']\n", + "val_acc = history.history['val_accuracy']\n", + "loss = history.history['loss']\n", + "val_loss = history.history['val_loss']\n", + "plt.figure(figsize = (8,8))\n", + "epochs = range(1, len(acc) + 1)\n", + "plt.plot(epochs, acc, 'wo', label='Training acc')\n", + "plt.plot(epochs, val_acc, 'w', label='Validation acc')\n", + "plt.title('Training and validation accuracy')\n", + "plt.legend()" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 60 + }, + { + "output_type": "display_data", + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "_orawE5havtv", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 499 + }, + "outputId": "0a67400a-eaea-4a26-823d-2a3b71e80816" + }, + "source": [ + "plt.figure(figsize = (8,8))\n", + "plt.plot(epochs, loss, 'wo', label='Training loss')\n", + "plt.plot(epochs, val_loss, 'w', label='Validation loss')\n", + "plt.title('Training and validation loss')\n", + "plt.legend()\n", + "plt.show()" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeoAAAHiCAYAAAAnPo9XAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3deXRV9dn28SsTEEIgQFQeMoOAQX1qwOCEoGJBRAgVrQQFRAoOUMGKYmMtoC2t2sehWtEVAXEgERAQ6oCooFQFDiSBYBJIwgkZQDESIMzh5Pf+QTkvx0DmsHeS72ete62z53vHrl7s2UuSEQAAsCVvqxsAAADnRlADAGBjBDUAADZGUAMAYGMENQAANkZQAwBgYwQ1cIaPP/5YY8aMqfd5reR0OjVgwIB6X68xRl27dpUkzZkzR3/605+qNW9NjRo1SqtWrarVspXp37+/CgoK6n29QEMwFNWYq7S01F0ul8scOXLEPTxq1CjL+7O6nE6nGTBgQL2v1xhjunbtWq/zRkREGGOM8fHxafC/S//+/U1BQYHl/30oqqryFdDIBQYGun87nU797ne/0xdffFFhPh8fH7lcrvPZGgDUGae+0WSdPrX5+OOPa8+ePZo/f76CgoK0cuVK7d27V/v27dPKlSsVEhLiXmbNmjUaP368JGns2LFat26dnn/+ee3bt087d+7ULbfcUqt5IyMj9dVXX+ngwYNavXq1Xn31Vb3zzjtn7bs6PT799NP6z3/+o4MHD2rVqlXq2LGje/o999yjvLw8FRcXKyEh4Zx/nz59+mjPnj3y9v7//zcwfPhwbdmyRZIUGxurb7/9ViUlJdq9e7deeeUV+fn5nXVd8+fP1zPPPOMenjZtmnbv3q2ioiKNGzfOY95bb71VKSkpOnDggPLz8zVjxgz3tK+//lqStH//fpWWlurqq692/21Pu+aaa7Rx40bt379fGzdu1DXXXFPtv01lLrnkEq1Zs0YlJSXatm2bhg4d6p42ePBgff/99zp48KAKCwv16KOPSpI6duyolStXqqSkRD///LO+/vpreXl5VWt7QHUR1GjSOnXqpA4dOigiIkITJ06Ut7e35s+fr4iICIWHh+vo0aN69dVXz7n8VVddpe3btys4OFjPPfec5s6dW6t5Fy5cqI0bN6pjx46aOXOmRo8efc71VKfHUaNGady4cbrwwgvVokULTZs2TZIUHR2tOXPmaPTo0ercubM6duyo0NDQs25n48aNOnz4sG666SaP9S5cuFCS5HK59Mgjjyg4OFjXXHONBgwYoIceeuicfZ82aNAgTZs2Tb/+9a/VrVs33XzzzR7TDx8+rDFjxigoKEhDhgzRgw8+qLi4OElSv379JJ36x0pgYKDWr1/vsWz79u310Ucf6Z///Kc6duyoF154QR999JE6dOhQ5d+mMr6+vlq5cqU+++wzXXjhhfr973+v9957T927d5ckzZ07V/fff7/atm2ryy67TF9++aUk6dFHH1VhYaEuuOACXXTRRUpISJAxpsrtATVl+fl3iqqvOvN6bP/+/c3x48dNy5Ytzzn/r371K7Nv3z738Jo1a8z48eONJDN27FiTnZ3tnubv72+MMeaiiy6q0bxhYWGmrKzM+Pv7u6e/88475p133qnWPp2txyeffNI9/OCDD5pPPvnESDJPPfWUSUpKck9r3bq1OX78+DmvUT/zzDNm7ty5RpJp06aNOXTokAkPDz/rvFOmTDFLly51D5953Xn+/PnmmWeeMZLM3Llzzd/+9jf3fN26dav0GvWLL75oXnjhBSOd/Rr12LFjzbp164wkc88995gNGzZ4LP/tt9+asWPHVvm3+WWdeY26b9++Zs+ePcbLy8s9feHChWbGjBlGktm1a5eZOHGiCQwM9FjHrFmzzPLly6t9rZ6ialMcUaNJ++mnn3T8+HH3sL+/v15//XXl5eXpwIED+vrrr9W+fXuP079n+uGHH9y/jx49Kklq06ZNjebt3Lmz9u3b5x4nqdK7javT45nbOnLkiLunzp07e6z7yJEj+vnnn8+5rYULF+r2229XixYtdPvttyslJUX5+fmSpG7dumnlypXas2ePDhw4oNmzZys4OPic6zrtlz3s2rXLY3qfPn305Zdfau/evdq/f78eeOCBaq339Lp/ub5du3Z5XBo419+mOj2feTR85npHjBihW2+9Vbt27dLatWt19dVXS5Kef/555eTk6LPPPlNubq6mT59erf0AaoKgRpP2y9OQjz76qHr06KGrrrpK7dq1c59qbcjrinv27FGHDh3k7+/vHhcWFnbO+evS4549ezzW7e/vX+k12szMTO3atUuDBw/2OO0tnXrkKisrS926dVO7du2UkJBQqx7Cw8M9pi9cuFArVqxQWFiYgoKC9Prrr7vXW9Vp4927dysiIsJjXHh4uIqKiqrsq6r1hoWFeezfmevdtGmThg8frgsvvFDLly/XokWLJEmHDh3StGnT1LVrVw0bNkx/+MMfPC4lAPWBoEazEhgYqKNHj2r//v1q3769x41MDSU/P1+bNm3SzJkz5efnp6uvvtrjRqX67HHJkiW67bbbdN1118nPz09PP/30Oc8WnLZw4UJNmTJF/fr10+LFiz36OHjwoA4dOqQePXrowQcfrFYPixYt0r333qvo6Gj5+/tX6D8wMFD79u3T8ePHFRsbq1GjRrmn/fTTT3K5XOrSpctZ1/3xxx+re/fuio+Pl4+Pj37729+qZ8+e+ve//12t3s5lw4YNOnLkiB5//HH5+vqqf//+Gjp0qJKTk+Xn56dRo0apbdu2OnnypA4ePKjy8nJJ0pAhQ9zPhx84cEAul8s9DagvBDWalZdeekn+/v4qLi7W+vXr9emnn56X7d5999265ppr9PPPP+svf/mL3n//fY9T8vXVY0ZGhiZNmqSFCxdqz549KikpUWFhYaXLJCUlqX///vryyy89TpNPmzZNo0aNUmlpqRITE/X+++9Xq4dPP/1UL730kr788kvl5OS4b7w67aGHHtLTTz+tgwcP6s9//rP76FQ6dcngr3/9q7755huVlJToqquu8lh23759uu222/Too4/q559/1uOPP67bbrut0tP71VFWVqahQ4dq8ODBKi4u1muvvaYxY8Zo+/btkqTRo0e7L0U88MADuvvuuyWdujzw+eef69ChQ/ruu+/02muvae3atXXqBfglL526WA3gPEpOTlZWVpZmzpxpdSsAbI4jauA8uPLKK9WlSxd5eXlp0KBBiouL0/Lly61uC0AjwJvJgPOgU6dOWrp0qTp27KjCwkI9+OCDSktLs7otAI0Ap74BALAxTn0DAGBjBDUAADZmu2vUe/furfDmIQAAmrKIiAhdeOGFZ51mu6DetWuXYmNjrW4DAIDzxuFwnHMap74BALAxghoAABsjqAEAsDHbXaMGANRM+/btNXXqVEVGRjbol+BQN8YY5eXl6aWXXlJJSUm1lyOoAaCRmzp1qjZt2qSnn35aLpfL6nZwDj4+PhoyZIimTp1ao6/iceobABq5yMhIffzxx4S0zblcLn300UeKjIys0XIENQA0cl5eXoR0I+FyuWp8eYKgBgDUSYcOHZSamqrU1FTt2bNHhYWF7mE/P79Kl+3du7defvnlKrfxzTff1Euv/fv318qVK+tlXecLQQ0AzUx8fLycTqdcLpecTqfi4+PrtL59+/YpJiZGMTExev311/Xiiy+6h8vKyuTj43POZTdv3qwpU6ZUuY3rrruuTj02ZgQ1ADQj8fHxSkxMVGRkpLy9vRUZGanExMQ6h/UvzZ8/X3PmzNH69ev13HPPKTY2Vt9++61SUlL0zTffqHv37pI8j3BnzJihuXPnas2aNcrNzdXvf/979/pKS0vd869Zs0aLFy9WZmam3n33Xfc8gwcPVmZmpjZt2qSXX365yiPn9u3ba9myZdqyZYu+++47XX755ZKkfv36uc8IpKSkqE2bNurUqZO++uorpaamKj09XX379q3Xv1dluOsbAJqR2bNnKyAgwGNcQECAZs+eraSkpHrdVmhoqK699lqVl5crMDBQ119/vVwulwYMGKDZs2frjjvuqLDMJZdcohtvvFGBgYHavn275syZo5MnT3rMExMTo0svvVS7d+/WN998o+uuu06bNm3SG2+8oX79+ikvL08LFy6ssr9Zs2YpNTVVv/nNb3TjjTfq7bffVkxMjKZNm6ZJkybp22+/VUBAgI4dO6aJEydq1apVmj17try9vdW6det6+ztVhaAGgGYkPDy8RuPrYvHixSovL5cktWvXTgsWLFC3bt1kjDnnteuPPvpIJ06c0M8//6y9e/fqoosuUlFRkcc8GzdudI9LS0tTZGSkDh06pJ07dyovL0+SlJSUpIkTJ1baX9++fTVixAhJ0po1a9SxY0cFBgbqm2++0QsvvKD33ntPS5cuVVFRkRwOh+bNmyc/Pz8tX75cW7ZsqcufpkY49Q0AzUh+fn6NxtfF4cOH3b+feeYZrVmzRpdffrmGDh2qVq1anXWZ48ePu3+7XC75+lY8nqzOPHXx7LPP6ne/+538/f31zTffqEePHlq3bp369eunoqIivfXWWxo9enS9brMy1QrqQYMGKSsrS9nZ2Zo+ffo557v99ttljFHv3r0lnfps15EjR9zn+ufMmVM/XQMAaiUhIcEjQKVTgZqQkNCg223Xrp37KPjee++t9/Vv375dXbp0UUREhCTprrvuqnKZdevW6e6775Z06tp3cXGxSktL1aVLF23btk3PPfecHA6HLrnkEoWHh+vHH3/Um2++qTfffFO9evWq9304lyr/GeLt7a1//etf+vWvf63CwkI5HA6tWLFCmZmZHvO1adNGU6ZM0fr16z3G5+bmKiYmpn67BgDUyunr0LNnz1Z4eLjy8/OVkJBQ79enf+m5557TggUL9Kc//UkfffRRva//2LFjeuihh/Tpp5/q8OHDlX428rSZM2dq3rx52rJli44cOaKxY8dKOvWmtxtvvFHl5eX6/vvv9cknn2jkyJF67LHHVFZWpkOHDmnMmDH1vg+VMZXV1VdfbT799FP38BNPPGGeeOKJCvO9+OKL5tZbbzVr1qwxvXv3NpJMRESESU9Pr3T9vyyHw1Gj+Sur+Ph443Q6jcvlMk6n08THx9fbuimKouxSb7/9tuU92KECAgLcv//1r3+ZqVOnWt5Tdf97VZZ9VZ76DgkJUUFBgXu4sLBQISEhHvPExMQoLCxMH3/8cYXlo6KilJKSorVr157zdvYJEybI4XDI4XAoODi4qpaq5Xw9ggAAsIcJEyYoNTVV33//vdq1a6c33njD6pbqTaXJP2LECJOYmOgevueee8wrr7ziHvby8jJr1qwxERERRpLHEXWLFi1Mhw4djCTTq1cvk5+fbwIDAyvdXn0dUTudTnM2TqfT8n9NURRF1WdxRN24qt6PqIuKihQWFuYeDg0N9bhVPjAwUJdddpnWrl0rp9Opq6++WitWrFDv3r114sQJ7du3T5KUkpKi3Nxc90PuDe18PoIAAEBDqTKoHQ6HunXrpsjISPn5+WnkyJFasWKFe/rBgwd1wQUXKCoqSlFRUVq/fr2GDRumzZs3Kzg4WN7epzYRFRWlbt26aefOnQ23N2c4n48gAADQUKoMapfLpcmTJ2vVqlXKzMzUokWLlJGRoVmzZmno0KGVLtuvXz9t3bpVqampWrJkiR544IEafSy7Lqx6BAEAgPpm+fn6M4u7vimKompWXKNuXFXv16gbs6SkJEVFRcnHx0dRUVEN/pwgADRHX375pQYOHOgxbsqUKXrttdfOucyaNWvcL8f66KOP1K5duwrzzJgxQ48++mil246Li1N0dLR7eNasWRowYEBN2j8rO30Os0kHNQCg4SUlJWnkyJEe40aOHFntg6MhQ4bowIEDtdr28OHD1bNnT/fwjBkz9MUXX9RqXXZFUAMA6mTJkiUaMmSI+0MbERER6ty5s9atW6fXXntNDodD27Zt08yZM8+6vNPpVMeOHSWdur9o+/btWrdunXr06OGe53e/+502btyotLQ0LVmyRP7+/rrmmms0bNgwPf/880pNTVWXLl00f/5894c2brrpJqWkpGjr1q2aO3euWrRo4d7ezJkztXnzZm3dutVjO2dj9ecw+XoWADQhL774oq644op6XWdaWpoeeeSRc04vKSnRxo0bNXjwYK1YsUIjR47UokWLJElPPvmkSkpK5O3trS+++EKXX3650tPTz7qeXr16aeTIkbriiivk6+urlJQUbd68WZK0dOlSvfnmm5JOfeBj/PjxevXVV7VixQr9+9//1gcffOCxrpYtW+qtt97SgAEDlJ2drQULFujBBx/Uyy+/LEkqLi5W79699eCDD2ratGmaMGHCOffP6s9hckQNAKizM09/n3na+7e//a02b96s1NRUXXrppR6nqX/p+uuv17Jly3T06FGVlpZ6PAp82WWX6euvv9bWrVt1991369JLL620nx49esjpdCo7O1uStGDBAvXr1889fenSpZKkzZs3KzIystJ19e3bV++8846ks38O8/e//72CgoLkcrnkcDg0btw4zZgxQ5dffrkOHTpU6bqrgyNqAGhCKjvybUgffvihXnzxRcXExKh169ZKSUlRZGSkpk2bptjYWO3fv1/z588/5+ctq/LWW29p+PDh2rp1q8aOHasbbrihTv2e/lRmXT6T+eyzz+qjjz7Srbfeqm+++UaDBg1yfw5zyJAheuutt/TCCy+4Q762OKIGANTZ4cOHtWbNGs2bN899NN22bVsdPnxYBw4c0IUXXqjBgwdXuo6vv/5aw4cPV6tWrdSmTRuPd3UEBgZqz5498vX1dX+aUpJKS0sVGBhYYV3bt29XZGSkunbtKkkaPXq0vvrqq1rtm9Wfw+SIGgBQL5KSkrR8+XL3KfDTL7zKyspSQUGBvvnmm0qXT01N1fvvv68tW7Zo7969Hp+qfOqpp7Rhwwb99NNP2rBhgzuck5OTlZiYqIcfflh33HGHe/7jx49r3LhxWrx4sXx9feVwOPT666/Xar/s8DlMyx/+PrPq84UnFEVRzaF44UnjKl54AgBAE0JQAwBgYwQ1AAA2RlADQCNnjJGPj4/VbaAafHx8ZIyp0TIENQA0cnl5eRoyZAhhbXM+Pj4aMmSI8vLyarQcj2cBQCP30ksvaerUqRoxYoS8vLysbgfnYIxRXl6eXnrppRotR1ADQCNXUlKiGTNmWN0GGginvgEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABsjqAEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABsjqAEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABsjqAEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABsjqAEAsDGCGgAAGyOoAQCwsWoF9aBBg5SVlaXs7GxNnz79nPPdfvvtMsaod+/e7nFPPPGEsrOzlZWVpYEDB9a9YwAAmhlTWXl7e5ucnBwTFRVl/Pz8TFpamomOjq4wX5s2bcxXX31lvvvuO9O7d28jyURHR5u0tDTTokULExkZaXJycoy3t3el23M4HJVOpyiKoqimVpVlX5VH1H369FFOTo6cTqfKysqUnJysuLi4CvM988wzevbZZ3Xs2DH3uLi4OCUnJ+vEiRPKy8tTTk6O+vTpU9UmAQDAf1UZ1CEhISooKHAPFxYWKiQkxGOemJgYhYWF6eOPP67xsgAA4Nx867oCLy8vvfDCC7r33ntrvY4JEyZo4sSJkqTg4OC6tgQAQJNRZVAXFRUpLCzMPRwaGqqioiL3cGBgoC677DKtXbtWktSpUyetWLFCw4YNq3LZ0xITE5WYmChJcjgctd4ZAACaokovcPv4+Jjc3FwTGRnpvpmsZ8+e55x/zZo17pvJevbs6XEzWW5uLjeTURRFUdQvqrLsq/KI2uVyafLkyVq1apV8fHw0b948ZWRkaNasWdq0aZNWrlx5zmUzMjK0aNEiZWRk6OTJk5o0aZLKy8ur2iQAAPgvL51KbNtwOByKjY21ug0AAM6byrKPN5MBAGBjBDUAADZGUAMAYGMENQAANkZQAwBgYwQ1AAA2RlADAGBjBDUAADZGUAMAYGMENQAANkZQAwBgYwQ1AAA2RlADAGBjBDUAADZGUAMAYGMENQAANkZQAwBgYwQ1AAA2RlADAGBjBDUAADZGUAMAYGMENQAANkZQAwBgYwQ1AAA2RlADAGBjBDUAADZGUAMAYGMENQAANkZQAwBgYwQ1AAA2RlADAGBjBDUAADZGUAMAYGMENQAANkZQAwBgYwQ1AAA2RlADAGBjBDUAADZGUAMAYGMENQAANkZQAwBgYwQ1AAA2RlADAGBj1QrqQYMGKSsrS9nZ2Zo+fXqF6ffff7+2bt2q1NRUrVu3TtHR0ZKkiIgIHTlyRKmpqUpNTdWcOXPqt3sAAJoBU1l5e3ubnJwcExUVZfz8/ExaWpqJjo72mCcwMND9e+jQoeaTTz4xkkxERIRJT0+vdP2/LIfDUaP5KYqiKKqxV2XZV+URdZ8+fZSTkyOn06mysjIlJycrLi7OY57S0lL374CAABljqlotAACohiqDOiQkRAUFBe7hwsJChYSEVJjvoYceUk5Ojp577jk9/PDD7vFRUVFKSUnR2rVr1bdv33pqGwCA5qHebiZ77bXXdPHFF2v69On605/+JEnas2ePwsPD1atXL/3hD3/QwoULFRgYWGHZCRMmyOFwyOFwKDg4uL5aAgCg0asyqIuKihQWFuYeDg0NVVFR0TnnT05O1vDhwyVJJ06c0L59+yRJKSkpys3NVffu3Sssk5iYqNjYWMXGxqq4uLjGOwEAQFNVZVA7HA5169ZNkZGR8vPz08iRI7VixQqPeS6++GL37yFDhig7O1uSFBwcLG/vU5uIiopSt27dtHPnzvrsHwCAJs23qhlcLpcmT56sVatWycfHR/PmzVNGRoZmzZqlTZs2aeXKlZo8ebJuvvlmlZWVqaSkRGPHjpUk9evXT08//bTKyspUXl6uBx54QCUlJQ2+UwAANBVeOnX7t204HA7FxsZa3QYAAOdNZdnHm8kAALAxghoAABsjqAEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABsjqAEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABsjqAEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABsjqAEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABsjqAEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABsjqAEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABsjqAEAsDGCGgAAGyOoAQCwMYIaAAAbI6gBALAxghoAABurVlAPGjRIWVlZys7O1vTp0ytMv//++7V161alpqZq3bp1io6Odk974oknlJ2draysLA0cOLD+OgcAoJkwlZW3t7fJyckxUVFRxs/Pz6SlpZno6GiPeQIDA92/hw4daj755BMjyURHR5u0tDTTokULExkZaXJycoy3t3el23M4HJVOpyiKoqimVpVlX5VH1H369FFOTo6cTqfKysqUnJysuLg4j3lKS0vdvwMCAmSMkSTFxcUpOTlZJ06cUF5ennJyctSnT5+qNgkAAP7Lt6oZQkJCVFBQ4B4uLCzUVVddVWG+hx56SH/4wx/UokUL3XTTTe5l169f77FsSEhIhWUnTJigiRMnSpKCg4NrvhcAADRR9XYz2WuvvaaLL75Y06dP15/+9KcaLZuYmKjY2FjFxsaquLi4vloCAKDRqzKoi4qKFBYW5h4ODQ1VUVHROedPTk7W8OHDa7UsAADwVGVQOxwOdevWTZGRkfLz89PIkSO1YsUKj3kuvvhi9+8hQ4YoOztbkrRixQqNHDlSLVq0UGRkpLp166aNGzfW8y4AANB0VXmN2uVyafLkyVq1apV8fHw0b948ZWRkaNasWdq0aZNWrlypyZMn6+abb1ZZWZlKSko0duxYSVJGRoYWLVqkjIwMnTx5UpMmTVJ5eXmD7xQAAE2Fl07d/m0bDodDsbGxVrcBAMB5U1n28WYyAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbIygBgDAxghqAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbIygBgDAxghqAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbIygBgDAxghqAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbIygBgDAxghqAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbIygBgDAxghqAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbIygBgDAxghqAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbIygBgDAxghqAABsrFpBPWjQIGVlZSk7O1vTp0+vMP2RRx7R999/ry1btujzzz9XeHi4e9rJkyeVmpqq1NRUffjhh/XXOQAAzYSprLy9vU1OTo6Jiooyfn5+Ji0tzURHR3vMc8MNNxh/f38jyTzwwAMmOTnZPa20tLTS9f+yHA5HjeanKIqiqMZelWVflUfUffr0UU5OjpxOp8rKypScnKy4uDiPedauXaujR49KktavX6/Q0NCqVgsAAKqhyqAOCQlRQUGBe7iwsFAhISHnnH/8+PH65JNP3MOtWrWSw+HQd999VyHgAQBA5Xzrc2V33323rrzySvXv3989LiIiQrt371ZUVJS+/PJLpaena+fOnR7LTZgwQRMnTpQkBQcH12dLAAA0alUeURcVFSksLMw9HBoaqqKiogrzDRgwQE8++aSGDRumEydOuMfv3r1bkuR0OrV27VrFxMRUWDYxMVGxsbGKjY1VcXFxrXYEAICmqtIL3D4+PiY3N9dERka6bybr2bOnxzxXXHGFycnJMRdffLHH+KCgINOiRQsjyXTs2NHs2LGjwo1ovyxuJqMoiqKaW1WWfVWe+na5XJo8ebJWrVolHx8fzZs3TxkZGZo1a5Y2bdqklStX6vnnn1ebNm20ePFiSVJ+fr7i4q4DOZgAAB22SURBVOIUHR2tN954Q+Xl5fL29tbf//53ZWZmVrVJAADwX146ldi24XA4FBsba3UbAACcN5VlH28mAwDAxghqAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbIygBgDAxghqAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbIygBgDAxghqAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbIygBgDAxghqAABsjKAGAMDGCGoAAGyMoAYAwMYIagAAbKxJB7W3t7f+8pe/6NJLL7W6FQAAaqVJB3VwcLDuu+8+LV++XEFBQVa3AwBAjTXpoN67d69GjBih8PBwJSUlydu7Se8uAKAJavLJ9d1332nSpEm65ZZb9Ne//tXqdgAAqBFfqxs4H95880316tVLTzzxhNLS0vT+++9b3RIAANXS5I+oT5syZYrWrVunefPm6Ve/+pXV7QAAUC3NJqjLysp0xx13aN++fVq+fLk6duxodUsAAFSp2QS1dOrmst/85jfq1KmTFi1aJB8fH6tbAgCgUs0qqCVp06ZNmjhxom666Sb94x//sLodAAAq1SxuJvuld955RzExMXrkkUeUmpqqt99+2+qWAAA4q2Z3RH3aY489pi+++EJvvPGGrrzySqvbAQDgrJptULtcLt1111364YcftGzZMl100UVWtwQAQAXNNqgl6eeff9bw4cPVoUMHLVmyRH5+fla3BACAh2Yd1JK0ZcsWjRs3Tn379tXLL79sdTsAAHholjeT/dKiRYsUExOjJ554QqmpqUpMTLS6JQAAJHFE7fbkk0/qk08+0auvvqprr73W6nYAAJBEULuVl5dr1KhR2rVrlz744AN17tzZ6pYAACCoz7R//34NHz5cAQEBWrp0qVq2bGl1SwCAZo6g/oWMjAyNGTNGV111lebMmWN1OwCAZq5aQT1o0CBlZWUpOztb06dPrzD9kUce0ffff68tW7bo888/V3h4uHvamDFjtGPHDu3YsUNjxoypv84b0PLlyzVr1iyNGzdOkydPtrodAEAzZyorb29vk5OTY6Kiooyfn59JS0sz0dHRHvPccMMNxt/f30gyDzzwgElOTjaSTPv27U1ubq5p3769CQoKMrm5uSYoKKjS7Tkcjkqnn6/y8vIyy5cvN2VlZaZ///6W90NRFEU13aos+6o8ou7Tp49ycnLkdDpVVlam5ORkxcXFecyzdu1aHT16VJK0fv16hYaGSjp1JL569WqVlJRo//79Wr16tW655ZaqNmkLxhiNHj1a2dnZWrx4scdZAgAAzpcqgzokJEQFBQXu4cLCQoWEhJxz/vHjx+uTTz6p1bJ2U1paqri4OLVo0ULLli2Tv7+/1S0BAJqZer2Z7O6779aVV16p559/vkbLTZgwQQ6HQw6HQ8HBwfXZUp1lZ2crPj5eV1xxhd58802r2wEANDNVBnVRUZHCwsLcw6GhoSoqKqow34ABA/Tkk09q2LBhOnHiRI2WTUxMVGxsrGJjY1VcXFyrHWlIn3zyiZ588kmNGjVKjz76qNXtAACamUovcPv4+Jjc3FwTGRnpvpmsZ8+eHvNcccUVJicnx1x88cUe49u3b2927txpgoKCTFBQkNm5c6dp3759rS+oW12LFi0yJ0+eNAMHDrS8F4qiKKrpVBXZV/UKBg8ebLZv325ycnJMQkKCkWRmzZplhg4daiSZ1atXmx9++MGkpqaa1NRU8+GHH7qXHTdunMnOzjbZ2dnm3nvvrWuzllZAQIDZsmWL2bdvn+natavl/VAURVFNo+oc1DZq1vKKiooyxcXFJj093QQEBFjeD0VRFNX4q06PZ8GT0+nUXXfdpejoaC1YsMDqdgAATRxBXQtffPGFHnvsMY0YMUJPPvmk1e0AAJowgrqWXnzxRb3zzjt6+umnddttt1ndDgCgiSKo62DixIlKTU3Vu+++qx49eljdDgCgCSKo6+DYsWP6zW9+o+PHj2v58uVq27at1S0BAJoYgrqOCgoKdMcdd6hr165677335OXlZXVLAIAmhKCuB+vWrdOUKVN02223adasWVa3AwBoQnytbqCpmDNnjnr16qWnnnpKaWlpWrp0qdUtAQCaAI6o69GkSZP03XffacGCBbrsssusbgcA0AQQ1PXoxIkTGjFihA4ePKjly5erffv2VrcEAGjkCOp6tmfPHt1+++0KDQ1VcnKyfHx8rG4JANCIEdQNYMOGDZo0aZIGDhyov/3tb1a3AwBoxLiZrIHMnTtXMTExeuyxx5SamqqkpCSrWwIANEIcUTegqVOn6quvvnKHNgAANUVQN6CTJ0/qzjvvVHFxsZYtW6bg4GCrWwIANDIEdQP76aefNHz4cF144YVavHixfH252gAAqD6C+jxISUnRhAkTdMMNN+j//u//rG4HANCIcHh3nrz33nuKiYnRo48+qtTUVL311ltWtwQAaAQ4oj6Ppk+frs8//1yvv/66+vTpY3U7AIBGgKA+j1wul+666y7t3r1bS5cuVadOnaxuCQBgcwT1ebZv3z7FxcUpKChIH3zwgVq0aGF1SwAAGyOoLZCenq57771X1157rf75z39a3Q4AwMYIaossWbJEs2fP1v3336/777/f6nYAADZFUFvoqaee0scff6xXXnlFffv2tbodAIANEdQWKi8v16hRo+R0OrVkyRKFhIRY3RIAwGYIaosdOHBAcXFx8vf317Jly9SyZUurWwIA2AhBbQNZWVkaPXq0YmNj9cYbb1jdDgDARghqm1ixYoVmzJihsWPH6uGHH/aYFh8fL6fTKZfLJafTqfj4eIu6BABYwdipHA6H5T1YVV5eXmbp0qWmrKzM3HjjjUaSiY+PN4cOHTJnOnTokImPj7e8X4qiKKp+qorss77BGjTb5KtNmzZm27Zt5qeffjKRkZHG6XSas3E6nZb3SlEURdVPVZZ9nPq2mUOHDmn48OHy8fHRsmXLFB4eftb5zjUeANC0ENQ2lJOTo/j4eP3v//6vjh49etZ58vPzz3NXAAArENQ2tWrVKv3xj39UQECATpw44THt8OHDSkhIsKgzAMD5RFDb2HPPPaf3339fvr6++vHHH1VeXq68vDxNmDBBSUlJVrcHADgPfK1uAJW777771KNHD0VERKhHjx7KycmxuiUAwHlEUNvckSNHNHz4cG3atEn//ve/tWjRIh08eNCjDhw4UGGcy+WyunUAQD0gqBuBXbt26c4779S7776rhIQE+fj4VLnMkSNHzhrg5wr2c40/efLkedhDAMC5ENSNxNq1axUaGipJat26tdq2beuudu3aeQyfa3yXLl08xvv6Vv2f/+jRo/US+GVlZQ39JwKAJomgboSOHDmiI0eO6IcffqjTevz9/WsV+JGRkR7j/fz8qtzWsWPHPAJ8165dev/997VixQodO3asTvsBAE0ZQd2MHT16VEePHtWPP/5Yp/W0atWqxoEfGxur22+/XQcPHtQHH3ygd999V2vXrlV5eXk97R0ANA0ENers2LFjOnbsmPbu3VvtZby8vNSvXz/dc889uvPOOzVu3DgVFRVp4cKFeu+997Rly5YG7BgAGhfL33F6ZjX3d303x2rZsqUZMWKEWbZsmTl+/Lgxxpj09HQzffp0ExYWZnl/FEVRDV18lINqNNWhQwdz//33m3Xr1rk/QLJmzRozfvx4065dO8v7oyiKaogiqKlGWZGRkSYhIcFkZmYaY4w5duyYWbJkiRk+fLhp0aKF5f1RFEXVV9U5qAcNGmSysrJMdna2mT59eoXp119/vdm8ebMpKyszI0aM8Jh28uRJk5qaalJTU82HH35Y12apZlq9evUyL7zwgtmzZ48xxph9+/aZ119/3fTt29d4eXlZ3h9FUVRdqk5B7e3tbXJyckxUVJTx8/MzaWlpJjo62mOeiIgIc/nll5sFCxZUCOrS0tL6bJZq5uXj42MGDhxo3n77bVNaWur+Nvdf/vKXCv+7pCiKaixVp+9R9+nTRzk5OXI6nSorK1NycrLi4uI85tm1a5fS09N5tAYNzuVy6bPPPtOYMWN00UUXadSoUcrIyND06dOVkZGhzZs365FHHlGnTp2sbhUA6kWVQR0SEqKCggL3cGFhoUJCQqq9gVatWsnhcOi7776rEPCnTZgwQQ6HQw6HQ8HBwdVeN5q3I0eOKCkpSUOGDFFISIgefvhhnTx5Ui+88IIKCwu1atUqjR49Wm3atLG6VQCotQb/zGVERIRiY2M1atQovfTSS+rSpUuFeRITExUbG6vY2FgVFxc3dEtogvbu3atXXnlFV111lXr06KG//vWvuvjii/X222/rxx9/1Hvvvadbb721Wq9NBQA7qTKoi4qKFBYW5h4ODQ1VUVFRtTewe/duSZLT6dTatWsVExNTizaB6tuxY4dmzJihrl276tprr9Vbb72lQYMG6aOPPtLu3bv1z3/+U3369LG6TQCotkovcPv4+Jjc3FwTGRnpvpmsZ8+eZ513/vz5HjeTBQUFuR+j6dixo9mxY0eVN/xwMxnVEOXn52duu+02k5ycbI4ePWqMMWbHjh3mz3/+s+natavl/VEU1byrzo9nDR482Gzfvt3k5OSYhIQEI8nMmjXLDB061EgyV155pSkoKDCHDh0yxcXFZtu2bUaSueaaa8zWrVtNWlqa2bp1q7nvvvvq2ixF1bnatm1r7r33XvP5558bl8tljDHm22+/NZMmTTLBwcGW90dRVPMrXnhCUeeokJAQM23aNJOWlmaMMebEiRNm5cqV5q677jL+/v6W90dRVPMogpqiqlGXX365+fvf/27y8/ONMcYcPHjQzJ8/3wwYMMB4e3tb3h9FUU23CGqKqkF5eXmZ/v37m8TERLN//35jjDFFRUXmH//4h7niiiss74+iqKZXBDVF1bLO9mWvbdu2mSeeeMKEh4db3h9FUU2jKss+r//+sA2Hw6HY2Fir2wAq6NChg+68807dc8896tu3ryTp66+/1qeffqqMjAxlZmYqNzdXLpfL4k4BNDaVZR9BDdRCZGSkRo0apbvvvls9e/Z0jz9x4oR27NihzMxMd3hnZmZq+/btOn78uIUdA7AzghpoQIGBgbrkkksUHR2tnj17Kjo6WtHR0erSpYt8fHwknXpHudPprBDgmZmZKi0ttXgPAFitsuzjfYpAHZWWlrrfVX+mli1bqnv37hUCfNCgQWrRooV7vsLCwgoBnpGRwet0myFfX1+FhYUpKipKkZGRioyMVIcOHeTl5SVJ8vLyqvHv2i5X3+uTTp1xOnLkiI4cOaKjR4+6f1c27mzjm9vZKYIaaCDHjx9Xenq60tPTPcb7+PioS5cuFQJ8/PjxHh8QKS4urhDemZmZKiwsPN+7gnri4+Oj0NBQdwifGchRUVEKCQlxn4WRTp2JKSkpkTFGxpw6+Vmb37Vdrj7XJ0l+fn5q3bq1R7Vs2bLGf8fy8vIaBXtt/3Fw7NixGvfWEDj1jfMqPj5es2fPVnh4uPLz85WQkKCkpCSr27IFLy8vhYaGVgjwnj17qmPHju75SktLPU6dnw7wnTt38qlZi3l7e6tz584VAvj077CwMI8Pw5SXl6uoqEhOp1N5eXnuOj1cWFiokydPWrhHDc/b21v+/v4VAvxs4+oyvlWrVjXurby8vEKQnx7+4x//qP/85z/19nfg1DdsIT4+XomJiQoICJB06oasxMRESSKsderoo6CgQAUFBfrss888pl1wwQUeAd6zZ0/dfPPNGjt2rHueY8eOnfVGth07dujEiRPne3eaJC8vL/3P//xPhQA+PRweHi4/Pz+PZYqKipSXl6dvv/3WI5CdTqcKCgpUVlZm0d7YQ3l5uQ4fPqzDhw836HZ++Q+Cuv5D4Hw+3cERNc4bp9OpyMjICuPz8vIUFRV1/htqAtq2bes+8j7zCDwqKkre3qc+judyuZSbm3vWG9ka+v8cG6OLLrronEfEERERFU7V/vDDDxUC+PTv/Pz8Znc9FbXDXd+wBZfL5Q6PM5WXl3tcl0PdtWrVSt27d/c4Ao+Ojla3bt08bmTLz893B3hWVpYOHDigEydO1Koay/PjF1xwwVmvD58OYn9/f4/59+7dWyGAT//Oz8/X0aNHLdoTNCWc+oYt5Ofnn/WIOj8///w308QdO3ZMW7du1datWz3G+/r6qmvXrh5H39HR0Zo4caL7kkRtlZeXVwjvsrKyWgd/bausrEze3t4KDw8/6ynqX+5ncXGx8vLytG3bNq1cudLjWnFeXp6OHDlSp78LUFcENc6bhIQEj2vUknT48GElJCRY2FXzcvLkSW3fvl3bt2/X8uXL3eO9vLwUEhKigIAAtWjR4pzl5+dX6fSaVJs2baqcpzZ3BP9SSUmJnE6ntm/frlWrVnkcEe/atYvn2GF7BDXOm9M3jHHXt/0YY2z72Jevr2+N//Hg5eWl/Px87dq1SwcOHLB6F4A6IahxXiUlJRHMqJGTJ0/q5MmTnIJGs1Xxzh4AAGAbBDUAADZGUAMAYGMENQAANkZQAwBgYwQ1AAA2RlADAGBjBDUAADZGUAMAYGMENQAANkZQAwBgYwQ1UEvx8fFyOp1yuVxyOp2Kj4+3uiUATRAf5QBqIT4+3uOTnZGRkUpMTJQkPjoCoF5xRA3UwuzZsz2+qy1JAQEBmj17tkUdAWiqCGqgFsLDw2s0HgBqi6AGaiE/P79G4wGgtghqoBYSEhJ0+PBhj3GHDx9WQkKCRR0BaKoIaqAWkpKSNGHCBOXl5am8vFx5eXmaMGECN5IBqHfc9Q3UUlJSEsEMoMFxRA0AgI0R1AAA2BhBDQCAjRHUAADYGEENAICNEdQA+MAIYGM8ngU0c3xgBLA3jqiBZo4PjAD2Vq2gHjRokLKyspSdna3p06dXmH799ddr8+bNKisr04gRIzymjRkzRjt27NCOHTs0ZsyY+ukaQL3hAyOA/ZnKytvb2+Tk5JioqCjj5+dn0tLSTHR0tMc8ERER5vLLLzcLFiwwI0aMcI9v3769yc3NNe3btzdBQUEmNzfXBAUFVbo9h8NR6XSKouq3nE6nORun02l5bxTVXKqy7KvyiLpPnz7KycmR0+lUWVmZkpOTFRcX5zHPrl27lJ6ervLyco/xgwYN0urVq1VSUqL9+/dr9erVuuWWW6raJIDziA+MAPZWZVCHhISooKDAPVxYWKiQkJBqrbwuywI4P/jACGBvtrjre8KECZo4caIkKTg42OJugOaHD4wA9lXlEXVRUZHCwsLcw6GhoSoqKqrWyqu7bGJiomJjYxUbG6vi4uJqrRsAgOagyqB2OBzq1q2bIiMj5efnp5EjR2rFihXVWvmqVas0cOBABQUFKSgoSAMHDtSqVavq3DQAAM1JlXejDR482Gzfvt3k5OSYhIQEI8nMmjXLDB061EgyV155pSkoKDCHDh0yxcXFZtu2be5lx40bZ7Kzs012dra5995763TnG0VRVGUVHx9vnE6ncblcxul0mvj4eMt7oqjqVBXZZ32DNWiWoijqrBUfH28OHTrk8YjZoUOHCGuqUVSdHs8CgMaAN6yhqSKoATQJvGENTRVBDaBJyM/Pr9F4oLEgqAE0CbxhDU0VQQ2gSeANa2iqbPFmMgCoD7xhDU0RR9QAANgYQQ0ANhQfHy+n0ymXyyWn06n4+HirW4JFOPUNADYTHx+vxMRE93PhkZGRSkxMlCRO7TdDHFEDgM3w8haciaAGAJvh5S04E0ENADbDy1twJoIaAGyGl7fgTAQ1ANgML2/BmbjrGwBsiJe34DSOqAEADYpnwuuGI2oAQIPhmfC644gaANBgeCa87ghqAECD4ZnwuiOoAQANhmfC646gBgA0GJ4JrzuCGgDQYHgmvO4IagBAg0pKSlJUVJR8fHwUFRXVqEPaikfNeDwLAIBqsOpRM46oAQCoBqseNSOoAQCoBqseNSOoAQCoBqseNSOoAQCoBqseNSOoAQCoBqseNeOubwAAqsmKz49yRA0AgI0R1AAA2BhBDQCAjRHUAADYGEENAICNEdQAANgYQQ0AgI0R1AAA2BhBDQCAjRHUAADYGEENAICNEdQAANgYQQ0AgI0R1AAA2BhBDQCAjXlJMlY3caa9e/dq165d9brO4OBgFRcX1+s6rdBU9kNiX+yqqexLU9kPiX2xq/rel4iICF144YXnnG6aejkcDst7YD/Yl8ZQTWVfmsp+sC/2rfO5L5z6BgDAxghqAABszEfSTKubOB9SUlKsbqFeNJX9kNgXu2oq+9JU9kNiX+zqfO2L7W4mAwAA/x+nvgEAsLEmHdRz587Vjz/+qPT0dKtbqZPQ0FB9+eWX+v7777Vt2zY9/PDDVrdUay1bttSGDRuUlpambdu2aebMmVa3VCfe3t5KSUnRypUrrW6lTpxOp7Zu3arU1FQ5HA6r26mTdu3aafHixcrMzFRGRoauvvpqq1uqle7duys1NdVdBw4c0JQpU6xuq1amTp2qbdu2KT09XQsXLlTLli2tbqnWHn74YaWnp2vbtm3n9b+H5be5N1Rdf/31JiYmxqSnp1veS12qU6dOJiYmxkgybdq0Mdu3bzfR0dGW91XbCggIMJKMr6+vWb9+vbnqqqss76m29cgjj5j33nvPrFy50vJe6lJOp9N07NjR8j7qo9566y0zfvx4I8n4+fmZdu3aWd5TXcvb29vs2bPHhIeHW95LTatz585m586dplWrVkaSef/9983YsWMt76s2demll5r09HTj7+9vfHx8zOrVq03Xrl0b/r+/mrB169Zp3759VrdRZz/88INSU1MlSYcOHVJmZqZCQkIs7qr2Dh8+LEny8/OTn5+fjDEWd1Q7ISEhGjJkiN58802rW8F/tW3bVv369dPcuXMlSWVlZTpw4IDFXdXdgAEDlJubq/z8fKtbqRVfX1/5+/vLx8dHrVu31u7du61uqVaio6O1YcMGHT16VC6XS1999ZVuv/32Bt9ukw7qpigiIkIxMTHasGGD1a3Umre3t1JTU7V3716tXr1aGzdutLqlWnnppZf0+OOPq7y83OpW6swYo88++0ybNm3ShAkTrG6n1qKiovTTTz9p/vz5SklJUWJiolq3bm11W3U2cuRIJSUlWd1GrezevVv/+Mc/lJ+frz179ujAgQNavXq11W3VyrZt23T99derQ4cO8vf316233qqwsLAG3y5B3YgEBATogw8+0NSpU1VaWmp1O7VWXl6umJgYhYaGqk+fPrr00kutbqnGhgwZor179zaZR0369u2r3r17a/DgwZo0aZKuv/56q1uqFV9fX/Xq1Utz5sxRr169dPjwYT3xxBNWt1Unfn5+GjZsmBYvXmx1K7USFBSkuLg4RUVFqXPnzgoICNDdd99tdVu1kpWVpWeffVafffaZPv30U6WlpcnlcjX4dgnqRsLX11cffPCB3nvvPS1btszqdurFgQMHtGbNGt1yyy1Wt1Jj1113nYYNGyan06nk5GTddNNNeuedd6xuq9ZOn4r86aeftGzZMvXp08fijmqnsLBQhYWF7rM0S5YsUa9evSzuqm4GDx6slJQU7d271+pWauXmm2+W0+lUcXGxTp48qaVLl+raa6+1uq1amzdvnq688kr1799fJSUl2rFjR4Nvk6BuJObOnavMzEy9+OKLVrdSJ8HBwWrXrp0kqVWrVvr1r3+trKwsi7uquYSEBIWFhSkqKkojR47Ul19+qdGjR1vdVq20bt1abdq0cf8eOHCgtm3bZnFXtfPjjz+qoKBA3bt3l3Tq2m5GRobFXdVNfHx8oz3tLUn5+fm6+uqr5e/vL+nUf5PMzEyLu6q9Cy64QJIUFham22+/XQsXLjwv27X8TrqGqoULF5rdu3ebEydOmIKCAnPfffdZ3lNt6rrrrjPGGLNlyxaTmppqUlNTzeDBgy3vqzZ1+eWXm5SUFLNlyxaTnp5unnrqKct7qmv179+/Ud/1HRUVZdLS0kxaWprZtm2bSUhIsLynutSvfvUr43A4zJYtW8yyZctMUFCQ5T3Vtlq3bm2Ki4tN27ZtLe+lLjVz5kyTmZlp0tPTzdtvv21atGhheU+1ra+//tp8//33Ji0tzdx0003nZZu8mQwAABvj1DcAADZGUAMAYGMENQAANkZQAwBgYwQ1AAA2RlADAGBjBDUAADZGUAMAYGP/D9jw8cRkIhVTAAAAAElFTkSuQmCC\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "-ygvHcHU4Nkj" + }, + "source": [ + "\n", + "# Part 5: Test the model\n" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "SYIAQDjhsHNQ" + }, + "source": [ + "# Convert from one-hot encoding (3D array) to 2D array \n", + "test_padded_tags_pred = model.predict(test_padded_sequences)\n", + "test_padded_tags_pred = np.argmax(test_padded_tags_pred, axis=-1)\n", + "test_padded_tags_true = np.argmax(test_padded_tags, axis=-1)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "aS4oaqkBE4XX" + }, + "source": [ + "# Convert the index to tag\n", + "test_tags_pred =[0]*len(test_padded_tags_pred)\n", + "for idx, row in enumerate(test_padded_tags_pred):\n", + " add = []\n", + " for i in row:\n", + " add.append(reverse_tag_map[i]) if i != 0 else add.append(\"PAD\")\n", + " test_tags_pred[idx] = add\n", + "\n", + "test_tags_true =[0]*len(test_padded_tags_true)\n", + "for idx, row in enumerate(test_padded_tags_true):\n", + " add = []\n", + " for i in row:\n", + " add.append(reverse_tag_map[i]) if i != 0 else add.append(\"PAD\")\n", + " test_tags_true[idx] = add" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "OO5Pg9r2MmV8", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "96156078-c726-4d4d-8f4e-43886bdc7f1c" + }, + "source": [ + "print(\"Micro F1-score is : {:.1%}\".format(f1_score(test_tags_true, test_tags_pred)))\n", + "print(\"Micro Precision-score is : {:.1%}\".format(precision_score(test_tags_true, test_tags_pred)))\n", + "print(\"Micro Recall-score is : {:.1%}\".format(recall_score(test_tags_true, test_tags_pred)))\n" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.7/dist-packages/seqeval/metrics/sequence_labeling.py:171: UserWarning: PAD seems not to be NE tag.\n", + " warnings.warn('{} seems not to be NE tag.'.format(chunk))\n" + ], + "name": "stderr" + }, + { + "output_type": "stream", + "text": [ + "Micro F1-score is : 69.6%\n", + "Micro Precision-score is : 66.4%\n", + "Micro Recall-score is : 73.0%\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "znKBJcadMr9l", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "ef4323f4-fa78-45b7-c69f-bb68f54bdfb0" + }, + "source": [ + "report = flat_classification_report(y_pred=test_tags_pred, y_true=test_tags_true)\n", + "print(report)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1272: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n", + " _warn_prf(average, modifier, msg_start, len(result))\n" + ], + "name": "stderr" + }, + { + "output_type": "stream", + "text": [ + " precision recall f1-score support\n", + "\n", + " B-Actor 0.91 0.95 0.93 1274\n", + " B-Award 0.61 0.52 0.56 66\n", + "B-Character_Name 0.65 0.48 0.55 283\n", + " B-Director 0.87 0.87 0.87 425\n", + " B-Genre 0.78 0.83 0.81 789\n", + " B-Opinion 0.47 0.53 0.50 195\n", + " B-Origin 0.54 0.27 0.36 190\n", + " B-Plot 0.55 0.43 0.48 1577\n", + " B-Quote 0.00 0.00 0.00 47\n", + " B-Relationship 0.80 0.62 0.70 171\n", + " B-Soundtrack 0.00 0.00 0.00 8\n", + " B-Year 0.94 0.97 0.96 661\n", + " I-Actor 0.91 0.95 0.93 1553\n", + " I-Award 0.64 0.74 0.69 147\n", + "I-Character_Name 0.64 0.41 0.50 227\n", + " I-Director 0.91 0.90 0.91 411\n", + " I-Genre 0.84 0.62 0.71 544\n", + " I-Opinion 0.50 0.02 0.04 143\n", + " I-Origin 0.68 0.69 0.69 808\n", + " I-Plot 0.90 0.92 0.91 14661\n", + " I-Quote 0.60 0.62 0.61 344\n", + " I-Relationship 0.59 0.36 0.44 289\n", + " I-Soundtrack 0.00 0.00 0.00 30\n", + " I-Year 0.71 0.57 0.63 44\n", + " O 0.85 0.89 0.86 14143\n", + " PAD 1.00 1.00 1.00 99704\n", + "\n", + " accuracy 0.96 138734\n", + " macro avg 0.65 0.58 0.60 138734\n", + " weighted avg 0.96 0.96 0.96 138734\n", + "\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "HE7Q2oD2M9rL", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "df0d97c9-fa16-4333-876c-915231440dc6" + }, + "source": [ + "# At every execution model picks some random test sample from test set.\n", + "i = np.random.randint(0,test_padded_sequences.shape[0]) # choose a random number between 0 and len(X_te)b\n", + "p = model.predict(np.array([test_padded_sequences[i]]))\n", + "p = np.argmax(p, axis=-1)\n", + "true = np.argmax(test_padded_tags[i], -1)\n", + "\n", + "print(\"Sample number {} of {} (Test Set)\".format(i, test_padded_sequences.shape[0]))\n", + "# Visualization\n", + "print(\"{:20}||{:20}||{}\".format(\"Word\", \"True\", \"Pred\"))\n", + "print(60 * \"=\")\n", + "for word, tag, pred in zip(test_padded_sequences[i], true, p[0]):\n", + " if word != 0:\n", + " print(\"{:20}: {:20} {}\".format(reverse_vocab[word], reverse_tag_map[tag], reverse_tag_map[pred]))\n", + "\n" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Sample number 1382 of 1954 (Test Set)\n", + "Word ||True ||Pred\n", + "============================================================\n", + "a : O O\n", + "animated : B-Genre B-Genre\n", + "movie : O O\n", + "about : O O\n", + "a : B-Plot B-Plot\n", + "father : I-Plot I-Plot\n", + "clown : I-Plot I-Plot\n", + "fish : I-Plot I-Plot\n", + "that : I-Plot I-Plot\n", + "ha : I-Plot I-Plot\n", + "lost : I-Plot I-Plot\n", + "his : I-Plot I-Plot\n", + "son : I-Plot I-Plot\n", + "in : I-Plot I-Plot\n", + "the : I-Plot I-Plot\n", + "deep : I-Plot I-Plot\n", + "blue : I-Plot I-Plot\n", + "ocean : I-Plot I-Plot\n", + "and : I-Plot I-Plot\n", + "wiwill : I-Plot I-Plot\n", + "stop : I-Plot I-Plot\n", + "at : I-Plot I-Plot\n", + "nothing : I-Plot I-Plot\n", + "to : I-Plot I-Plot\n", + "find : I-Plot I-Plot\n", + "him : I-Plot I-Plot\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "sNtJLetx4Sb_" + }, + "source": [ + "\n", + "# Part 6: Test with your own sentence" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "V0niUrOnVskX", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 758 + }, + "outputId": "77c0aa7b-7836-4909-edda-72903c0e893f" + }, + "source": [ + "# if this cell fails => run the 2nd time, it will work\n", + "\n", + "original_to_test = [\"is michael scofield the protagonist in prison break\", \n", + " \"what is the highest rated romantic movie in all time\"]\n", + "\n", + "to_test = apply_preproc(original_to_test)\n", + "\n", + "vocab_tokenizer.fit_on_texts(to_test)\n", + "to_test = vocab_tokenizer.texts_to_sequences(to_test)\n", + "\n", + "to_test = pad_sequences(to_test,\n", + " maxlen=max_length, \n", + " truncating=trunc_type, \n", + " padding=pad_type)\n", + "\n", + "to_test_tag_pred = model.predict(to_test)\n", + "to_test_tag_pred = np.argmax(to_test_tag_pred, axis=-1)\n", + "\n", + "for i, row in enumerate(to_test_tag_pred):\n", + " print(\"\\n{:20}||{}\".format(\"Word\", \"Pred\"))\n", + " print(40 * \"=\")\n", + " for j, pred in enumerate(row):\n", + " words = original_to_test[i].split(' ')\n", + " length = len(words)\n", + " if pred != 0 and j < length:\n", + " print(\"{:20}: {}\".format(words[j], reverse_tag_map[pred]))\n", + " " + ], + "execution_count": null, + "outputs": [ + { + "output_type": "error", + "ename": "InvalidArgumentError", + "evalue": "ignored", + "traceback": [ + "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", + "\u001b[0;31mInvalidArgumentError\u001b[0m Traceback (most recent call last)", + "\u001b[0;32m\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 14\u001b[0m padding=pad_type)\n\u001b[1;32m 15\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 16\u001b[0;31m \u001b[0mto_test_tag_pred\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpredict\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mto_test\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 17\u001b[0m \u001b[0mto_test_tag_pred\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margmax\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mto_test_tag_pred\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 18\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py\u001b[0m in \u001b[0;36mpredict\u001b[0;34m(self, x, batch_size, verbose, steps, callbacks, max_queue_size, workers, use_multiprocessing)\u001b[0m\n\u001b[1;32m 1725\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mstep\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mdata_handler\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msteps\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1726\u001b[0m \u001b[0mcallbacks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mon_predict_batch_begin\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mstep\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1727\u001b[0;31m \u001b[0mtmp_batch_outputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpredict_function\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0miterator\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 1728\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mdata_handler\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshould_sync\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1729\u001b[0m \u001b[0mcontext\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0masync_wait\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/def_function.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *args, **kwds)\u001b[0m\n\u001b[1;32m 887\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 888\u001b[0m \u001b[0;32mwith\u001b[0m \u001b[0mOptionalXlaContext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_jit_compile\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 889\u001b[0;31m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 890\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 891\u001b[0m \u001b[0mnew_tracing_count\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexperimental_get_tracing_count\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/def_function.py\u001b[0m in \u001b[0;36m_call\u001b[0;34m(self, *args, **kwds)\u001b[0m\n\u001b[1;32m 922\u001b[0m \u001b[0;31m# In this case we have not created variables on the first call. So we can\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 923\u001b[0m \u001b[0;31m# run the first trace but we should fail if variables are created.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 924\u001b[0;31m \u001b[0mresults\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_stateful_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 925\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_created_variables\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 926\u001b[0m raise ValueError(\"Creating variables on a non-first call to a function\"\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/function.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m 3022\u001b[0m filtered_flat_args) = self._maybe_define_function(args, kwargs)\n\u001b[1;32m 3023\u001b[0m return graph_function._call_flat(\n\u001b[0;32m-> 3024\u001b[0;31m filtered_flat_args, captured_inputs=graph_function.captured_inputs) # pylint: disable=protected-access\n\u001b[0m\u001b[1;32m 3025\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 3026\u001b[0m \u001b[0;34m@\u001b[0m\u001b[0mproperty\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/function.py\u001b[0m in \u001b[0;36m_call_flat\u001b[0;34m(self, args, captured_inputs, cancellation_manager)\u001b[0m\n\u001b[1;32m 1959\u001b[0m \u001b[0;31m# No tape is watching; skip to running the function.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1960\u001b[0m return self._build_call_outputs(self._inference_function.call(\n\u001b[0;32m-> 1961\u001b[0;31m ctx, args, cancellation_manager=cancellation_manager))\n\u001b[0m\u001b[1;32m 1962\u001b[0m forward_backward = self._select_forward_and_backward_functions(\n\u001b[1;32m 1963\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/function.py\u001b[0m in \u001b[0;36mcall\u001b[0;34m(self, ctx, args, cancellation_manager)\u001b[0m\n\u001b[1;32m 594\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 595\u001b[0m \u001b[0mattrs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mattrs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 596\u001b[0;31m ctx=ctx)\n\u001b[0m\u001b[1;32m 597\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 598\u001b[0m outputs = execute.execute_with_cancellation(\n", + "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/execute.py\u001b[0m in \u001b[0;36mquick_execute\u001b[0;34m(op_name, num_outputs, inputs, attrs, ctx, name)\u001b[0m\n\u001b[1;32m 58\u001b[0m \u001b[0mctx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mensure_initialized\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 59\u001b[0m tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,\n\u001b[0;32m---> 60\u001b[0;31m inputs, attrs, num_outputs)\n\u001b[0m\u001b[1;32m 61\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0mcore\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_NotOkStatusException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 62\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mname\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;31mInvalidArgumentError\u001b[0m: 2 root error(s) found.\n (0) Invalid argument: indices[0,2] = 9113 is not in [0, 9113)\n\t [[node model/embedding/embedding_lookup (defined at :16) ]]\n (1) Invalid argument: indices[0,2] = 9113 is not in [0, 9113)\n\t [[node model/embedding/embedding_lookup (defined at :16) ]]\n\t [[model/embedding/embedding_lookup/_21]]\n0 successful operations.\n0 derived errors ignored. [Op:__inference_predict_function_14054]\n\nErrors may have originated from an input operation.\nInput Source operations connected to node model/embedding/embedding_lookup:\n model/embedding/embedding_lookup/13534 (defined at /usr/lib/python3.7/contextlib.py:112)\n\nInput Source operations connected to node model/embedding/embedding_lookup:\n model/embedding/embedding_lookup/13534 (defined at /usr/lib/python3.7/contextlib.py:112)\n\nFunction call stack:\npredict_function -> predict_function\n" + ] + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "WZJApvzvaJcy" + }, + "source": [ + "\n", + "# Part 7: Analyse the incorrect predictions" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "fb8q_rD4aNRH" + }, + "source": [ + "def get_incorrect(y_pred, y_true, X_test):\n", + " y_pred.flatten()\n", + " y_true.flatten()\n", + " X_test.flatten()\n", + " where_incorrect = y_true != y_pred\n", + " incorrect_idxes = np.where(where_incorrect==1)[0]\n", + " incorrect_tokens = X_test[incorrect_idxes]\n", + " incorrect_tokens = dict(Counter(incorrect_tokens.flatten()))\n", + " incorrect_tags = y_true[incorrect_idxes]\n", + " incorrect_tags = dict(Counter(incorrect_tags.flatten()))\n", + " return incorrect_tokens, incorrect_tags\n", + "\n", + "\n", + "incorrect_tokens, incorrect_tags = get_incorrect(test_padded_tags_pred, \n", + " test_padded_tags_true, \n", + " test_padded_sequences) \n", + "\n", + "incorrect_tokens = sorted(incorrect_tokens.items(), key=lambda x:x[1], reverse=True)\n", + "incorrect_tags = sorted(incorrect_tags.items(), key=lambda x:x[1], reverse=True)\n", + "\n", + "print(\"{:^20}||{:^15}\".format(\"Incorrect word\", \"Frequency\"))\n", + "print(37 * \"=\")\n", + "for idx, count in incorrect_tokens[:20]:\n", + " if idx != 0:\n", + " print(\"{:20}: {:15}\".format(reverse_vocab[idx], count))\n", + "\n", + "print(\"\\n{:^20}||{:^15}\".format(\"Incorrect tag\", \"Frequency\"))\n", + "print(37 * \"=\")\n", + "for idx, count in incorrect_tags[:20]:\n", + " if idx != 0:\n", + " print(\"{:20}: {:15}\".format(reverse_tag_map[idx], count))\n" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "5rOw_59Ztzy7" + }, + "source": [ + "## Conclusion after analysis " + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "NiAN2Zy--j-_" + }, + "source": [ + "1. A lot of incorrect words are stopwords, due to their occurence in the plot. This is the problem of token label consistency.\n", + "2. The values of macro-average are much lower than micro-average of Precision, Recall and F1-score, as a result of imbalanced classes.\n", + "3. The number of incorrect Plot entities are so high, as Plot entities are the longest ones.\n", + "4. Lots of abbreviations that need to be cleaned. For example: country names (US -> u, s) or person name (J.K.Rowling => j, k, rowling)\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "N2J2LDiBRkyq" + }, + "source": [ + "## Potential improvements" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "9aBUF7SIHVhd" + }, + "source": [ + "1. To tackle imbalanced classes problem, use over-sample to gain more examples of tags from the minority groups.\n", + "2. To tackle label inconsistency, there're 3 solutions:\n", + "\n", + " * Use larger context. For example, use longer sentences, or combine 2 or more sentences that have similar/corelated meaning.\n", + " * Use CRF decoder layer.\n", + " * Use Character/Subword-level encoders like ELMO, Flair, CNN and BERT. \n", + "\n", + "3. To regconize long entity better:\n", + " * Incorporate POS tagging beside BIO tagging. \n", + " * Use ensemble to combine multiple models.\n", + "\n", + "\n", + "\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "NkJ8Go7BVVgz" + }, + "source": [ + "\n", + "# Export result to .tsv file" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "3EQxLZgzRdL9" + }, + "source": [ + "# write result to a new txt file\n", + "with open('/tmp/pred.tsv', 'wt') as out_file:\n", + " tsv_writer = csv.writer(out_file, delimiter='\\t')\n", + " test_size = len(test_padded_sequences)\n", + " for i in range(test_size):\n", + " for pred, word in zip(test_padded_tags_pred[i], test_padded_sequences[i]):\n", + " if pred != 0 and word != 0:\n", + " tsv_writer.writerow([reverse_tag_map[pred], reverse_vocab[word]])\n", + " tsv_writer.writerow([])" + ], + "execution_count": null, + "outputs": [] + } + ] +} \ No newline at end of file