textPrediction.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# NLP using LSTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow.keras.preprocessing.text import Tokenizer\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "from tensorflow.keras.layers import Embedding, LSTM, Dense, Flatten, Conv1D, GRU, Dropout, Bidirectional\n",
    "from tensorflow.keras.models import Sequential\n",
    "import pandas as pd\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Preprocessing on Data\n",
    "# Removing emojis from the chat"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "WhatsApp chat contains a lot of emojis. So need to filter out the input."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['28/08/18, 10:16 am - mummy: bhiya', \"28/08/18, 10:16 am - mummy: aa no par bhi what's app kri didhu\", '28/08/18, 10:16 am - mummy: mummy jode j hse', '28/08/18, 10:17 am - mummy: aetle ame bhar hoe ae to bhi tre vat thay ghre', '28/08/18, 12:52 pm - rp: naaa']\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "#This will read emojis as well\n",
    "non_bmp_map = dict.fromkeys(range(0x10000, sys.maxunicode + 1), 0xfffd)\n",
    "\n",
    "data = open(\"chat.txt\",encoding=\"unicode-escape\").read()\n",
    "corpus = data.lower().split(\"\\n\")\n",
    "corpus = corpus[1:]\n",
    "\n",
    "input_data = []\n",
    "i = 0\n",
    "for line in corpus:\n",
    "    text = line.translate(non_bmp_map)\n",
    "    input_data.append(line)\n",
    "    if len(text.split(\"-\")) == 2:\n",
    "        _,b = text.split(\"-\")\n",
    "        if ((\"rp\" in b) & ((len(b.split(\":\")) == 2))):\n",
    "            _,j = b.split(\":\")\n",
    "            if (not \"media omitted\" in j):\n",
    "                input_data.append(j)\n",
    "len(input_data)\n",
    "print(input_data[:5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = Tokenizer()\n",
    "tokenizer.fit_on_texts(input_data)\n",
    "total_words = len(tokenizer.word_index) + 1\n",
    "\n",
    "input_sentences = []\n",
    "for line in input_data:\n",
    "    token_list = tokenizer.texts_to_sequences([line])[0]\n",
    "    for i in range(1,len(token_list)):\n",
    "        n_gram = token_list[:i+1]\n",
    "        input_sentences.append(n_gram)\n",
    "        \n",
    "max_sequence_len = max([len(i) for i in input_sentences])\n",
    "input_sentences = np.array(pad_sequences(input_sentences, padding='pre', maxlen = max_sequence_len))\n",
    "\n",
    "training_data = input_sentences[:,:-1]\n",
    "labels = input_sentences[:,-1]\n",
    "\n",
    "training_labels = tf.keras.utils.to_categorical(labels, num_classes = total_words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras import regularizers\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(total_words, 64, input_length=max_sequence_len-1))\n",
    "model.add(Bidirectional(LSTM(150,return_sequences = True)))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(Bidirectional(LSTM(100)))\n",
    "model.add(Dense(total_words/2, activation='relu', kernel_regularizer=regularizers.l2(0.01)))\n",
    "model.add(Dense(total_words,activation='softmax'))\n",
    "model.compile(loss='categorical_crossentropy',optimizer='adam',metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding (Embedding)        (None, 36, 64)            92800     \n",
      "_________________________________________________________________\n",
      "bidirectional (Bidirectional (None, 36, 300)           258000    \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 36, 300)           0         \n",
      "_________________________________________________________________\n",
      "bidirectional_1 (Bidirection (None, 200)               320800    \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 725)               145725    \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1450)              1052700   \n",
      "=================================================================\n",
      "Total params: 1,870,025\n",
      "Trainable params: 1,870,025\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 16599 samples\n",
      "Epoch 1/7\n",
      "16599/16599 [==============================] - 92s 6ms/sample - loss: 3.0860 - accuracy: 0.3574\n",
      "Epoch 2/7\n",
      "16599/16599 [==============================] - 86s 5ms/sample - loss: 3.0575 - accuracy: 0.3586\n",
      "Epoch 3/7\n",
      "16599/16599 [==============================] - 98s 6ms/sample - loss: 3.0304 - accuracy: 0.3612\n",
      "Epoch 4/7\n",
      "16599/16599 [==============================] - 97s 6ms/sample - loss: 3.0040 - accuracy: 0.3674\n",
      "Epoch 5/7\n",
      "16599/16599 [==============================] - 98s 6ms/sample - loss: 2.9810 - accuracy: 0.3686\n",
      "Epoch 6/7\n",
      "16599/16599 [==============================] - 105s 6ms/sample - loss: 2.9595 - accuracy: 0.3752\n",
      "Epoch 7/7\n",
      "16599/16599 [==============================] - 105s 6ms/sample - loss: 2.9354 - accuracy: 0.3747\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(training_data, training_labels,batch_size=500, epochs = 7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "acc = history.history['accuracy']\n",
    "loss = history.history['loss']\n",
    "epochs = range(len(acc))\n",
    "\n",
    "plt.plot(epochs, acc, 'b', label='Training accuracy')\n",
    "plt.title(\"Training accuracy\")\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'b', label='Training loss')\n",
    "plt.title('Training loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "kem karu 6u\n"
     ]
    }
   ],
   "source": [
    "seed_text = \"kem\"\n",
    "next_words = 12\n",
    "  \n",
    "for _ in range(next_words):\n",
    "    token_list = tokenizer.texts_to_sequences([seed_text])[0]\n",
    "    token_list = pad_sequences([token_list], maxlen=max_sequence_len-1,padding='pre')\n",
    "    predicted = model.predict_classes(token_list, verbose=0)\n",
    "    output_word = \"\"\n",
    "    for word, index in tokenizer.word_index.items():\n",
    "        if index == predicted:\n",
    "            output_word = word\n",
    "            break\n",
    "    if output_word not in seed_text:\n",
    "        seed_text += \" \" + output_word\n",
    "print(seed_text)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### References"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "1. https://www.tensorflow.org/tutorials/text/text_generation\n",
    "2. \"Natural Language Processing in TensorFlow\" by deeplearning.ai from coursera.org"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}