nlp-03.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Named Entity Recognition (NER)\n",
    "---\n",
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 734,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# !pip install python-crfsuite\n",
    "\n",
    "import nltk\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import codecs\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "from itertools import chain\n",
    "from sklearn.metrics import classification_report, confusion_matrix\n",
    "from sklearn.preprocessing import MultiLabelBinarizer\n",
    "mlb = MultiLabelBinarizer()\n",
    "\n",
    "import sklearn\n",
    "import pycrfsuite\n",
    "\n",
    "from nltk.corpus.reader import ConllChunkCorpusReader\n",
    "from seqeval.metrics import accuracy_score\n",
    "from seqeval.metrics import classification_report\n",
    "from seqeval.metrics import f1_score\n",
    "from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay\n",
    "from seqeval.scheme import IOB1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Reading Train and Test Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 735,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "15300 \n",
      "\n",
      "[('it', 'O'), ('is', 'O'), ('a', 'O'), ('series', 'O'), ('of', 'O'), ('badminton', 'O'), ('tournaments', 'O'), (',', 'O'), ('sanctioned', 'O'), ('by', 'O'), ('badminton', 'B-GRP'), ('world', 'I-GRP'), ('federation', 'I-GRP'), ('(', 'O'), ('bwf', 'O'), (')', 'O'), ('since', 'O'), ('2007', 'O'), ('.', 'O')]\n"
     ]
    }
   ],
   "source": [
    "mycorpus = ConllChunkCorpusReader(\"EN\", \"en_train_final.conll\", chunk_types=\"pos\")\n",
    "train_corpus = []\n",
    "for tree in mycorpus.tagged_sents():\n",
    "    train_corpus.append(tree)\n",
    "    \n",
    "print(len(train_corpus),'\\n')\n",
    "print(train_corpus[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 736,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "800 \n",
      "\n",
      "[('it', 'O'), ('was', 'O'), ('republished', 'O'), ('by', 'O'), ('mit', 'B-CORP'), ('press', 'I-CORP'), ('in', 'O'), ('1971', 'O'), ('and', 'O'), ('is', 'O'), ('still', 'O'), ('in', 'O'), ('print', 'O'), ('.', 'O')]\n"
     ]
    }
   ],
   "source": [
    "mycorpus = ConllChunkCorpusReader(\"EN\", \"en_dev_final.conll\",chunk_types=\"pos\")\n",
    "test_corpus = []\n",
    "for tree in mycorpus.tagged_sents():\n",
    "    test_corpus.append(tree)\n",
    "\n",
    "print(len(test_corpus),'\\n')\n",
    "print(test_corpus[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Defining Features\n",
    "Here we use word identity, word suffix, word shape. and some information from nearby words for the purpose."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 737,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Every word is represented by a set of features. CRF allows us to give any arbitrary set of features\n",
    "def word2features(sent, i):\n",
    "    word = sent[i][0]\n",
    "    features = [\n",
    "        'word[-3:]=' + word[-3:],                # last three letters of the word\n",
    "        'word[-2:]=' + word[-2:],                # last two letters of the word\n",
    "        'word.isdigit=%s' % word.isdigit()       # checking if the word contains all digits\n",
    "    ]\n",
    "    if i > 0:\n",
    "        word1 = sent[i-1][0]\n",
    "        features.extend([\n",
    "            '-1:word=' + word1                   # previous word\n",
    "        ])\n",
    "    else:\n",
    "        features.append('BOS')                   # tagging Beginning Of Sentence (BOS)\n",
    "        \n",
    "    if i < len(sent)-1:\n",
    "        word1 = sent[i+1][0]\n",
    "        features.extend([\n",
    "            '+1:word=' + word1                   # next word\n",
    "        ])\n",
    "    else:\n",
    "        features.append('EOS')                   # tagging End Of Sentence (EOS)\n",
    "                \n",
    "    return features\n",
    "\n",
    "def word2featuresTest(sent, i):\n",
    "    word = sent[i]\n",
    "    features = [\n",
    "        'word[-3:]=' + word[-3:],\n",
    "        'word[-2:]=' + word[-2:],\n",
    "        'word.isdigit=%s' % word.isdigit()\n",
    "    ]\n",
    "    if i > 0:\n",
    "        word1 = sent[i-1]\n",
    "        features.extend([\n",
    "            '-1:word=' + word1\n",
    "        ])\n",
    "    else:\n",
    "        features.append('BOS')\n",
    "        \n",
    "    if i < len(sent)-1:\n",
    "        word1 = sent[i+1]\n",
    "        features.extend([\n",
    "            '+1:word=' + word1\n",
    "        ])\n",
    "    else:\n",
    "        features.append('EOS')\n",
    "                \n",
    "    return features\n",
    "\n",
    "def sent2features(sent):\n",
    "    return [word2features(sent, i) for i in range(len(sent))]\n",
    "\n",
    "def sent2labels(sent):\n",
    "    return [label for token, label in sent]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 738,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[('it', 'O'), ('is', 'O'), ('a', 'O'), ('series', 'O'), ('of', 'O'), ('badminton', 'O'), ('tournaments', 'O'), (',', 'O'), ('sanctioned', 'O'), ('by', 'O'), ('badminton', 'B-GRP'), ('world', 'I-GRP'), ('federation', 'I-GRP'), ('(', 'O'), ('bwf', 'O'), (')', 'O'), ('since', 'O'), ('2007', 'O'), ('.', 'O')] \n",
      "\n",
      "('badminton', 'B-GRP')\n",
      "['word[-3:]=ton', 'word[-2:]=on', 'word.isdigit=False', '-1:word=by', '+1:word=world'] \n",
      "\n",
      "('2007', 'O')\n",
      "['word[-3:]=007', 'word[-2:]=07', 'word.isdigit=True', '-1:word=since', '+1:word=.'] \n",
      "\n"
     ]
    }
   ],
   "source": [
    "X_train = [sent2features(s) for s in train_corpus]\n",
    "y_train = [sent2labels(s) for s in train_corpus]\n",
    "\n",
    "X_test = [sent2features(s) for s in test_corpus]\n",
    "y_test = [sent2labels(s) for s in test_corpus]\n",
    "\n",
    "print(train_corpus[1],'\\n')\n",
    "\n",
    "print(train_corpus[1][10])\n",
    "print(X_train[1][10],'\\n')\n",
    "\n",
    "print(train_corpus[1][17])\n",
    "print(X_train[1][17],'\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 739,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# 1%%time\n",
    "trainer = pycrfsuite.Trainer(verbose=False)\n",
    "\n",
    "for xseq, yseq in zip(X_train, y_train):\n",
    "    trainer.append(xseq, yseq)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Hyperparameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 740,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "trainer.set_params({\n",
    "    'c1': 0.7,   # coefficient for L1 penalty\n",
    "    'c2': 0.03,  # coefficient for L2 penalty\n",
    "    'max_iterations': 75,  # stop earlier\n",
    "\n",
    "    # include transitions that are possible, but not observed\n",
    "    'feature.possible_transitions': True\n",
    "})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 741,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 13.3 s, sys: 1.42 ms, total: 13.3 s\n",
      "Wall time: 13.3 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "trainer.train('en.crfsuite')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 742,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<contextlib.closing at 0x7fb2c27f3ac0>"
      ]
     },
     "execution_count": 742,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tagger = pycrfsuite.Tagger()\n",
    "tagger.open('en.crfsuite')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Obtaining Predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 743,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 45.6 ms, sys: 3.89 ms, total: 49.5 ms\n",
      "Wall time: 48 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "y_pred = [tagger.tag(xseq) for xseq in X_test]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Dumping Predictions to Disk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 744,
   "metadata": {},
   "outputs": [],
   "source": [
    "with codecs.open('en.out.txt', 'w') as out_file:\n",
    "    for every_sent,pred_label in zip(test_corpus, y_pred):\n",
    "        for every_word_correct_label,predicted_label in zip(every_sent, pred_label):\n",
    "            out_file.write(every_word_correct_label[0] + \"\\t\" + every_word_correct_label[1] + \"\\t\" + predicted_label)\n",
    "            out_file.write(\"\\n\")\n",
    "        out_file.write(\"\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 745,
   "metadata": {},
   "outputs": [],
   "source": [
    "def conllReader(filename, word_field=0, label_field=1, prediction_field=2):\n",
    "    sentences_true_labels = []\n",
    "    sentences_pred_labels = []\n",
    "    true_list = []\n",
    "    pred_list = []\n",
    "    label_list = []\n",
    "    \n",
    "    with codecs.open(filename, 'r', errors='ignore', encoding='utf8') as f_in:\n",
    "        for line in f_in:\n",
    "            line = line.strip()\n",
    "            if line:\n",
    "                if line.startswith('#'):\n",
    "                    continue\n",
    "                label = line.split('\\t')[label_field]\n",
    "                pred = line.split('\\t')[prediction_field]\n",
    "                true_list.append( label )\n",
    "                pred_list.append( pred )\n",
    "                \n",
    "            else:\n",
    "                if len(true_list) > 0:\n",
    "                    sentences_true_labels.append( true_list )\n",
    "                    sentences_pred_labels.append( pred_list )\n",
    "                true_list = []\n",
    "                pred_list = []\n",
    "        f_in.close()\n",
    "        \n",
    "    return sentences_true_labels, sentences_pred_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 746,
   "metadata": {},
   "outputs": [],
   "source": [
    "true_labels, predicted_labels = conllReader(\"en.out.txt\")\n",
    "# predicted_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 747,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score is\n",
      "0.611214953271028 \n",
      "\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "        CORP       0.79      0.48      0.60       193\n",
      "          CW       0.41      0.16      0.23       176\n",
      "         GRP       0.74      0.59      0.66       190\n",
      "         LOC       0.65      0.59      0.62       234\n",
      "         PER       0.81      0.81      0.81       290\n",
      "        PROD       0.69      0.32      0.44       147\n",
      "\n",
      "   micro avg       0.72      0.53      0.61      1230\n",
      "   macro avg       0.68      0.49      0.56      1230\n",
      "weighted avg       0.69      0.53      0.59      1230\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('F1 Score is')\n",
    "print( f1_score(true_labels, predicted_labels) ,'\\n')\n",
    "\n",
    "print('Classification report')\n",
    "print( classification_report(true_labels, predicted_labels, scheme=IOB1) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 748,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7fb31979ce20>"
      ]
     },
     "execution_count": 748,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "cm = confusion_matrix(mlb.fit_transform(true_labels).argmax(axis=1), mlb.fit_transform(predicted_labels).argmax(axis=1))\n",
    "\n",
    "cm_display = ConfusionMatrixDisplay(cm)\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "fig.set_figheight(10)\n",
    "fig.set_figwidth(10)\n",
    "\n",
    "cmd.plot(xticks_rotation='vertical', ax =ax)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Attaching PoS Tagger-based features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 749,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<contextlib.closing at 0x7fb2c3968c10>"
      ]
     },
     "execution_count": 749,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "postagger = pycrfsuite.Tagger()\n",
    "postagger.open('en.crfsuite')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 750,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Every word is represented by a set of features. CRF allows us to give any arbitrary set of features\n",
    "def word2featurespos(sent, i):\n",
    "    word = sent[i][0]\n",
    "    postag = sent[i][2]\n",
    "    features = [\n",
    "        'word[-3:]=' + word[-3:],\n",
    "        'word[-2:]=' + word[-2:],\n",
    "        'word.isdigit=%s' % word.isdigit(),\n",
    "        'postag=' + postag,\n",
    "        'postag[:2]=' + postag[:2],\n",
    "    ]\n",
    "    if i > 0:\n",
    "        word1 = sent[i-1][0]\n",
    "        postag1 = sent[i-1][2]\n",
    "        features.extend([\n",
    "            '-1:word=' + word1,\n",
    "            '-1:postag=' + postag1,\n",
    "            '-1:postag[:2]=' + postag1[:2],\n",
    "        ])\n",
    "    else:\n",
    "        features.append('BOS')\n",
    "        \n",
    "    if i < len(sent)-1:\n",
    "        word1 = sent[i+1][0]\n",
    "        postag1 = sent[i+1][2]\n",
    "        features.extend([\n",
    "            '+1:word=' + word1,\n",
    "            '+1:postag=' + postag1,\n",
    "            '+1:postag[:2]=' + postag1[:2],\n",
    "        ])\n",
    "    else:\n",
    "        features.append('EOS')\n",
    "                \n",
    "    return features\n",
    "\n",
    "def sent2featurespos(sent, postagger):\n",
    "    tagged = postagger.tag(sent2features(sent))\n",
    "    sentNew = [word_nelabel+(pos, ) for word_nelabel,pos in zip(sent, tagged)]\n",
    "    return [word2featurespos(sentNew, i) for i in range(len(sentNew))]\n",
    "\n",
    "def sent2labelspos(sent):\n",
    "    return [label for token, label in sent]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 751,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[('his', 'O'), ('playlist', 'O'), ('includes', 'O'), ('sonny', 'B-PER'), ('sharrock', 'I-PER'), (',', 'O'), ('gza', 'B-PER'), (',', 'O'), ('country', 'B-GRP'), ('teasers', 'I-GRP'), ('and', 'O'), ('the', 'B-PER'), ('notorious', 'I-PER'), ('b.i.g.', 'I-PER')]\n",
      "\n",
      "\n",
      "[('his', 'O', 'O'), ('playlist', 'O', 'O'), ('includes', 'O', 'O'), ('sonny', 'B-PER', 'B-PER'), ('sharrock', 'I-PER', 'I-PER'), (',', 'O', 'O'), ('gza', 'B-PER', 'B-LOC'), (',', 'O', 'O'), ('country', 'B-GRP', 'O'), ('teasers', 'I-GRP', 'O'), ('and', 'O', 'O'), ('the', 'B-PER', 'O'), ('notorious', 'I-PER', 'O'), ('b.i.g.', 'I-PER', 'O')]\n"
     ]
    }
   ],
   "source": [
    "for sent in train_corpus:\n",
    "    print(sent)\n",
    "    print(\"\\n\")\n",
    "    s1 = postagger.tag(sent2features(sent))\n",
    "    sentNew = [word_nelabel+(pos, ) for word_nelabel,pos in zip(sent, s1)]\n",
    "    print(sentNew)\n",
    "    break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 752,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['word[-3:]=his', 'word[-2:]=is', 'word.isdigit=False', 'postag=O', 'postag[:2]=O', 'BOS', '+1:word=playlist', '+1:postag=O', '+1:postag[:2]=O']\n"
     ]
    }
   ],
   "source": [
    "X_train_pos = [sent2featurespos(s, postagger) for s in train_corpus]\n",
    "y_train_pos = [sent2labelspos(s) for s in train_corpus]\n",
    "\n",
    "X_test_pos = [sent2featurespos(s, postagger) for s in test_corpus]\n",
    "y_test_pos = [sent2labelspos(s) for s in test_corpus]\n",
    "\n",
    "print(X_train_pos[0][0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 753,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainerpos = pycrfsuite.Trainer(verbose=False)\n",
    "\n",
    "for xseq, yseq in zip(X_train_pos, y_train_pos):\n",
    "    trainerpos.append(xseq, yseq)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 754,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "trainerpos.set_params({\n",
    "    'c1': 1.0,   # coefficient for L1 penalty\n",
    "    'c2': 1e-3,  # coefficient for L2 penalty\n",
    "    'max_iterations': 100,  # stop earlier\n",
    "\n",
    "    # include transitions that are possible, but not observed\n",
    "    'feature.possible_transitions': True\n",
    "})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 755,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 23.6 s, sys: 2.95 ms, total: 23.6 s\n",
      "Wall time: 23.6 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "trainerpos.train('en-pos.crfsuite')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 756,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<contextlib.closing at 0x7fb2c54f2820>"
      ]
     },
     "execution_count": 756,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "taggerpos = pycrfsuite.Tagger()\n",
    "taggerpos.open('en-pos.crfsuite')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 757,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 39.2 ms, sys: 21 µs, total: 39.2 ms\n",
      "Wall time: 38.5 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "y_pred_pos = [taggerpos.tag(xseq) for xseq in X_test_pos]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 758,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "with codecs.open('en.out.pos.txt', 'w') as out_file:\n",
    "    for every_sent,pred_label in zip(test_corpus, y_pred_pos):\n",
    "        for every_word_correct_label,predicted_label in zip(every_sent, pred_label):\n",
    "            out_file.write(every_word_correct_label[0] + \"\\t\" + every_word_correct_label[1] + \"\\t\" + predicted_label)\n",
    "            out_file.write(\"\\n\")\n",
    "        out_file.write(\"\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 759,
   "metadata": {},
   "outputs": [],
   "source": [
    "true_labels, predicted_labels = conllReader(\"en.out.pos.txt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 760,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score is\n",
      "0.6122258516098927 \n",
      "\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "        CORP       0.79      0.48      0.60       193\n",
      "          CW       0.41      0.17      0.24       176\n",
      "         GRP       0.74      0.59      0.66       190\n",
      "         LOC       0.65      0.59      0.62       234\n",
      "         PER       0.81      0.81      0.81       290\n",
      "        PROD       0.70      0.33      0.44       147\n",
      "\n",
      "   micro avg       0.72      0.53      0.61      1230\n",
      "   macro avg       0.68      0.49      0.56      1230\n",
      "weighted avg       0.70      0.53      0.59      1230\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('F1 Score is')\n",
    "print( f1_score(true_labels, predicted_labels) ,'\\n')\n",
    "\n",
    "print('Classification report')\n",
    "print( classification_report(true_labels, predicted_labels, scheme=IOB1) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 761,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7fb31979ce20>"
      ]
     },
     "execution_count": 761,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "cm = confusion_matrix(mlb.fit_transform(true_labels).argmax(axis=1), mlb.fit_transform(predicted_labels).argmax(axis=1))\n",
    "\n",
    "cm_display = ConfusionMatrixDisplay(cm)\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "fig.set_figheight(10)\n",
    "fig.set_figwidth(10)\n",
    "\n",
    "cmd.plot(xticks_rotation='vertical', ax =ax)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Adding Gazetteer features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 762,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk.corpus.reader import plaintext"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 763,
   "metadata": {},
   "outputs": [],
   "source": [
    "personGazetteer = []\n",
    "\n",
    "mycorpus = plaintext.PlaintextCorpusReader(\"EN\", \"firstnames\")\n",
    "person = []\n",
    "for sent in mycorpus.words():\n",
    "    person.append(sent)\n",
    "    \n",
    "mycorpus = plaintext.PlaintextCorpusReader(\"EN\", \"lastnames\")\n",
    "for sent in mycorpus.words():\n",
    "    person.append(sent)\n",
    "\n",
    "person = [name.lower() for name in person]\n",
    "personGazetteer = set(person)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 764,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Every word is represented by a set of features. CRF allows us to give any arbitrary set of features\n",
    "def word2featuresGaz(sent, i, personGaz):\n",
    "    word = sent[i][0]\n",
    "    features = [\n",
    "        'word[-3:]=' + word[-3:],\n",
    "        'word[-2:]=' + word[-2:],\n",
    "        'word.isdigit=%s' % word.isdigit(),\n",
    "        'word.ispersongaz=%s' % (word in personGaz),\n",
    "    ]\n",
    "    if i > 0:\n",
    "        word1 = sent[i-1][0]\n",
    "        features.extend([\n",
    "            '-1:word=' + word1\n",
    "        ])\n",
    "    else:\n",
    "        features.append('BOS')\n",
    "        \n",
    "    if i < len(sent)-1:\n",
    "        word1 = sent[i+1][0]\n",
    "        features.extend([\n",
    "            '+1:word=' + word1\n",
    "        ])\n",
    "    else:\n",
    "        features.append('EOS')\n",
    "                \n",
    "    return features\n",
    "\n",
    "def word2featuresTest(sent, i):\n",
    "    word = sent[i]\n",
    "    features = [\n",
    "        'word[-3:]=' + word[-3:],\n",
    "        'word[-2:]=' + word[-2:],\n",
    "        'word.isdigit=%s' % word.isdigit()\n",
    "    ]\n",
    "    if i > 0:\n",
    "        word1 = sent[i-1]\n",
    "        features.extend([\n",
    "            '-1:word=' + word1\n",
    "        ])\n",
    "    else:\n",
    "        features.append('BOS')\n",
    "        \n",
    "    if i < len(sent)-1:\n",
    "        word1 = sent[i+1]\n",
    "        features.extend([\n",
    "            '+1:word=' + word1\n",
    "        ])\n",
    "    else:\n",
    "        features.append('EOS')\n",
    "                \n",
    "    return features\n",
    "\n",
    "def sent2featuresGaz(sent, personGaz):\n",
    "    return [word2featuresGaz(sent, i, personGaz) for i in range(len(sent))]\n",
    "\n",
    "def sent2labels(sent):\n",
    "    return [label for token, label in sent]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 765,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_gaz = [sent2featuresGaz(s, personGazetteer) for s in train_corpus]\n",
    "y_train_gaz = [sent2labels(s) for s in train_corpus]\n",
    "\n",
    "X_test_gaz = [sent2featuresGaz(s, personGazetteer) for s in test_corpus]\n",
    "y_test_gaz = [sent2labels(s) for s in test_corpus]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 766,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "trainergaz = pycrfsuite.Trainer(verbose=False)\n",
    "\n",
    "for xseq, yseq in zip(X_train_gaz, y_train_gaz):\n",
    "    trainergaz.append(xseq, yseq)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 767,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainergaz.set_params({\n",
    "    'c1': 1.0,   # coefficient for L1 penalty\n",
    "    'c2': 1e-3,  # coefficient for L2 penalty\n",
    "    'max_iterations': 150,  # stop earlier\n",
    "\n",
    "    # include transitions that are possible, but not observed\n",
    "    'feature.possible_transitions': True\n",
    "})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 768,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 26.8 s, sys: 891 µs, total: 26.8 s\n",
      "Wall time: 26.9 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "trainergaz.train('en-gaz.crfsuite')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 769,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<contextlib.closing at 0x7fb2c534b250>"
      ]
     },
     "execution_count": 769,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "taggergaz = pycrfsuite.Tagger()\n",
    "taggergaz.open('en-gaz.crfsuite')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 770,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 27.7 ms, sys: 996 µs, total: 28.7 ms\n",
      "Wall time: 28 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "y_pred_gaz = [taggergaz.tag(xseq) for xseq in X_test_gaz]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 771,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "with codecs.open('en.out.gaz.txt', 'w') as out_file:\n",
    "    for every_sent,pred_label in zip(test_corpus, y_pred_gaz):\n",
    "        for every_word_correct_label,predicted_label in zip(every_sent, pred_label):\n",
    "            out_file.write(every_word_correct_label[0] + \"\\t\" + every_word_correct_label[1] + \"\\t\" + predicted_label)\n",
    "            out_file.write(\"\\n\")\n",
    "        out_file.write(\"\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 772,
   "metadata": {},
   "outputs": [],
   "source": [
    "true_labels, predicted_labels = conllReader(\"en.out.gaz.txt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 773,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score is\n",
      "0.6121690664189503 \n",
      "\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "        CORP       0.76      0.46      0.57       193\n",
      "          CW       0.38      0.18      0.24       176\n",
      "         GRP       0.72      0.59      0.65       190\n",
      "         LOC       0.65      0.58      0.62       234\n",
      "         PER       0.82      0.83      0.83       290\n",
      "        PROD       0.74      0.34      0.47       147\n",
      "\n",
      "   micro avg       0.71      0.54      0.61      1230\n",
      "   macro avg       0.68      0.50      0.56      1230\n",
      "weighted avg       0.69      0.54      0.59      1230\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('F1 Score is')\n",
    "print( f1_score(true_labels, predicted_labels) ,'\\n')\n",
    "\n",
    "print('Classification report')\n",
    "print( classification_report(true_labels, predicted_labels, scheme=IOB1) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 774,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7fb31979ce20>"
      ]
     },
     "execution_count": 774,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "cm = confusion_matrix(mlb.fit_transform(true_labels).argmax(axis=1), mlb.fit_transform(predicted_labels).argmax(axis=1))\n",
    "\n",
    "cm_display = ConfusionMatrixDisplay(cm)\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "fig.set_figheight(10)\n",
    "fig.set_figwidth(10)\n",
    "\n",
    "cmd.plot(xticks_rotation='vertical', ax =ax)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Adding PoS Tag feature with the Gazetteer Feature"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 775,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Every word is represented by a set of features. CRF allows us to give any arbitrary set of features\n",
    "def word2featuresposGaz(sent, i, personGaz):\n",
    "    word = sent[i][0]\n",
    "    postag = sent[i][2]\n",
    "    features = [\n",
    "        'word[-3:]=' + word[-3:],\n",
    "        'word[-2:]=' + word[-2:],\n",
    "        'word.isdigit=%s' % word.isdigit(),\n",
    "        'postag=' + postag,\n",
    "        'postag[:2]=' + postag[:2],\n",
    "        'word.ispersongaz=%s' % (word in personGaz),\n",
    "    ]\n",
    "    if i > 0:\n",
    "        word1 = sent[i-1][0]\n",
    "        postag1 = sent[i-1][2]\n",
    "        features.extend([\n",
    "            '-1:word=' + word1,\n",
    "            '-1:postag=' + postag1,\n",
    "            '-1:postag[:2]=' + postag1[:2]\n",
    "        ])\n",
    "    else:\n",
    "        features.append('BOS')\n",
    "        \n",
    "    if i < len(sent)-1:\n",
    "        word1 = sent[i+1][0]\n",
    "        postag1 = sent[i+1][2]\n",
    "        features.extend([\n",
    "            '+1:word=' + word1,\n",
    "            '+1:postag=' + postag1,\n",
    "            '+1:postag[:2]=' + postag1[:2]\n",
    "        ])\n",
    "    else:\n",
    "        features.append('EOS')\n",
    "                \n",
    "    return features\n",
    "\n",
    "def sent2featuresposGaz(sent, personGaz, postagger):\n",
    "    tagged = postagger.tag(sent2features(sent))\n",
    "    sentNew = [word_nelabel+(pos, ) for word_nelabel,pos in zip(sent, tagged)]\n",
    "    return [word2featuresposGaz(sentNew, i, personGaz) for i in range(len(sentNew))]\n",
    "\n",
    "def sent2labels(sent):\n",
    "    return [label for token, label in sent]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 776,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_pos_gaz = [sent2featuresposGaz(s, personGazetteer, postagger) for s in train_corpus]\n",
    "y_train_pos_gaz = [sent2labels(s) for s in train_corpus]\n",
    "\n",
    "X_test_pos_gaz = [sent2featuresposGaz(s, personGazetteer,postagger) for s in test_corpus]\n",
    "y_test_pos_gaz = [sent2labels(s) for s in test_corpus]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 777,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "trainerposgaz = pycrfsuite.Trainer(verbose=False)\n",
    "\n",
    "for xseq, yseq in zip(X_train_pos_gaz, y_train_pos_gaz):\n",
    "    trainerposgaz.append(xseq, yseq)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 778,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "trainerposgaz.set_params({\n",
    "    'c1': 1.0,   # coefficient for L1 penalty\n",
    "    'c2': 1e-3,  # coefficient for L2 penalty\n",
    "    'max_iterations': 150,  # stop earlier\n",
    "\n",
    "    # include transitions that are possible, but not observed\n",
    "    'feature.possible_transitions': True\n",
    "})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 779,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 34.6 s, sys: 3.64 ms, total: 34.6 s\n",
      "Wall time: 34.7 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "trainerposgaz.train('en-pos-gaz.crfsuite')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 780,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<contextlib.closing at 0x7fb3193001c0>"
      ]
     },
     "execution_count": 780,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "taggerposgaz = pycrfsuite.Tagger()\n",
    "taggerposgaz.open('en-pos-gaz.crfsuite')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 781,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 52.2 ms, sys: 987 µs, total: 53.2 ms\n",
      "Wall time: 52.7 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "y_pred_pos_gaz = [taggerposgaz.tag(xseq) for xseq in X_test_pos_gaz]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 782,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "with codecs.open('en.out.pos.gaz.txt', 'w') as out_file:\n",
    "    for every_sent,pred_label in zip(test_corpus, y_pred_pos_gaz):\n",
    "        for every_word_correct_label,predicted_label in zip(every_sent, pred_label):\n",
    "            out_file.write(every_word_correct_label[0] + \"\\t\" + every_word_correct_label[1] + \"\\t\" + predicted_label)\n",
    "            out_file.write(\"\\n\")\n",
    "        out_file.write(\"\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 783,
   "metadata": {},
   "outputs": [],
   "source": [
    "true_labels, predicted_labels = conllReader(\"en.out.pos.gaz.txt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 784,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 Score is\n",
      "0.6131591227251516 \n",
      "\n",
      "Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "        CORP       0.79      0.48      0.60       193\n",
      "          CW       0.41      0.17      0.24       176\n",
      "         GRP       0.75      0.59      0.66       190\n",
      "         LOC       0.65      0.59      0.62       234\n",
      "         PER       0.81      0.81      0.81       290\n",
      "        PROD       0.70      0.33      0.44       147\n",
      "\n",
      "   micro avg       0.72      0.53      0.61      1230\n",
      "   macro avg       0.68      0.50      0.56      1230\n",
      "weighted avg       0.70      0.53      0.59      1230\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('F1 Score is')\n",
    "print( f1_score(true_labels, predicted_labels) ,'\\n')\n",
    "\n",
    "print('Classification report')\n",
    "print( classification_report(true_labels, predicted_labels, scheme=IOB1) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 785,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7fb31979ce20>"
      ]
     },
     "execution_count": 785,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "cm = confusion_matrix(mlb.fit_transform(true_labels).argmax(axis=1), mlb.fit_transform(predicted_labels).argmax(axis=1))\n",
    "\n",
    "cm_display = ConfusionMatrixDisplay(cm)\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "fig.set_figheight(10)\n",
    "fig.set_figwidth(10)\n",
    "\n",
    "cmd.plot(xticks_rotation='vertical', ax =ax)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Plotting the relation between values of c1, c2, maximum no. of iterations, and F1 Scores "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 786,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 20 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Using PairPlot to visualize the bivariate relations\n",
    "df = pd.DataFrame({'c1_values': [0.1, 0.1, 0.8, 0.5, 0.5, 0.7, 0.9, 0.2, 0.3, 0.2],\n",
    "                   'c2_values': [0.1, 0.02, 0.8, 0.1, 0.01, 0.5, 0.1, 0.001, 0.03, 0.1],\n",
    "                   'max_iterations': [100, 60, 90, 75, 75, 75, 100, 60, 300, 100],\n",
    "                   'F1_Scores': [62.0212, 61.5032, 61.2364, 61.3296, 62.4245, 61.3159, 62.9876, 63.8875, 63.4521, 61.0897]})\n",
    "sns.pairplot(df,\n",
    "             diag_kind=\"kde\",\n",
    "             height=2\n",
    "             )\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "interpreter": {
   "hash": "7553fb56ab976ad5aff909b9ab0e5f0d2a23e397f8c044edf0a757925c0fad4d"
  },
  "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": "3.9.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}