NER Model Tests.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# NER Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from gensim.models import KeyedVectors\n",
    "from nltk.corpus.reader.conll import ConllCorpusReader\n",
    "import nltk\n",
    "from nltk import conlltags2tree\n",
    "from IPython.display import display\n",
    "import re\n",
    "from math import floor\n",
    "import numpy as np\n",
    "import sys\n",
    "import pandas as pd\n",
    "import itertools\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.layers.recurrent import LSTM\n",
    "from keras.models import Sequential, Model, load_model\n",
    "from keras.layers import Dense, Bidirectional, Flatten, Dropout, TimeDistributed\n",
    "from keras.wrappers.scikit_learn import KerasClassifier\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.utils import to_categorical\n",
    "import keras.backend as K\n",
    "from keras_contrib.layers import CRF\n",
    "from keras_contrib.utils import save_load_utils\n",
    "from sklearn.model_selection import KFold, StratifiedKFold, cross_val_score, train_test_split\n",
    "from sklearn.metrics import confusion_matrix, f1_score, precision_score, recall_score"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "max_length = 70\n",
    "num_features = 300"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature Extraction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_padded_sentence_features(sentences, num_features, max_length, wv):\n",
    "    features = np.empty((0, max_length, num_features))\n",
    "    for i in range(len(sentences)):\n",
    "        if i % 100 == 0:\n",
    "            print(\"Processed\", i, \"of\", len(sentences))\n",
    "        sent = sentences[i]\n",
    "        new_sent = []\n",
    "        for j in range(max_length):\n",
    "            if 0 <= j < len(sent):\n",
    "                this_word = sent[j]\n",
    "                if this_word in wv.vocab:\n",
    "                    new_sent.append(wv.get_vector(this_word))\n",
    "                elif this_word == '':\n",
    "                    new_sent.append(np.zeros(num_features))\n",
    "                else:\n",
    "                    new_sent.append(np.random.uniform(-0.25,0.25, num_features))  # random vector for unknown\n",
    "            else:\n",
    "                new_sent.append(np.zeros(num_features))\n",
    "\n",
    "        feature_stack = np.dstack([[new_sent]])\n",
    "        features = np.vstack([features, feature_stack])\n",
    "        \n",
    "    return features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data Cleaning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "nums_regex = re.compile(r'0+')\n",
    "def clean_sents(sents, max_length):\n",
    "    cleaned = []\n",
    "    # remove sentences shorter than 5 words\n",
    "    for sent in sents:\n",
    "        if len(sent) > 4 and len(sent) <= max_length:\n",
    "            new_sent = []\n",
    "            # clean the words\n",
    "            for word in sent:\n",
    "                this_word = word.lower()\n",
    "                new_word = ''\n",
    "                # replace numbers with 0\n",
    "                for char in this_word:\n",
    "                    if char.isalpha():\n",
    "                        new_word = new_word + char\n",
    "                    elif char.isdigit():\n",
    "                        new_word = new_word + '0'\n",
    "                new_word = nums_regex.sub('0', new_word)\n",
    "                new_sent.append(this_word)\n",
    "            cleaned.append(new_sent)\n",
    "    return cleaned"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Helper function for getting the labels after prediction."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def arr2label(cats, labels):\n",
    "    new_labels = []\n",
    "    for i in range(len(cats)):\n",
    "        sent_labels = []\n",
    "        for j in range(len(cats[i])):\n",
    "            label = np.argmax(cats[i][j])\n",
    "            label = labels[label]\n",
    "            new_labels.append(label)\n",
    "    return new_labels\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load the English word vectors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "wv = KeyedVectors.load_word2vec_format('data/wiki.multi.en.vec.txt', binary=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define class variables for label encoding and decoding."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "classes = ['B-MISC', 'I-MISC', 'B-LOC', 'I-LOC', 'B-PER', 'I-PER', 'B-ORG', 'I-ORG', 'O']\n",
    "num_classes = len(classes)\n",
    "encoded_classes = range(num_classes)\n",
    "class2idx = {classes[enc]: enc for enc in encoded_classes}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Redefine the model before loading it from file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "dropout = 0.1\n",
    "recurrent_dropout = 0.3\n",
    "hidden_nodes = 100\n",
    "window_size = 70\n",
    "\n",
    "def create_model(num_classes, num_features, hidden_nodes=100):\n",
    "    model = Sequential()\n",
    "    model.add(Bidirectional(\n",
    "        LSTM(units=num_features, return_sequences=True, dropout=dropout, recurrent_dropout=recurrent_dropout),\n",
    "        input_shape=(window_size, num_features,),\n",
    "        merge_mode='concat'))\n",
    "    model.add(TimeDistributed(Dense(hidden_nodes, activation='relu')))\n",
    "    # add a CRF layer to enforce NER IOB rules\n",
    "    crf = CRF(num_classes, sparse_target=False)\n",
    "    model.add(crf)\n",
    "    print(\"Summary:\", model.summary())\n",
    "    model.compile(optimizer='rmsprop', loss=crf.loss_function, metrics=[crf.accuracy])\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load the model from file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "bidirectional_1 (Bidirection (None, 70, 600)           1442400   \n",
      "_________________________________________________________________\n",
      "time_distributed_1 (TimeDist (None, 70, 100)           60100     \n",
      "_________________________________________________________________\n",
      "crf_1 (CRF)                  (None, 70, 9)             1008      \n",
      "=================================================================\n",
      "Total params: 1,503,508\n",
      "Trainable params: 1,503,508\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Summary: None\n"
     ]
    }
   ],
   "source": [
    "\n",
    "model = create_model(num_classes, num_features)\n",
    "save_load_utils.load_all_weights(model,'models/full_train.h5', include_optimizer=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Test the Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run a test sentence."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_sent = \"My name is Tony DiPadova and I study at Dartmouth College\" "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processed 0 of 1\n"
     ]
    }
   ],
   "source": [
    "test_sent = nltk.word_tokenize(test_sent)  # tokenize the sentence\n",
    "test = clean_sents([test_sent], max_length)\n",
    "test_X = get_padded_sentence_features(test, num_features, max_length, wv)  # process the sentence"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['my', 'name', 'is', 'tony', 'dipadova', 'and', 'i', 'study', 'at', 'dartmouth', 'college']]\n",
      "11\n"
     ]
    }
   ],
   "source": [
    "pred = model.predict(test_X)  # predict\n",
    "labels = arr2label(pred, classes)  # get labels\n",
    "print(test)\n",
    "print(len(test[0]))\n",
    "labels = labels[:len(test[0])]  # remove padding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "tags = [(test_sent[i], labels[i]) for i in range(len(test_sent))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[('My', 'O'), ('name', 'O'), ('is', 'O'), ('Tony', 'B-PER'), ('DiPadova', 'I-PER'), ('and', 'O'), ('I', 'O'), ('study', 'O'), ('at', 'O'), ('Dartmouth', 'B-ORG'), ('College', 'I-ORG')]\n"
     ]
    }
   ],
   "source": [
    "print(tags)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "def tag2tree(tags):\n",
    "    string = '(S'\n",
    "    for i in range(len(tags)):\n",
    "        # beginning case\n",
    "        if i == 0 and tags[i][1] != 'O':\n",
    "            string += ' (' + tags[i][1][2:] + ' ' + tags[i][0]  # beginning is B-\n",
    "            if i + 1 < len(tags) and tags[i+1][1] == 'O':\n",
    "                string += ')'\n",
    "        elif tags[i][1] == 'O':\n",
    "            string += ' ' + tags[i][0]\n",
    "        else:\n",
    "            # middle cases\n",
    "            if i < len(tags) - 1:\n",
    "                if tags[i+1][1] == 'O' and tags[i][1][0] == 'I':\n",
    "                    string += ' ' + tags[i][0] + ')'\n",
    "                elif tags[i+1][1][0] == 'I' and tags[i][1][0] == 'I':\n",
    "                    string += ' ' + tags[i][0] \n",
    "                elif tags[i][1][0] == 'B' and tags[i+1][1] == 'O':\n",
    "                    string += ' (' + tags[i][1][2:] + ' ' + tags[i][0] + ')'\n",
    "                elif tags[i][1][0] == 'B' and tags[i+1][1][0] == 'I':\n",
    "                    string += ' (' + tags[i][1][2:] + ' ' + tags[i][0]\n",
    "            else:\n",
    "                # end cases\n",
    "                if tags[i][1][0] == 'B':\n",
    "                    string += ' (' + tags[i][1][2:] + ' ' + tags[i][0] + ')'\n",
    "                elif tags[i][1][0] == 'I':\n",
    "                    string += ' ' + tags[i][0] + ')'\n",
    "                \n",
    "    string += ')'\n",
    "    tree = nltk.Tree.fromstring(string)\n",
    "    return tree\n",
    "                \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(S\n",
      "  My\n",
      "  name\n",
      "  is\n",
      "  (PER Tony DiPadova)\n",
      "  and\n",
      "  I\n",
      "  study\n",
      "  at\n",
      "  (ORG Dartmouth College))\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "Tree('S', ['My', 'name', 'is', Tree('PER', ['Tony', 'DiPadova']), 'and', 'I', 'study', 'at', Tree('ORG', ['Dartmouth', 'College'])])"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tree = tag2tree(tags)\n",
    "tree.pprint()\n",
    "display(tree)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}