ToxicWikiCommentsEmbeddings.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import os\n",
    "import re\n",
    "import string\n",
    "import shutil\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import scipy as sp\n",
    "import seaborn as sn\n",
    "\n",
    "import sklearn as sk\n",
    "from sklearn.feature_selection import mutual_info_classif, SelectKBest, chi2\n",
    "from sklearn.feature_extraction.text import CountVectorizer, TfidfTransformer, TfidfVectorizer\n",
    "from sklearn.naive_bayes import GaussianNB\n",
    "from sklearn.linear_model import SGDClassifier\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.ensemble import AdaBoostClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.utils.class_weight import compute_class_weight\n",
    "from sklearn.utils.class_weight import compute_sample_weight\n",
    "\n",
    "from gensim.models import KeyedVectors\n",
    "\n",
    "import keras\n",
    "from keras.models import Sequential, Model\n",
    "from keras.layers import Dense, Activation, Dropout, LSTM, Input, GlobalMaxPooling1D, Conv1D, MaxPooling1D, Embedding, Flatten, Concatenate\n",
    "from keras.layers.embeddings import Embedding\n",
    "from keras.preprocessing import sequence\n",
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from keras.utils import to_categorical\n",
    "from tqdm import tqdm, tqdm_notebook, tnrange, trange"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "42d1135655084d99ba5b4f9039626b88",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "tqdm.monitor_interval = 0\n",
    "tqdm_notebook().pandas()\n",
    "sn.set_context('notebook')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prep Text Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_pickle('CLEAN/train.P.gz', compression='gzip')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "test = pd.read_pickle('CLEAN/test.P.gz', compression='gzip')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(159571, 9)\n",
      "Index(['id', 'comment_text', 'toxic', 'severe_toxic', 'obscene', 'threat',\n",
      "       'insult', 'identity_hate', 'cleaned_comments'],\n",
      "      dtype='object')\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>comment_text</th>\n",
       "      <th>toxic</th>\n",
       "      <th>severe_toxic</th>\n",
       "      <th>obscene</th>\n",
       "      <th>threat</th>\n",
       "      <th>insult</th>\n",
       "      <th>identity_hate</th>\n",
       "      <th>cleaned_comments</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0000997932d777bf</td>\n",
       "      <td>Explanation\\nWhy the edits made under my usern...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>explanation \\n why the edit make under my user...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>000103f0d9cfb60f</td>\n",
       "      <td>D'aww! He matches this background colour I'm s...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>daww he match this background colour I be seem...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>000113f07ec002fd</td>\n",
       "      <td>Hey man, I'm really not trying to edit war. It...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>hey man I be really not try to edit war its ju...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0001b41b1c6bb37e</td>\n",
       "      <td>\"\\nMore\\nI can't make any real suggestions on ...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>more \\n I can not make any real suggestion on ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0001d958c54c6e35</td>\n",
       "      <td>You, sir, are my hero. Any chance you remember...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>you sir be my hero any chance you remember wha...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 id                                       comment_text  toxic  \\\n",
       "0  0000997932d777bf  Explanation\\nWhy the edits made under my usern...  False   \n",
       "1  000103f0d9cfb60f  D'aww! He matches this background colour I'm s...  False   \n",
       "2  000113f07ec002fd  Hey man, I'm really not trying to edit war. It...  False   \n",
       "3  0001b41b1c6bb37e  \"\\nMore\\nI can't make any real suggestions on ...  False   \n",
       "4  0001d958c54c6e35  You, sir, are my hero. Any chance you remember...  False   \n",
       "\n",
       "   severe_toxic  obscene  threat  insult  identity_hate  \\\n",
       "0         False    False   False   False          False   \n",
       "1         False    False   False   False          False   \n",
       "2         False    False   False   False          False   \n",
       "3         False    False   False   False          False   \n",
       "4         False    False   False   False          False   \n",
       "\n",
       "                                    cleaned_comments  \n",
       "0  explanation \\n why the edit make under my user...  \n",
       "1  daww he match this background colour I be seem...  \n",
       "2  hey man I be really not try to edit war its ju...  \n",
       "3  more \\n I can not make any real suggestion on ...  \n",
       "4  you sir be my hero any chance you remember wha...  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(train.shape)\n",
    "print(train.columns)\n",
    "train.iloc[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(153164, 3)\n",
      "Index(['id', 'comment_text', 'cleaned_comments'], dtype='object')\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>comment_text</th>\n",
       "      <th>cleaned_comments</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>00001cee341fdb12</td>\n",
       "      <td>Yo bitch Ja Rule is more succesful then you'll...</td>\n",
       "      <td>yo bitch ja rule be more succesful then you wi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0000247867823ef7</td>\n",
       "      <td>== From RfC == \\n\\n The title is fine as it is...</td>\n",
       "      <td>from rfc the title be fine as it be imo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>00013b17ad220c46</td>\n",
       "      <td>\" \\n\\n == Sources == \\n\\n * Zawe Ashton on Lap...</td>\n",
       "      <td>source zawe ashton on lapland</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>00017563c3f7919a</td>\n",
       "      <td>:If you have a look back at the source, the in...</td>\n",
       "      <td>if you have a look back at the source the info...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>00017695ad8997eb</td>\n",
       "      <td>I don't anonymously edit articles at all.</td>\n",
       "      <td>I do not anonymously edit article at all</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 id                                       comment_text  \\\n",
       "0  00001cee341fdb12  Yo bitch Ja Rule is more succesful then you'll...   \n",
       "1  0000247867823ef7  == From RfC == \\n\\n The title is fine as it is...   \n",
       "2  00013b17ad220c46  \" \\n\\n == Sources == \\n\\n * Zawe Ashton on Lap...   \n",
       "3  00017563c3f7919a  :If you have a look back at the source, the in...   \n",
       "4  00017695ad8997eb          I don't anonymously edit articles at all.   \n",
       "\n",
       "                                    cleaned_comments  \n",
       "0  yo bitch ja rule be more succesful then you wi...  \n",
       "1            from rfc the title be fine as it be imo  \n",
       "2                      source zawe ashton on lapland  \n",
       "3  if you have a look back at the source the info...  \n",
       "4           I do not anonymously edit article at all  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(test.shape)\n",
    "print(test.columns)\n",
    "test.iloc[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "all_comments = pd.concat([train.cleaned_comments,test.cleaned_comments])\n",
    "tokenizer = Tokenizer()\n",
    "tokenizer.fit_on_texts(all_comments)\n",
    "word_index = tokenizer.word_index\n",
    "print('Found %s unique tokens.' % len(word_index))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 396618 unique tokens.\n"
     ]
    }
   ],
   "source": [
    "train_sequences = tokenizer.texts_to_sequences(train.cleaned_comments)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[581,\n",
       " 85,\n",
       " 2,\n",
       " 41,\n",
       " 47,\n",
       " 194,\n",
       " 29,\n",
       " 625,\n",
       " 3683,\n",
       " 8179,\n",
       " 774,\n",
       " 1,\n",
       " 148,\n",
       " 46,\n",
       " 1,\n",
       " 10,\n",
       " 228,\n",
       " 51,\n",
       " 4835,\n",
       " 18,\n",
       " 66,\n",
       " 1969,\n",
       " 156,\n",
       " 4,\n",
       " 517,\n",
       " 32,\n",
       " 125,\n",
       " 1084,\n",
       " 7899,\n",
       " 2144,\n",
       " 7,\n",
       " 52,\n",
       " 14,\n",
       " 10,\n",
       " 98,\n",
       " 2,\n",
       " 319,\n",
       " 30,\n",
       " 2,\n",
       " 56,\n",
       " 24,\n",
       " 152,\n",
       " 4,\n",
       " 1,\n",
       " 2655,\n",
       " 142857]"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_sequences[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|█████████████████████████████████████████████████████████████████████████████████████████████████| 159571/159571 [00:00<00:00, 780291.16it/s]\n"
     ]
    }
   ],
   "source": [
    "lengths = []\n",
    "for s in tqdm_notebook(train_sequences):\n",
    "    lengths.append(len(s))\n",
    "train[\"seq_len\"] = lengths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(153164, 3)\n",
      "153164\n"
     ]
    }
   ],
   "source": [
    "test_sequences = tokenizer.texts_to_sequences(test.cleaned_comments)\n",
    "print(test.shape)\n",
    "print(len(test_sequences))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|█████████████████████████████████████████████████████████████████████████████████████████████████| 153164/153164 [00:00<00:00, 741706.39it/s]\n"
     ]
    }
   ],
   "source": [
    "test_lengths = []\n",
    "for s in tqdm_notebook(test_sequences):\n",
    "    test_lengths.append(len(s))\n",
    "test[\"seq_len\"] = test_lengths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x20e11634e10>"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x20e2a7f1438>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sn.distplot(a=train['seq_len'])\n",
    "sn.distplot(a=test['seq_len'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1403\n",
      "2142\n"
     ]
    }
   ],
   "source": [
    "print(max(train['seq_len']))\n",
    "print(max(test['seq_len']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "# sequence_pad_len=2150\n",
    "sequence_pad_len=250\n",
    "train_data = pad_sequences(train_sequences, maxlen=sequence_pad_len)\n",
    "test_data = pad_sequences(test_sequences, maxlen=sequence_pad_len)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(159571, 250)"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(153164, 250)"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_data.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preparing the embedding Layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "embedding_dim = 25\n",
    "import pathlib\n",
    "sourcedir = pathlib.Path('GLOVE')\n",
    "embeddings_index = {}\n",
    "with open(sourcedir / (\"glove.twitter.27B.%dd.txt\" % embedding_dim), 'r', encoding='utf-8') 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",
    "# word2vec = KeyedVectors.load_word2vec_format(\n",
    "#             os.path.join(\"W2V\",\n",
    "#                 'w2v.twitter.27B.%dd.txt' % embedding_dim),\n",
    "#             binary=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "embedding_matrix = np.zeros((len(word_index) + 1, embedding_dim))\n",
    "for word, i in word_index.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"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Save the Embeddings and Sequences as pickles"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "os.makedirs(\"NN_INPUTS\", exist_ok=True)\n",
    "with open(\"NN_INPUTS/train.seq.P\", 'wb') as f:\n",
    "    pickle.dump(train_data, f)\n",
    "with open(\"NN_INPUTS/test.seq.P\", 'wb') as f:\n",
    "    pickle.dump(test_data, f)\n",
    "with open(\"NN_INPUTS/embedding.matrix.P\", 'wb') as f:\n",
    "    pickle.dump(embedding_matrix, f)\n",
    "with open(\"NN_INPUTS/word_index.P\", 'wb') as f:\n",
    "    pickle.dump(word_index, f)"
   ]
  },
  {
   "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}