011-RNN-LSTM-Toxic.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# RNN - LSTM - Toxic Comments\n",
    "\n",
    "A corpus of manually labeled comments - classifying each comment by its type of toxicity is available on Kaggle. We will aim to do a binary classification of whether a comment is toxic or not.\n",
    "\n",
    "Approach:\n",
    "- Learning Embedding with the Task\n",
    "- LSTM\n",
    "- BiLSTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import keras\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import vis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Get the Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Uncomment these shell lines to get the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# !wget http://bit.do/deep_toxic_train -P data/\n",
    "# !mv data/deep_toxic_train data/train.zip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(\"data/train.zip\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "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",
       "    </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>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</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>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</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>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</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>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</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>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</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...      0   \n",
       "1  000103f0d9cfb60f  D'aww! He matches this background colour I'm s...      0   \n",
       "2  000113f07ec002fd  Hey man, I'm really not trying to edit war. It...      0   \n",
       "3  0001b41b1c6bb37e  \"\\nMore\\nI can't make any real suggestions on ...      0   \n",
       "4  0001d958c54c6e35  You, sir, are my hero. Any chance you remember...      0   \n",
       "\n",
       "   severe_toxic  obscene  threat  insult  identity_hate  \n",
       "0             0        0       0       0              0  \n",
       "1             0        0       0       0              0  \n",
       "2             0        0       0       0              0  \n",
       "3             0        0       0       0              0  \n",
       "4             0        0       0       0              0  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Import the required libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create the Input & Output Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    Explanation\\nWhy the edits made under my usern...\n",
       "1    D'aww! He matches this background colour I'm s...\n",
       "2    Hey man, I'm really not trying to edit war. It...\n",
       "3    \"\\nMore\\nI can't make any real suggestions on ...\n",
       "4    You, sir, are my hero. Any chance you remember...\n",
       "Name: comment_text, dtype: object"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_sentences = df[\"comment_text\"]\n",
    "train_sentences.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Pre-processing the train data**\n",
    "- Tokenization: \"This is an apple\" -> [\"This\", \"is\", \"an\", \"apple\"]\n",
    "- Indexing: {0: \"This\", 1: \"is\", 2: \"an\", 3: \"apple\"}\n",
    "- Index Representation: [0, 1, 2, 3]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Tokenizer\n",
    "max_words = 20000\n",
    "tokenizer = Tokenizer(num_words=max_words, oov_token='UNK')\n",
    "tokenizer.fit_on_texts(list(train_sentences))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Tokenizer Fix from https://github.com/keras-team/keras/issues/8092"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer.word_index = {e:i for e,i in tokenizer.word_index.items() if i <= max_words} # <= because tokenizer is 1 indexed\n",
    "tokenizer.word_index[tokenizer.oov_token] = max_words + 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Index Representation\n",
    "tokenized_train = tokenizer.texts_to_sequences(train_sentences)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Selecting Padding\n",
    "# find length of each sentence and plot the length\n",
    "number_of_words = [len(comment) for comment in tokenized_train]\n",
    "plt.hist(number_of_words, bins = np.arange(0, 500, 10));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Padding to make it uniform\n",
    "maxlen = 200\n",
    "X = pad_sequences(tokenized_train, maxlen = maxlen)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = df.iloc[:,2].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9041555169799024"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Baseline Benchmark \n",
    "1 - df.iloc[:,2].sum()/df.iloc[:,2].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.utils import to_categorical\n",
    "y = to_categorical(labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((127656, 200), (31915, 200), (127656, 2), (31915, 2))"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)\n",
    "X_train.shape, X_test.shape, y_train.shape, y_test.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 2: Create the Model Architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, LSTM, Embedding, Dropout, Bidirectional"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Embedding(max_words, output_dim=128, mask_zero=True))\n",
    "model.add(LSTM(60))\n",
    "model.add(Dropout(0.1))\n",
    "model.add(Dense(2, activation='sigmoid'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 3: Compile the Model & Fit on the Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 102124 samples, validate on 25532 samples\n",
      "Epoch 1/5\n",
      "102124/102124 [==============================] - 398s 4ms/step - loss: 0.1625 - acc: 0.9454 - val_loss: 0.1185 - val_acc: 0.9601\n",
      "Epoch 2/5\n",
      "102124/102124 [==============================] - 352s 3ms/step - loss: 0.1096 - acc: 0.9618 - val_loss: 0.1050 - val_acc: 0.9629\n",
      "Epoch 3/5\n",
      "102124/102124 [==============================] - 436s 4ms/step - loss: 0.1006 - acc: 0.9642 - val_loss: 0.1022 - val_acc: 0.9642\n",
      "Epoch 4/5\n",
      "102124/102124 [==============================] - 333s 3ms/step - loss: 0.0954 - acc: 0.9654 - val_loss: 0.1071 - val_acc: 0.9635\n",
      "Epoch 5/5\n",
      "102124/102124 [==============================] - 329s 3ms/step - loss: 0.0923 - acc: 0.9665 - val_loss: 0.1026 - val_acc: 0.9640\n"
     ]
    }
   ],
   "source": [
    "output = model.fit(X_train, y_train, batch_size=128, epochs=5, validation_split=0.2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 4: Evaluate the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.vegalite.v2+json": {
       "$schema": "https://vega.github.io/schema/vega-lite/v2.4.3.json",
       "config": {
        "view": {
         "height": 300,
         "width": 400
        }
       },
       "data": {
        "values": [
         {
          "batch": 1,
          "metrics": "loss",
          "type": "validation",
          "value": 0.11854443359333913
         },
         {
          "batch": 2,
          "metrics": "loss",
          "type": "validation",
          "value": 0.10498868064149763
         },
         {
          "batch": 3,
          "metrics": "loss",
          "type": "validation",
          "value": 0.10217374689280567
         },
         {
          "batch": 4,
          "metrics": "loss",
          "type": "validation",
          "value": 0.10708715411239557
         },
         {
          "batch": 5,
          "metrics": "loss",
          "type": "validation",
          "value": 0.10257058775660276
         },
         {
          "batch": 1,
          "metrics": "acc",
          "type": "validation",
          "value": 0.9600892997490245
         },
         {
          "batch": 2,
          "metrics": "acc",
          "type": "validation",
          "value": 0.9629484568851674
         },
         {
          "batch": 3,
          "metrics": "acc",
          "type": "validation",
          "value": 0.9642213692714405
         },
         {
          "batch": 4,
          "metrics": "acc",
          "type": "validation",
          "value": 0.9634772051226732
         },
         {
          "batch": 5,
          "metrics": "acc",
          "type": "validation",
          "value": 0.9639667868240676
         },
         {
          "batch": 1,
          "metrics": "loss",
          "type": "training",
          "value": 0.16249100447295206
         },
         {
          "batch": 2,
          "metrics": "loss",
          "type": "training",
          "value": 0.10956678512316465
         },
         {
          "batch": 3,
          "metrics": "loss",
          "type": "training",
          "value": 0.10063836782718955
         },
         {
          "batch": 4,
          "metrics": "loss",
          "type": "training",
          "value": 0.09544190182134443
         },
         {
          "batch": 5,
          "metrics": "loss",
          "type": "training",
          "value": 0.09227849499671095
         },
         {
          "batch": 1,
          "metrics": "acc",
          "type": "training",
          "value": 0.9453654380878612
         },
         {
          "batch": 2,
          "metrics": "acc",
          "type": "training",
          "value": 0.9617768595274783
         },
         {
          "batch": 3,
          "metrics": "acc",
          "type": "training",
          "value": 0.9642297598717146
         },
         {
          "batch": 4,
          "metrics": "acc",
          "type": "training",
          "value": 0.9653999059826239
         },
         {
          "batch": 5,
          "metrics": "acc",
          "type": "training",
          "value": 0.9664770279128264
         }
        ]
       },
       "facet": {
        "column": {
         "field": "metrics",
         "type": "nominal"
        }
       },
       "resolve": {
        "scale": {
         "color": "shared",
         "x": "shared",
         "y": "independent"
        }
       },
       "spec": {
        "layer": [
         {
          "encoding": {
           "color": {
            "field": "type",
            "type": "nominal"
           },
           "tooltip": [
            {
             "field": "batch",
             "type": "quantitative"
            },
            {
             "field": "value",
             "type": "quantitative"
            }
           ],
           "x": {
            "field": "batch",
            "type": "quantitative"
           },
           "y": {
            "field": "value",
            "type": "quantitative"
           }
          },
          "height": 300,
          "mark": {
           "size": 50,
           "type": "circle"
          },
          "width": 300
         },
         {
          "encoding": {
           "color": {
            "field": "type",
            "type": "nominal"
           },
           "x": {
            "field": "batch",
            "type": "quantitative"
           },
           "y": {
            "field": "value",
            "type": "quantitative"
           }
          },
          "height": 300,
          "mark": "line",
          "width": 300
         }
        ]
       }
      },
      "image/png": "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",
      "text/plain": [
       "<VegaLite 2 object>\n",
       "\n",
       "If you see this message, it means the renderer has not been properly enabled\n",
       "for the frontend that you are using. For more information, see\n",
       "https://altair-viz.github.io/user_guide/troubleshooting.html\n"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vis.metrics(output.history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "31915/31915 [==============================] - 98s 3ms/step\n"
     ]
    }
   ],
   "source": [
    "score = model.evaluate(X_test, y_test, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.10434365738769805\n",
      "Test accuracy: 0.9634654551151496\n"
     ]
    }
   ],
   "source": [
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 5: Visualise evaluation & Make a prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "predict_classes = model.predict_classes(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "actual_classes = np.dot(y_test,np.array([[0],[1]])).reshape(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "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>col_0</th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>row_0</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0.0</th>\n",
       "      <td>28472</td>\n",
       "      <td>387</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1.0</th>\n",
       "      <td>785</td>\n",
       "      <td>2271</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "col_0      0     1\n",
       "row_0             \n",
       "0.0    28472   387\n",
       "1.0      785  2271"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.crosstab(actual_classes, predict_classes)"
   ]
  },
  {
   "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": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}