Trollaganda.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "#IMPORTING ALL THE REQUIRED PACKAGES.\n",
    "import numpy as np \n",
    "import pandas as pd \n",
    "import matplotlib.pyplot as plt\n",
    "from subprocess import check_output\n",
    "import gensim\n",
    "import nltk\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.tokenize import RegexpTokenizer\n",
    "import os\n",
    "import re\n",
    "import codecs\n",
    "import keras\n",
    "import seaborn as sns\n",
    "import zipfile\n",
    "from keras.models import Model\n",
    "from keras.models import model_from_json\n",
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from keras.utils import to_categorical\n",
    "from keras.layers import Conv1D, MaxPooling1D, Embedding\n",
    "from keras.layers import LSTM, Bidirectional\n",
    "from keras.layers import Dense, Input, Flatten, Dropout, Concatenate\n",
    "from keras.callbacks import EarlyStopping\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package stopwords to\n",
      "[nltk_data]     /Users/jonathanl/nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Download Stuff\n",
    "nltk.download('stopwords')\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not os.path.isfile('./output.csv'):\n",
    "    zip_ref = zipfile.ZipFile('./troll-or-not.zip', 'r')\n",
    "    zip_ref.extractall('./')\n",
    "    zip_ref.close()\n",
    "    \n",
    "if not os.path.isfile('./GoogleNews-vectors-negative300.bin.gz'):\n",
    "    zip_ref = zipfile.ZipFile('./googles-trained-word2vec-model-in-python.zip', 'r')\n",
    "    zip_ref.extractall('./')\n",
    "    zip_ref.close()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "stopwords0 = set(stopwords.words('english'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0",
    "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a"
   },
   "outputs": [],
   "source": [
    "#SETTING GLOBAL VARIABLES\n",
    "EMBEDDINGDIM = 300\n",
    "MAXVOCABSIZE = 175303 \n",
    "MAXSEQLENGTH = 200 \n",
    "batchsize = 256 \n",
    "epochs = 3\n",
    "FROM_JSON = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "_uuid": "9076a1eab4a4923b85bc06d86ee4b8c891934d68"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "link said that tag is a little bitch\n"
     ]
    }
   ],
   "source": [
    "#CLEANING UP THE TEXT\n",
    "#Function to clean up the text\n",
    "\n",
    "def standardizeText(text):\n",
    "    text = text.str.replace(r\"http\\S+\", \"LINK\")\n",
    "    text = text.str.replace(r\"@\\S+\", \"TAG\")\n",
    "    text = text.str.replace(r\"[^A-Za-z0-9(),!?@\\'\\`\\\"\\_\\n]\", \" \")\n",
    "    text = text.str.replace(r\"@\", \"AT \")\n",
    "    text = text.str.lower()\n",
    "    return text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "_uuid": "c460c348800f37e20a449de7ebc5d7b6730cf7aa"
   },
   "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>message</th>\n",
       "      <th>isTroll</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>\"we have a sitting democrat us senator on tria...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>marshawn lynch arrives to game in anti trump s...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>daughter of fallen navy sailor delivers powerf...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>just in  president trump dedicates presidents ...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>19,000 respecting our national anthem!  standf...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             message  isTroll\n",
       "0  \"we have a sitting democrat us senator on tria...        1\n",
       "1  marshawn lynch arrives to game in anti trump s...        1\n",
       "2  daughter of fallen navy sailor delivers powerf...        1\n",
       "3  just in  president trump dedicates presidents ...        1\n",
       "4  19,000 respecting our national anthem!  standf...        1"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv('./output.csv',delimiter=',',encoding='utf8',names=[\"message\", \"isTroll\"])\n",
    "df.message = df.message.astype(str)\n",
    "\n",
    "df = df.iloc[:10000]\n",
    "\n",
    "df.message = standardizeText(df.message)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "_uuid": "17e12ac79eb36e3c8622b879f9ac536bafbafe0d"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Number of troll or not messages')"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Understand the distribution better\n",
    "sns.countplot(df.isTroll)\n",
    "plt.xlabel('Label')\n",
    "plt.title('Number of troll or not messages')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "_uuid": "f4e39a2748467ab03ee4cfc2423bdd54cb87f5a1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.series.Series'>\n"
     ]
    }
   ],
   "source": [
    "X = df.message\n",
    "Y = df.isTroll\n",
    "le = LabelEncoder()\n",
    "Y = le.fit_transform(Y)\n",
    "Y = Y.reshape(-1,1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "_uuid": "412e30bf594fc895b8cb9fb1448522715e523961"
   },
   "outputs": [],
   "source": [
    "# 9;10 * 10\n",
    "X_train,X_test,Y_train,Y_test = train_test_split(X,Y,test_size=0.15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "_cell_guid": "811d9d63-aa94-447a-b532-7fb7fe3f0cff",
    "_uuid": "5f4b8646b4358f91ef877e7c65146133c70dfcbb"
   },
   "outputs": [],
   "source": [
    "def tokenizeMessages(messages):\n",
    "    #TOKENIZING THE TEXT\n",
    "    regexTokenizer = RegexpTokenizer(r'\\w+')\n",
    "    messages = messages.apply(regexTokenizer.tokenize)\n",
    "    # delete Stop Words\n",
    "    messages = messages.apply(lambda vec: [word for word in vec if word not in stopwords0])\n",
    "    return messages\n",
    "\n",
    "train_tokens = tokenizeMessages(X_train)\n",
    "# test_tokens = tokenizeMessages(X_test)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "_cell_guid": "495c41d0-2afc-479c-b733-bb69206c4f70",
    "_uuid": "1273ae2dc9b3840eb073a70f2264ec7458d22baf"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "70882 words total, with a vocabulary size of 12518\n",
      "Max sentence length is 22\n"
     ]
    }
   ],
   "source": [
    "alltrainingwords = [word for tokens in train_tokens for word in tokens]\n",
    "trainingsentencelengths = [len(tokens) for tokens in train_tokens]\n",
    "TRAININGVOCAB = sorted(list(set(alltrainingwords)))\n",
    "\n",
    "print(\"%s words total, with a vocabulary size of %s\" % (len(alltrainingwords), len(TRAININGVOCAB)))\n",
    "print(\"Max sentence length is %s\" % max(trainingsentencelengths))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "_cell_guid": "f23d823d-ec6c-49a2-ba86-50b15d4a4593",
    "_uuid": "87f59235841f6549e372b74cc484ace578293441"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading W2V\n",
      "W2V Loaded\n"
     ]
    }
   ],
   "source": [
    "#WORD2VEC\n",
    "word2vecpath = \"./GoogleNews-vectors-negative300.bin.gz\"\n",
    "\n",
    "print(\"Loading W2V\")\n",
    "word2vec = gensim.models.KeyedVectors.load_word2vec_format(word2vecpath, binary=True)\n",
    "print(\"W2V Loaded\")\n",
    "\n",
    "def getaverageword2vec(tokenslist, vector, generatemissing=False, k=300):\n",
    "    if len(tokenslist)<1:\n",
    "        return np.zeros(k)\n",
    "    if generatemissing:\n",
    "        vectorized = [vector[word] if word in vector else np.random.rand(k) for word in tokenslist]\n",
    "    else:\n",
    "        vectorized = [vector[word] if word in vector else np.zeros(k) for word in tokenslist]\n",
    "    length = len(vectorized)\n",
    "    summed = np.sum(vectorized, axis=0)\n",
    "    averaged = np.divide(summed, length)\n",
    "    return averaged\n",
    "\n",
    "#GETTING EMBEDDINGS\n",
    "def getword2vecembeddings(vectors, messages, generatemissing=False):\n",
    "    embeddings = messages.apply(lambda x: getaverageword2vec(x, vectors,\n",
    "                                                                          generatemissing=generatemissing))\n",
    "    return list(embeddings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "_cell_guid": "b3595e29-0db7-4cc6-bd77-b6a8eb5214e7",
    "_uuid": "d342e226c056fbed5d4bdd5ac07068f6fc527362"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 13471 unique tokens.\n",
      "(13472, 300)\n"
     ]
    }
   ],
   "source": [
    "#TRAIN EMBEDDING\n",
    "trainingembeddings = getword2vecembeddings(word2vec, X_train, generatemissing=True)\n",
    "tokenizer = Tokenizer(num_words=MAXVOCABSIZE, lower=True, char_level=False)\n",
    "tokenizer.fit_on_texts(X_train)\n",
    "trainingsequences = tokenizer.texts_to_sequences(X_train)\n",
    "\n",
    "trainwordindex = tokenizer.word_index\n",
    "print('Found %s unique tokens.' % len(trainwordindex))\n",
    "\n",
    "traincnndata = pad_sequences(trainingsequences, maxlen=MAXSEQLENGTH)\n",
    "\n",
    "trainembeddingweights = np.zeros((len(trainwordindex)+1, EMBEDDINGDIM))\n",
    "for word,index in trainwordindex.items():\n",
    "    trainembeddingweights[index,:] = word2vec[word] if word in word2vec else np.random.rand(EMBEDDINGDIM)\n",
    "print(trainembeddingweights.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "_cell_guid": "e9d59cc2-5bdd-4899-9a7c-3cabf46dcb56",
    "_uuid": "f52236073f542da9a7225ca82835c4f168e450ab"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[    0     0     0 ...    18  3130     1]\n",
      " [    0     0     0 ...     0   604  3309]\n",
      " [    0     0     0 ...    37   197   403]\n",
      " ...\n",
      " [    0     0     0 ...    43 10970   476]\n",
      " [    0     0     0 ...   392    19     1]\n",
      " [    0     0     0 ...    40    59     1]]\n"
     ]
    }
   ],
   "source": [
    "testsequences = tokenizer.texts_to_sequences(X_test)\n",
    "testcnndata = pad_sequences(testsequences, maxlen=MAXSEQLENGTH)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "_cell_guid": "dca2a8d0-bf4b-4d5f-873d-60abe362cabc",
    "_uuid": "f0a5a6760bc4123971b1047a453bb98bc96fd1d9"
   },
   "outputs": [],
   "source": [
    "#DEFINING THE CNN\n",
    "def ConvNet(embeddings, max_sequence_length, num_words, embedding_dim, trainable=False, extra_conv=True):\n",
    "    \n",
    "    embedding_layer = Embedding(num_words,\n",
    "                            embedding_dim,\n",
    "                            weights=[embeddings],\n",
    "                            input_length=max_sequence_length,\n",
    "                            trainable=trainable)\n",
    "\n",
    "    sequence_input = Input(shape=(max_sequence_length,), dtype='int32')\n",
    "    embedded_sequences = embedding_layer(sequence_input)\n",
    "\n",
    "    # Yoon Kim model (https://arxiv.org/abs/1408.5882)\n",
    "    convs = []\n",
    "    filter_sizes = [3,4,5]\n",
    "\n",
    "    for filter_size in filter_sizes:\n",
    "        l_conv = Conv1D(filters=128, kernel_size=filter_size, activation='relu')(embedded_sequences)\n",
    "        l_pool = MaxPooling1D(pool_size=3)(l_conv)\n",
    "        convs.append(l_pool)\n",
    "\n",
    "    l_merge = Concatenate(axis=1)(convs)\n",
    "\n",
    "    # add a 1D convnet with global maxpooling, instead of Yoon Kim model\n",
    "    conv = Conv1D(filters=128, kernel_size=3, activation='relu')(embedded_sequences)\n",
    "    pool = MaxPooling1D(pool_size=3)(conv)\n",
    "\n",
    "    if extra_conv==True:\n",
    "        x = Dropout(0.5)(l_merge)  \n",
    "    else:\n",
    "        # Original Yoon Kim model\n",
    "        x = Dropout(0.5)(pool)\n",
    "    x = Flatten()(x)\n",
    "    x = Dense(128, activation='relu')(x)\n",
    "    x = Dropout(0.5)(x)\n",
    "    # Finally, we feed the output into a Sigmoid layer.\n",
    "    # The reason why sigmoid is used is because we are trying to achieve a binary classification(1,0) \n",
    "    # for each of the 6 labels, and the sigmoid function will squash the output between the bounds of 0 and 1.\n",
    "    preds = Dense(1, activation='sigmoid')(x)\n",
    "\n",
    "    model = Model(sequence_input, preds)\n",
    "    model.compile(loss='binary_crossentropy',\n",
    "                  optimizer='adam',\n",
    "                  metrics=['acc'])\n",
    "    model.summary()\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "_cell_guid": "a97112a8-7502-4dd1-b012-3594eaea2c6b",
    "_uuid": "f17a51cf41ebc2c2ab966b7b41847b2e53da996d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /Users/jonathanl/PycharmProjects/trollaganda/venv/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "WARNING:tensorflow:From /Users/jonathanl/PycharmProjects/trollaganda/venv/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            (None, 200)          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "embedding_1 (Embedding)         (None, 200, 300)     4041600     input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv1d_1 (Conv1D)               (None, 198, 128)     115328      embedding_1[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "conv1d_2 (Conv1D)               (None, 197, 128)     153728      embedding_1[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "conv1d_3 (Conv1D)               (None, 196, 128)     192128      embedding_1[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling1d_1 (MaxPooling1D)  (None, 66, 128)      0           conv1d_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling1d_2 (MaxPooling1D)  (None, 65, 128)      0           conv1d_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling1d_3 (MaxPooling1D)  (None, 65, 128)      0           conv1d_3[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_1 (Concatenate)     (None, 196, 128)     0           max_pooling1d_1[0][0]            \n",
      "                                                                 max_pooling1d_2[0][0]            \n",
      "                                                                 max_pooling1d_3[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "dropout_1 (Dropout)             (None, 196, 128)     0           concatenate_1[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "flatten_1 (Flatten)             (None, 25088)        0           dropout_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_1 (Dense)                 (None, 128)          3211392     flatten_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_2 (Dropout)             (None, 128)          0           dense_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_2 (Dense)                 (None, 1)            129         dropout_2[0][0]                  \n",
      "==================================================================================================\n",
      "Total params: 7,714,305\n",
      "Trainable params: 3,672,705\n",
      "Non-trainable params: 4,041,600\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = ConvNet(trainembeddingweights, MAXSEQLENGTH, len(trainwordindex)+1, EMBEDDINGDIM, False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "_cell_guid": "88718843-bb8a-400a-8cfd-4ff4948580ca",
    "_uuid": "fe710bb2fc540742c06edabd2287068b469a6169"
   },
   "outputs": [],
   "source": [
    "#DEFINING CALLBACKS\n",
    "earlystopping = EarlyStopping(monitor='val_loss', min_delta=0.01, patience=4, verbose=1)\n",
    "callbackslist = [earlystopping]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "_cell_guid": "73063c77-cdf5-471c-8d56-a54bf439d886",
    "_uuid": "ba2258b0f57bf11c2f666c4a56c7cb3891cdbee8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training the Model\n",
      "WARNING:tensorflow:From /Users/jonathanl/PycharmProjects/trollaganda/venv/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Train on 7650 samples, validate on 850 samples\n",
      "Epoch 1/3\n",
      "7650/7650 [==============================] - 31s 4ms/step - loss: 0.4490 - acc: 0.8007 - val_loss: 0.3443 - val_acc: 0.8153\n",
      "Epoch 2/3\n",
      "7650/7650 [==============================] - 30s 4ms/step - loss: 0.3328 - acc: 0.8263 - val_loss: 0.2897 - val_acc: 0.8835\n",
      "Epoch 3/3\n",
      "7650/7650 [==============================] - 30s 4ms/step - loss: 0.2976 - acc: 0.8627 - val_loss: 0.2890 - val_acc: 0.8718\n"
     ]
    }
   ],
   "source": [
    "#USING PREMADE NETWORK\n",
    "if FROM_JSON:\n",
    "    json_file = open('./model.json', 'r')\n",
    "    loaded_model_json = json_file.read()\n",
    "    json_file.close()\n",
    "    loaded_model = model_from_json(loaded_model_json)\n",
    "    loaded_model.load_weights('./model.h5')\n",
    "    loaded_model.compile(loss='binary_crossentropy',\n",
    "                  optimizer='adam',\n",
    "                  metrics=['acc'])\n",
    "    model = loaded_model\n",
    "    print(\"Loaded Model from disk\")\n",
    "else:\n",
    "    #TRAINING THE NETWORK\n",
    "    print(\"Training the Model\")\n",
    "    hist = model.fit(traincnndata, Y_train, epochs=epochs, callbacks=callbackslist, validation_split=0.1, shuffle=True, batch_size=batchsize)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "_cell_guid": "9d530456-c94b-4255-9ccc-0314e6d333a4",
    "_uuid": "6738aa50eb235b3bf8ae51cea4311efc9690feb3"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1500/1500 [==============================] - 4s 3ms/step\n",
      "1500/1500 [==============================] - 2s 2ms/step\n",
      "Test set\n",
      "  Loss: 0.310\n",
      "  Accuracy: 0.869\n"
     ]
    }
   ],
   "source": [
    "ytest = model.predict(testcnndata, batch_size=1024, verbose=1)\n",
    "yresults = model.evaluate(testcnndata, Y_test)\n",
    "\n",
    "print('Test set\\n  Loss: {:0.3f}\\n  Accuracy: {:0.3f}'.format(yresults[0],yresults[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "_cell_guid": "44976f8e-e1d7-4bb5-b815-34ac33d0dcf1",
    "_uuid": "d50b4877cc148ad31b31d6355d0d9d315c60e390"
   },
   "outputs": [],
   "source": [
    "#CREATING THE SUBMISSION.CSV FILE\n",
    "# submissiondf = pd.DataFrame(columns=['id'] + labelnames)\n",
    "# submissiondf['id'] = testcomments['id'].values \n",
    "# submissiondf[labelnames] = ytest \n",
    "# submissiondf.to_csv(\"./cnn_submission.csv\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "_cell_guid": "2c0f0ac9-642a-45cf-bada-2735d8a1a457",
    "_uuid": "3afe119dd9db5940bda4c7c31e093629518b1c97"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#GENERATING THE GRAPHS\n",
    "plt.figure()\n",
    "plt.plot(hist.history['loss'], lw=2.0, color='b', label='train')\n",
    "plt.plot(hist.history['val_loss'], lw=2.0, color='r', label='val')\n",
    "plt.title('CNN sentiment')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Cross-Entropy Loss')\n",
    "plt.legend(loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "_cell_guid": "8eefaa45-ad86-4458-8218-e258572f4c48",
    "_uuid": "341a0047130d5f90add22851e6d42f64d4a23edd"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure()\n",
    "plt.plot(hist.history['acc'], lw=2.0, color='b', label='train')\n",
    "plt.plot(hist.history['val_acc'], lw=2.0, color='r', label='val')\n",
    "plt.title('CNN sentiment')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend(loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "_uuid": "535abeb5ec4d6e4ad4a51f0d052a5cad107a12df"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "history_dict = hist.history\n",
    "\n",
    "acc = history_dict['acc']\n",
    "val_acc = history_dict['val_acc']\n",
    "loss = history_dict['loss']\n",
    "val_loss = history_dict['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\n",
    "\n",
    "\n",
    "plt.clf() # clear figure\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving Model to ./model.json, weights to ./model.h5\n",
      "Model Saved\n"
     ]
    }
   ],
   "source": [
    "#Saving Model\n",
    "\n",
    "print(\"Saving Model to ./model.json, weights to ./model.h5\")\n",
    "model_json = model.to_json()\n",
    "\n",
    "with open('./model.json', \"w\") as json_file:\n",
    "    json_file.write(model_json)\n",
    "\n",
    "model.save_weights('./model.h5')\n",
    "print(\"Model Saved\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Message: [Some message @ https://google.com where @liran23 said he wants to read it], Prediction: [Troll]\n",
      "Message: [Yossi is the man and he loves to post links online], Prediction: [Troll]\n"
     ]
    }
   ],
   "source": [
    "# Prediction\n",
    "\n",
    "\n",
    "def tokenizePredictions(messages):\n",
    "    series = pd.Series(messages)\n",
    "    series = series.astype(str)\n",
    "    standardizeText(series)\n",
    "    tokenized = tokenizeMessages(series)\n",
    "    \n",
    "    predictSequences = tokenizer.texts_to_sequences(tokenized)\n",
    "    predictData = pad_sequences(predictSequences, maxlen=MAXSEQLENGTH)\n",
    "\n",
    "    return predictData\n",
    "\n",
    "def predictTrollOrNot(model, messages):\n",
    "    data = tokenizePredictions(messages)\n",
    "    predictions = model.predict(data)\n",
    "\n",
    "    for i in range(len(messages)):\n",
    "        print(\"Message: [%s], Prediction: [%s]\" % (messages[i], \"Troll\" if predictions[i] > 0.5 else \"Not a Troll\"))\n",
    "    \n",
    "    \n",
    "\n",
    "predictTrollOrNot(model, [\"Some message @ https://google.com where @liran23 said he wants to read it\", \"Yossi is the man and he loves to post links online\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": 1
}