detectStanceLSTMDemon.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Stance Detection using Demonitisation Twitter DataSet and LSTM Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 2817 entries, 0 to 2816\n",
      "Data columns (total 3 columns):\n",
      "ID        2817 non-null object\n",
      "text      2814 non-null object\n",
      "Stance    2811 non-null object\n",
      "dtypes: object(3)\n",
      "memory usage: 66.1+ KB\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "file = \"H:\\Project Files\\stance-detection\\Dataset-Demonitization\\csv_files/demoapril.csv\"\n",
    "\n",
    "df = pd.read_csv(file,encoding='utf-8')\n",
    "df.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Tweet Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "import re\n",
    "import nltk\n",
    "from nltk.corpus import stopwords\n",
    "\n",
    "df = df.reset_index(drop=True)\n",
    "REPLACE_BY_SPACE_RE = re.compile('[/(){}\\[\\]\\|@,;]')\n",
    "BAD_SYMBOLS_RE = re.compile('[^0-9a-z #+_]')\n",
    "STOPWORDS = set(stopwords.words('english'))\n",
    "\n",
    "def clean_text(text):\n",
    "    \"\"\"\n",
    "        text: a string\n",
    "        \n",
    "        return: modified initial string\n",
    "    \"\"\"\n",
    "    text = str(text)\n",
    "    text = text.lower() # lowercase text\n",
    "    text = REPLACE_BY_SPACE_RE.sub(' ', text) # replace REPLACE_BY_SPACE_RE symbols by space in text. substitute the matched string in REPLACE_BY_SPACE_RE with space.\n",
    "    text = BAD_SYMBOLS_RE.sub('', text) # remove symbols which are in BAD_SYMBOLS_RE from text. substitute the matched string in BAD_SYMBOLS_RE with nothing. \n",
    "    text = text.replace('x', '')\n",
    "#     text = re.sub(r'\\W+', '', text)\n",
    "    text = ' '.join(word for word in text.split() if word not in STOPWORDS) # remove stopwors from text\n",
    "    return text\n",
    "\n",
    "df['text'] = df.text + ' ' + 'Demonitization'\n",
    "df['text'] = df['text'].apply(clean_text)\n",
    "df['text'] = df['text'].str.replace('\\d+', '')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 6540 unique tokens.\n"
     ]
    }
   ],
   "source": [
    "from keras.preprocessing.text import Tokenizer\n",
    "\n",
    "# The maximum number of words to be used. (most frequent)\n",
    "MAX_NB_WORDS = 50000\n",
    "# Max number of words in each complaint.\n",
    "MAX_SEQUENCE_LENGTH = 100\n",
    "# This is fixed.\n",
    "EMBEDDING_DIM = 100\n",
    "tokenizer = Tokenizer(num_words=MAX_NB_WORDS, filters='!\"#$%&()*+,-./:;<=>?@[\\]^_`{|}~', lower=True)\n",
    "tokenizer.fit_on_texts(df['text'].values)\n",
    "word_index = tokenizer.word_index\n",
    "print('Found %s unique tokens.' % len(word_index))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of data tensor: (2817, 100)\n"
     ]
    }
   ],
   "source": [
    "from keras.preprocessing.sequence import pad_sequences\n",
    "X = tokenizer.texts_to_sequences(df['text'].values)\n",
    "X = pad_sequences(X, maxlen=MAX_SEQUENCE_LENGTH)\n",
    "\n",
    "print('Shape of data tensor:', X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of label tensor: (2817, 3)\n"
     ]
    }
   ],
   "source": [
    "Y = pd.get_dummies(df['Stance']).values\n",
    "\n",
    "print('Shape of label tensor:', Y.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2535, 100) (2535, 3)\n",
      "(282, 100) (282, 3)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "X_train, X_test, Y_train, Y_test = train_test_split(X,Y, test_size = 0.1, random_state = 42)\n",
    "print(X_train.shape,Y_train.shape)\n",
    "print(X_test.shape,Y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LSTM Model - 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_15 (Embedding)     (None, 100, 100)          5000000   \n",
      "_________________________________________________________________\n",
      "spatial_dropout1d_8 (Spatial (None, 100, 100)          0         \n",
      "_________________________________________________________________\n",
      "lstm_14 (LSTM)               (None, 100)               80400     \n",
      "_________________________________________________________________\n",
      "dense_26 (Dense)             (None, 3)                 303       \n",
      "=================================================================\n",
      "Total params: 5,080,703\n",
      "Trainable params: 5,080,703\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 2281 samples, validate on 254 samples\n",
      "Epoch 1/5\n",
      "2281/2281 [==============================] - 16s 7ms/step - loss: 1.0664 - acc: 0.4244 - val_loss: 0.9976 - val_acc: 0.5669\n",
      "Epoch 2/5\n",
      "2281/2281 [==============================] - 11s 5ms/step - loss: 0.8043 - acc: 0.6550 - val_loss: 0.7266 - val_acc: 0.7087\n",
      "Epoch 3/5\n",
      "2281/2281 [==============================] - 10s 5ms/step - loss: 0.4467 - acc: 0.8632 - val_loss: 0.5495 - val_acc: 0.7835\n",
      "Epoch 4/5\n",
      "2281/2281 [==============================] - 10s 5ms/step - loss: 0.2274 - acc: 0.9404 - val_loss: 0.4750 - val_acc: 0.8189\n",
      "Epoch 5/5\n",
      "2281/2281 [==============================] - 11s 5ms/step - loss: 0.1197 - acc: 0.9667 - val_loss: 0.5121 - val_acc: 0.8110\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout\n",
    "from keras.layers import Embedding\n",
    "from keras.layers import LSTM\n",
    "from keras.layers import SpatialDropout1D\n",
    "from keras.callbacks import EarlyStopping\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(MAX_NB_WORDS, EMBEDDING_DIM, input_length=X.shape[1]))\n",
    "model.add(SpatialDropout1D(0.2))\n",
    "model.add(LSTM(100, dropout=0.2, recurrent_dropout=0.2))\n",
    "model.add(Dense(3, activation='softmax'))\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "epochs = 5\n",
    "batch_size = 64\n",
    "\n",
    "print(model.summary())\n",
    "\n",
    "history = model.fit(X_train, Y_train, epochs=epochs, batch_size=batch_size,validation_split=0.1,callbacks=[EarlyStopping(monitor='val_loss', patience=3, min_delta=0.0001)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "282/282 [==============================] - 0s 778us/step\n",
      "Test set\n",
      "  Loss: 0.705\n",
      "  Accuracy: 0.723\n"
     ]
    }
   ],
   "source": [
    "accr = model.evaluate(X_test,Y_test)\n",
    "print('Test set\\n  Loss: {:0.3f}\\n  Accuracy: {:0.3f}'.format(accr[0],accr[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "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": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.title('Loss')\n",
    "plt.plot(history.history['loss'], label='train')\n",
    "plt.plot(history.history['val_loss'], label='test')\n",
    "plt.legend()\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEICAYAAACktLTqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3Xl8VNXdx/HPjxCyEEJCwp4gqIggIGLYqrZuKLgrPhYVFVGx7m3tom211adPazdrbW0tKgIq4AIqKiraYrFuZJF9B0GGNQQIW0K28/xxJxBiIBPIrPm+X6+8zMyc3PvLlfnOybn3nmPOOUREJLY0C3cBIiLS+BTuIiIxSOEuIhKDFO4iIjFI4S4iEoMU7iIiMUjhLiISgxTuEnXM7CMz22FmCeGuRSRSKdwlqphZV+AswAGXhXC/zUO1L5HGoHCXaHMj8DkwAbip+kkzSzKzP5nZOjMrNrP/mlmS/7UzzexTM9tpZuvNbLT/+Y/M7NYa2xhtZv+t8diZ2V1mthJY6X/uL/5t7DKzfDM7q0b7ODP7mZmtNrPd/tezzewpM/tTzV/CzN4ys+8H4wCJgMJdos+NwEv+rwvNrL3/+T8CpwPfAtoAPwGqzKwL8C7wV6At0A+Y14D9XQEMAnr5H+f6t9EGmAy8amaJ/td+CFwLXASkAmOAfcBE4FozawZgZpnAecCUhvziIg2hcJeoYWZnAscBrzjn8oHVwHX+0BwD3Oec2+Ccq3TOfeqc2w9cD3zonJvinCt3zhU55xoS7r91zm13zpUAOOde9G+jwjn3JyAB6OFveyvwC+fccueZ7287FyjGC3SAkcBHzrktx3hIRA5L4S7R5CZglnNum//xZP9zmUAiXtjXln2Y5wO1vuYDM7vfzJb6h352Aq39+69vXxOBUf7vRwEvHENNIvXSSSKJCv7x82uAODPb7H86AUgDOgKlwAnA/Fo/uh4YeJjN7gWSazzuUEebA9Om+sfXf4rXA1/snKsysx2A1djXCcCiOrbzIrDIzE4FegJvHKYmkUahnrtEiyuASryx737+r57Ax3jj8OOBx82sk//E5hD/pZIvAeeb2TVm1tzMMsysn3+b84CrzCzZzE4EbqmnhlZABVAINDezh/HG1qs9C/yvmXU3T18zywBwzvnwxutfAKZVD/OIBIvCXaLFTcDzzrmvnXObq7+Av+GNqz8ALMQL0O3A74Bmzrmv8U5w3u9/fh5wqn+bfwbKgC14wyYv1VPD+3gnZ1cA6/D+Wqg5bPM48AowC9gFPAck1Xh9ItAHDclICJgW6xAJDTP7Nt7wTFfnXFW465HYpp67SAiYWTxwH/Csgl1CQeEuEmRm1hPYiXfi94kwlyNNhIZlRERikHruIiIxKGzXuWdmZrquXbuGa/ciIlEpPz9/m3OubX3twhbuXbt2JS8vL1y7FxGJSma2LpB2GpYREYlBCncRkRikcBcRiUERNXFYeXk5Pp+P0tLScJcSVImJiWRlZREfHx/uUkQkRkVUuPt8Plq1akXXrl0xs/p/IAo55ygqKsLn89GtW7dwlyMiMSqihmVKS0vJyMiI2WAHMDMyMjJi/q8TEQmviAp3IKaDvVpT+B1FJLwialhGRCTWVFY5Nu8qxbd9H74dJazfsY9zT25H36y0oO5X4V7Dzp07mTx5MnfeeWeDfu6iiy5i8uTJpKUF93+WiESeyirH1t2l+HaU4Nuxj/Xbvf96j0vYuLOEiqqDc3iZQUZKgsI9lHbu3Mnf//73b4R7ZWUlcXFxh/25mTNnBrs0EQmTqipH4Z79BwJ7/faDwe3bsY8NO0sorzx0Asa2rRLITk+iX3Yal/TtSHabZLLSk8hKT6ZTWiIJzQ+fJ41F4V7DAw88wOrVq+nXrx/x8fGkpKTQsWNH5s2bx5IlS7jiiitYv349paWl3HfffYwdOxY4OJXCnj17GD58OGeeeSaffvopnTt35s033yQpKamePYtIuDhXHd4ltcJ7Hxt2lODbWUJZxaFT8GemtCArPZnenVszrHdHstKTDgR457QkEuODH971idhwf+StxSzZuKtRt9mrUyq/vPSUw77+2GOPsWjRIubNm8dHH33ExRdfzKJFiw5csjh+/HjatGlDSUkJAwYMYMSIEWRkZByyjZUrVzJlyhSeeeYZrrnmGqZNm8aoUaPq2p2IhIBzjqK9ZYcZNvH+u79WeLdp2YLs9CR6dkxlaK/2Xq+7TTLZ6Ul0TksmqUX4w7s+ERvukWDgwIGHXIv+5JNP8vrrrwOwfv16Vq5c+Y1w79atG/36eesvn3766axduzZk9Yo0Rc45duwrr2PY5ODwSUl55SE/k5YcT3Z6Mie1b8W5J7cjKz2Z7DbesEnntCRaJkR/NEbsb3CkHnaotGzZ8sD3H330ER9++CGfffYZycnJnH322XVeq56QkHDg+7i4OEpKtMi9yLFwzlFcUn7YE5a+HfvYW3ZoeKcmNicrPZnj27bk2ye1PTDend3GGzZplRj7d4dHbLiHQ6tWrdi9e3edrxUXF5Oenk5ycjLLli3j888/D3F1IrHLC++6T1j6dpSwZ3/FIe1TEpqTlZ5El4xkvnViBlnp3nh3dnoyndOTaJ0U++FdH4V7DRkZGZxxxhn07t2bpKQk2rdvf+C1YcOG8fTTT9O3b1969OjB4MGDw1ipSHTZXVpe5wnL6uu+d5ceGt7JLeLI9gf24OMz/D1vf+87PZnUpOa6GbAeYVtDNScnx9VerGPp0qX07NkzLPWEWlP6XSX27d1fUWPYpEbPe6c3jFJcUn5I+6T4uAOBXfMywered1pyvML7MMws3zmXU1879dxFpMGccyzwFTOtwMe7izZTuHv/Ia8nNG92ILD7ZacdCG6v551Em5YtFN5BpnAXkYBtKi7h9S83ML1gA6u27qFF82YM7dmeUzqnHtLzzkxReIebwl1EjmhfWQXvL97MtPwNfLJ6G85BznHp/PaqPlzUp6NOXkYohbuIfENVleOLr7Z7wy4LN7G3rJKs9CTuObc7I/p35riMlvVvRMJK4S4iB3y1bS/TC3xML9jAhp0lpCQ05+K+HRnRP4sBXdvQrJmGWqKFwl2kiSveV87bCzcyLd9Hwdc7aWZwxomZ/GRYDy7o1SEqbrWXb1K413C0U/4CPPHEE4wdO5bk5OQgVCbSuCoqq5izspBp+Rv4YOkWyiqq6N4uhQeGn8wV/TrToXViuEuUY6Rwr+FwU/4G4oknnmDUqFEKd4loSzbuYlqBjzfnbWDbnjLSk+O5bmAXRvTPonfnVF3hEkMU7jXUnPJ36NChtGvXjldeeYX9+/dz5ZVX8sgjj7B3716uueYafD4flZWVPPTQQ2zZsoWNGzdyzjnnkJmZyezZs8P9q4gcsHV3KTPmbeS1fB/LNu8mPs449+R2jOifxdk92tGiecSttimNIHLD/d0HYPPCxt1mhz4w/LHDvlxzyt9Zs2bx2muvMXfuXJxzXHbZZcyZM4fCwkI6derEO++8A3hzzrRu3ZrHH3+c2bNnk5mZ2bg1ixyF0vJKPly6hWn5Puas3EZllePU7DQevfwULu3bifSWLcJdogRZ5IZ7mM2aNYtZs2Zx2mmnAbBnzx5WrlzJWWedxY9+9CN++tOfcskll3DWWWeFuVIRj3OOgq938Fr+Bt5esJHdpRV0SE1k7LePZ0T/zpzYrlW4S5QQitxwP0IPOxScczz44IPcfvvt33gtPz+fmTNn8uCDD3LBBRfw8MMPh6FCEc/67fv8d436WFu0j6T4OIb17sCI/lkMOSGDOF2+2CQFFO5mNgz4CxAHPOuce6zW68cB44G2wHZglHPO18i1Bl3NKX8vvPBCHnroIa6//npSUlLYsGED8fHxVFRU0KZNG0aNGkVKSgoTJkw45Gc1LCOhsGd/BTMXbmJavo8vvtoOwJDjM7jrnBMZ3qcjKTGw2IQcm3r/BZhZHPAUMBTwAblmNsM5t6RGsz8Ck5xzE83sXOC3wA3BKDiYak75O3z4cK677jqGDBkCQEpKCi+++CKrVq3ixz/+Mc2aNSM+Pp5//OMfAIwdO5bhw4fTsWNHnVCVoKiscny6ehvT8n28t3gzpeVVdMtsyf1DT+LK/p3JSteVWnJQvVP+mtkQ4FfOuQv9jx8EcM79tkabxcCFzjmfeddSFTvnUo+0XU3523R+Vzk2q7bu5rX8Dbzx5QY27yolNbE5l5zaiRH9s+jfJU2XLzYxjTnlb2dgfY3HPmBQrTbzgRF4QzdXAq3MLMM5VxRgvSJSw/a9Zbw1fyPTCnws8BUT18w4+6S2PHRJL87r2Y7EeN01KkcWSLjX1S2o3d3/EfA3MxsNzAE2ABW1f8jMxgJjAbp06dKgQkViXVlFFbOXb2Vavo/Zy7dSXuno1TGVhy7pxWWndqJtq4T6NyLiF0i4+4DsGo+zgI01GzjnNgJXAZhZCjDCOVdce0POuXHAOPCGZeramXMu5v/MDNfqVxJ5qhe9mF7gY8b8jezYV05mSgKjv9WVq/pn0bPjEUc3RQ4rkHDPBbqbWTe8HvlI4LqaDcwsE9junKsCHsS7cqbBEhMTKSoqIiMjI2YD3jlHUVERiYmau6Mp21xcyutfbmBage/AohcX9GrPiP5ZnNU9k+ZxumtUjk294e6cqzCzu4H38S6FHO+cW2xmjwJ5zrkZwNnAb83M4Q3L3HU0xWRlZeHz+SgsLDyaH48aiYmJZGVlhbsMCbF9ZRXMWryFaQU+/rtKi15IcEXUAtkisaZ60YvpBT5m1lj04qr+WVx1Wme6ZmrRC2kYLZAtEkZrqxe9+HIDvh0HF724qn8WA7XohYSAwl2kkRSXlPPOgk1MK/CRv27HgUUvfnyhFr2Q0FO4ixyDisoqPl65jdcKfHywRIteSORQuIschSUbdzG9wMcb8zaybc/+A4teXNW/M306t47Zq70keijcRQJUuHs/b87bwLSCDSzdtEuLXkhEU7iLHEFpeSX/WrqVaQU+/rOi0Fv0Iqu1Fr2QiKdwF6mletGLaQUbeHv+RnZp0QuJQgp3kRoW+oq5b+qXrNm2V4teSFRTuIv4LfDtZNSzX9AqMZ4/XN1Xi15IVNO/XBEOBntqUjxTxw7WwhcS9XR6X5q8+et3cv2zX9A6WcEusUM9d2nS5q/fyajnviAtOZ6pY4fQOS0p3CWJNAr13KXJmqdglximcJcmad76ndzw7BekJ7dQsEtM0rCMNDkHgr1lC6aOHUwnBbvEIIW7NClffr2DG5+bq2CXmKdhGWkyCvzB3iZFwS6xT+EuTULB1zu4yR/sU25TsEvs07CMxLzqHnuGv8fesbWCXWKfeu4S06qDPVPBLk2Mwl1iVv66g8E+RcEuTYyGZSQm5a/bzk3jc/099iFa7k6aHIW7xJzqYG/bKoEptw1WsEuTpHCXmJK/bjs3PjeXdqmJCvZgqaqCneugcBlsXQJFa6BZM4hPhuaJ3n/jkw79al79fTLE12hT/XzzRG8b0mgU7hIz8tZu56bxCvZG4xzs2gBblx78KlwKhcuhfN/BdikdwMx7rrwEKsuObn/NEw9+ABzyIZF4bB8ch3wlQ1x84xyfCKdwl5hQHeztUxOZMnYw7VMV7AFzDvZs8Ye3vze+dZn3/f5dB9uldIB2J8Ppo6HtydCuF7TtAYmph26vqtIL+fISqCg5+H15ifcBUFF68Pvy0oMfChUltZ6v0b50k/9xrfauquG/r8V984PjSB8GdX2o1PlBU6ttmP8aUbhL1Mtdu53RCvbA7C3yet+1e+MlOw62SWrjBXff73ph3q6XF+bJbQLbR7M4SEjxvoLJOe+vhGP94KhuU14CpcWwe/M321fuP7oaq/8aqf3XxRn3Qc9LG/d41N51ULcuEmS5/h57BwX7oUqLvd731iWH9sb3bj3YJqE1tOsJvS4/GODtekLLtt4wS6Qzg+YJ3ldSWnD3VVUZwAdHXX+h7PvmaxUl0Cz4Q0MBhbuZDQP+AsQBzzrnHqv1ehdgIpDmb/OAc25mI9cqcoiawT517GDaNcVg37/HGwOv3RvfvfFgm/iWXg/8pAugbc+DvfFWHaMjxCNBszho0dL7ihL1hruZxQFPAUMBH5BrZjOcc0tqNPsF8Ipz7h9m1guYCXQNQr0iAMz9ajujn59Lh9aJTL2tCQR7eQlsW/HN3vjOrw+2aZ4ImSdBt28fOpzSOltXojRBgfTcBwKrnHNrAMxsKnA5UDPcHVB9VqU1sBGRIKkO9o6tvatiYirYK8qgaFWNAPf3xHd8dfDkYbN4L8SzBkD/G/298Z6Q3tXrYYoQWLh3BtbXeOwDBtVq8ytglpndA7QEzq9rQ2Y2FhgL0KVLl4bWKsIXa4q4eUJu9Ad7ZYUX2NVj4dVhXrQKqiq8NhYHGSdA+1Ogz/8c7I23Ob7JXM4nRy+QcK9rUM7VenwtMME59yczGwK8YGa9nTv0OiXn3DhgHEBOTk7tbYgc0SHBPnYw7VpFQbBXVcHOtbWGU5Z6QywHrgc3r9fdrhecfPHB4ZTM7t7JQpGjEEi4+4DsGo+z+Oawyy3AMADn3GdmlghkAlsRaQRfrCli9PO5dE5PYvJtgyIv2J2DYt+hV6ZsXeKFeM0bflpne0MoJ5zrhXi7kyGzB7RIDl/tEpMCCfdcoLuZdQM2ACOB62q1+Ro4D5hgZj2BRKCwMQuVpuvzNUXcHCnBfuCGn1rDKVuXQdnug+1adfR636ffXOPkZg9IaBW+2qVJqTfcnXMVZnY38D7eZY7jnXOLzexRIM85NwO4H3jGzH6AN2Qz2jmnYRc5Zp+tLmLMBC/Yp9w2mLatQjhMsXfbN+/a3LoESncebJOc6fXE+13r/bf6UsOk9NDVKVIHC1cG5+TkuLy8vLDsW6JDyIN97zZYNB2WvwNbFsPeGn98JraucaOPfzilbU9IaRvcmkRqMbN851xOfe10h6pEpOpgz0pPYnIwg71sHyyfCQtehlX/AlfphfdJw/w98eobfjrohh+JKgp3iTifrS7i5glzyU5PDk6wV1bAV/+BBa/AsrehbA+kZsEZ90Kfa6B9r8bdn0gYKNwlony6ehtjJuSSnZ7MlLGDyUxppGB3DjbN8wJ90TTvpGhCa+g9wpsgq8sQ3cUpMUXhLhHj01XbGDMxly5tvB57owT7jrWw8FUv1LetgLgW0P0CL9C7X+DN0CcSgxTuEhGqg/24Ni156bZBxxbs+7bD4te9QF//uffccWfAkLu8GRB1JYs0AQp3CbtPVm3jlmMN9vISWPGeF+grP4Cqcu9k6Hm/hD5XQ5qmu5CmReEuYfXJKm+MvWtGSybfNoiMhgR7VSWs/a8X6EtneKsGteoIg7/nnRjt0EdXuEiTpXCXsKkO9m6ZLXnp1gYE++ZF3qWLC1/z5i1v0Qp6XQZ9r4GuZ2lmRBEU7hIm/13pDcUEHOzFvoMnRrcugWbN4cShcOH/QY/h3hJmInKAwl1C7uOVhdw6Ma/+YC/ZCUve9AJ93X+957IHwcV/gl5XQsuM0BUtEmUU7hJSNYN98m2DadOyxaENKvbDylnesMuK971pcTNOhHN+7p0YbXN8eAoXiTIKdwmZOSsKuW1SHcFeVQVffwYLX/EuYSwt9hZpzrnFG0fvdJpOjIo0kMJdQqI62I9vm8JLtw7ygn3rMv+J0VeheD3EJ0PPS71A73Y2xOmfp8jR0rtHgm7OikJunZTHCW1TmDzyONLn/9ML9c0LvaXkTjgXznsYelwECSnhLlckJijcJaj+s6KQ70+aw+2pC7g39Uvi/zEHcNCpPwz7HfS+ClLahbtMkZijcJfgqChj0Zzp7PloAp83zydhXxns6grf+Yl3g1HmieGuUCSmKdyl8TgHvlxY8DLl86fRu2wH2XGpuH6joP+1kDVAJ0ZFQkThLsdu20rvWvSFr8COtVTGJTCrvD+5qUO5b+z3SExtGe4KRZochbscnT1bvXnRF7wCGwsAg+O/w9KT7uC6T9rSqV07Xrp1EGnJLerdlIg0PoW7BK5sLyx7x7vSZfVsb0m6Dn3hgv+D3iOYvSmO2yfl0719ioJdJMwU7nJklRWw5iMv0Je9DeX7oHUXOOM+73r0dj0BmL18K7dPyuekDim8eIuCXSTcFO7yTc55Qy3VS9LtLYTENC/M+34XsgcfsiTd7GVbuf0FBbtIJFG4y0Hb18CCV70To0WrvCXpThrmX5JuKDT/5gRf/162he+9UKBgF4kwCvembm8RLJ7u9dJ9c73nup4F37rXvyRd2mF/tDrYe3RoxYu3DKJ1cnyIihaR+ijcm6KyfbDiXS/QV30IVRXQrhec/yvofTWkZde7iX8t3cIdLyrYRSKVwr2pqKqEr+b4l6R7C8p2Q6tOMPhOb9ilQ++AN/WvpVv43ov59OyYygtjFOwikUjhHut2b4FPn/SWpNuzGRJS4ZTLvSkAup7Z4CXpDgn2WwbROknBLhKJAgp3MxsG/AWIA551zj1W6/U/A+f4HyYD7Zxzhx+sldBYNA3euR/274buF0Lf//FOkB7lknQfLtnCHS8p2EWiQb3hbmZxwFPAUMAH5JrZDOfckuo2zrkf1Gh/D3BaEGqVQO0tgpn3ewtfdOoPVz4NbXsc0yY/WLKFO1/Kp1fHVCYp2EUiXiA994HAKufcGgAzmwpcDiw5TPtrgV82TnnSYMtmwlv3QckOOPchOOP7x7zohYJdJPoE8q7vDKyv8dgHDKqroZkdB3QD/n2Y18cCYwG6dOnSoEKlHqXF8N6DMO8laN8bbpgOHfoc82ZnLd7MXZML6NWpNZPGDFSwi0SJQMK9rjla3WHajgRec85V1vWic24cMA4gJyfncNuQhlr1L5hxD+zeDGf9CL7zU2h+7DcT1Qz2F24ZSGqigl0kWgQS7j6g5oXPWcDGw7QdCdx1rEVJgPbvgQ8egrzxkHkS3PIBZJ3eKJtWsItEt0DCPRfobmbdgA14AX5d7UZm1gNIBz5r1Aqlbms/gTfvhB3rYMjdcO4vjvoqmNreX7yZu14qoHfn1kxSsItEpXrD3TlXYWZ3A+/jXQo53jm32MweBfKcczP8Ta8FpjrnNNwSTOUl8O9fw2dPQfpxcPNMOO5bjbb56mDvk9WaiWMU7CLRKqDLKJxzM4GZtZ57uNbjXzVeWVInXz688T3YtgJyboGhj0JCSqNt/r1Fm7l7soJdJBboDtVoUFEG//kd/PfP0KoDjJoOJ57XqLt4b9Em7p78JX2yvKtiWinYRaKawj3SbV4Ir98BWxbCqdfBsN8ecabGo1Ed7H39PXYFu0j0U7hHqsoK+OQJ+OgxL8xHToaTL2703by7cBP3TFGwi8QahXskKlzhja1vyIdeV8DFj0PLjEbfTXWwn5qdxoSbByjYRWKIwj2SVFXBF/+Afz3qXdZ49XjoPSIou3p34SbunvIl/RTsIjFJ4R4ptn8Fb94F6z7xZm689C/eydMgmOnvsffLTmPimIGkJOifgUis0bs63JyD/Ofh/V+ANYPLn4J+14PVNevDsasO9tOy05igYBeJWXpnh1PxBphxN6z+N3T7jhfsASxxd7TeWbCJe6cq2EWaAr27w8E5mD8V3v0pVJXDRX/0bkpq1ixou3x7wUbumzqP/l3SeP5mBbtIrNM7PNT2bIW3vg/L34HswXDF3yHjhKDuUsEu0vToXR5Ki9+At38AZXvhgl97i1M3cA3Thnpv0eYDwT7h5oG0VLCLNAl6p4fCvu0w88ew6DXodBpc8TS0Oznou/18TRH3TvVuUFKwizQtercH2/L34K17YV8RnPNzOPMHEBf8a8qXbtrFbZPyyE5PYvxNAxTsIk2M3vHBUloM7/8MvnwR2p0C178KHU8Nya7Xb9/HTePn0rJFcybdMoj0lse+KpOIRBeFezCs+QjeuAt2b4QzfwhnPwDNE0Ky66I9+7lp/FxKyyt57Y5v0TmtcRbwEJHoonBvTGV74YNfQu4zkHEijJkF2QNCtvu9+ysYMyGXDTtLeOnWQZzUvlXI9i0ikUXh3li+/hxe/x7s+AoG3QHnPQwtkkO2+/LKKu54qYCFG4r55w055HRtE7J9i0jkUbgfq/JSmP1/8OlfvbtLb3obup0V0hKqqhw/eW0Bc1YU8rsRfRjaq31I9y8ikUfhfiw2FMAbd0DhMjh9tHftekLoh0J+++5SXv9yAz++sAffHdAl5PsXkcijcD8aFWXw8R9hzh8hpR1cPw26nx+WUsbNWc0zH3/FTUOO486zg3unq4hED4V7Q21Z7I2tb14Afb8Lw38HSelhKWV6gY/fzFzGxX078vClp2BBmklSRKKPwj1QVZXw6ZMw+zeQkArffRF6Xhq2cmYv38pPXlvAt07I4PFrTiWumYJdRA5SuAdi2ypv2TtfrhfolzwBLTPDVs6XX+/gzhcL6NGhFf+84XQSmgd3fhoRiT4K9yOpqoK54+DDX3k3IV31LPS5OmgLaQRideEexkzIpW2rBJ7X8ngichgK98PZsc5b9m7tx9D9Arj0SUjtGNaSNheXcuNzc4lrZkwaM5B2rRLDWo+IRC6Fe23OQcEkb14YDC77K5x2Q1h76wDFJeXcNH4uO/eV8fLtQ+ia2TKs9YhIZFO417RrI8y4F1Z9AF3P8hbSSAv/deOl5ZXcNjGPNdv28PzogfTu3DrcJYlIhFO4g9dbX/gqzPyRdw378D/AgFuDuuxdoCqrHPdO+ZLcddt5cuRpnNk9fCdyRSR6BJReZjbMzJab2Soze+Awba4xsyVmttjMJjdumUG0pxBeuQGm3waZPeCOT2DQ2IgIduccv3hjEbOWbOGXl/Ti0lM7hbskEYkS9fbczSwOeAoYCviAXDOb4ZxbUqNNd+BB4Azn3A4zaxesghvVkhnesnf7d8H5j8C37gn6sncN8ecPVzJl7tfcdc4JjD6jW7jLEZEoEsiwzEBglXNuDYCZTQUuB5bUaHMb8JRzbgeAc25rYxfaqEp2wMyfwMJXvAU0rngL2vcKd1WHeOHzdTz5r5Vck5PFjy7oEe5yRCTKBBLunYH1NR77gEG12pwEYGafAHHAr5xz79XekJmNBcYCdOkSphOVKz+AGffA3kI4+0E46/6QLHvXEDMXbuLhNxeSk65/AAAOI0lEQVRx3snt+M2VfTStgIg0WCDhXleyuDq20x04G8gCPjaz3s65nYf8kHPjgHEAOTk5tbcRXKW7YNbPvcsc2/aEa6d4i1VHmM9WF/H9qfPo3yWdv13Xn+Zx4R/7F5HoE0i4+4DsGo+zgI11tPncOVcOfGVmy/HCPrdRqjxWX83xlr0rXg9n3Adn/wziI+8GoMUbixk7KY/jMpJ57qYcklpEzvi/iESXQLqFuUB3M+tmZi2AkcCMWm3eAM4BMLNMvGGaNY1Z6FEp2wfv/hQmXgpxzWHM+zD00YgM9vXb9zH6+VxSEpszccxA0pK1qLWIHL16e+7OuQozuxt4H288fbxzbrGZPQrkOedm+F+7wMyWAJXAj51zRcEsvF7r53pT825fDQNvh/N/CS0i867ObXv2c8NzX1BWUcXk7w2hkxa1FpFjZM6Fdui7Wk5OjsvLy2v8DVfs96bl/fRJSO0Mlz8Fx3+n8ffTSPbsr+C6Zz5nxZbdvHTrYE4/Ljxzw4tIdDCzfOdcTn3tYusO1Y3zvGXvti7x5oO58DeQmBruqg6rrKKKO17MZ/HGXYy74XQFu4g0mtgI98py+PhxmPN7SM6E616Fky4Id1VHVFXl+NGr8/l45TZ+f3VfzuupRa1FpPFEf7hvXeqNrW+aB33+B4b/HpLbhLuqI3LO8et3ljJj/kZ+MqwH1+Rk1/9DIiINEL3hXlUJn/0N/v1rSGgF10yCXpeHu6qA/HPOGsZ/8hU3n9GVO76jRa1FpPFFZ7gXrfbG1td/ASdf4i17l9I23FUF5NW89Tz27jIuPbUTD13cS3efikhQRF+4z38Z3v4+NIuHK8dB32vCvpBGoP69bAsPTF/ImSdm8sf/6UszLWotIkESfeGe2hGOOwMu/Qu07hzuagJW8PUO7nypgF4dU3lai1qLSJBFX7h3+7b3FUVWbd3NmAm5dEhN5PmbB5CSEH2HXUSii2alCrJNxSXc+NxcmjdrxqQxg8hMSQh3SSLSBCjcg6h4n7eo9a7SCibcPIAuGcnhLklEmgiFe5CUlldyy8Rc1m7bx7gbTtei1iISUhr8DYKKyirunvwl+V/v4G/X9udbJ2pRaxEJLfXcG5lzjp+/vogPl27hkctO4eK+HcNdkog0QQr3Rvb4Byt4OW8995x7IjcO6RruckSkiVK4N6KJn67lr/9excgB2fxw6EnhLkdEmjCFeyN5e8FGfvXWYs7v2Z5fX9Fb0wqISFgp3BvBp6u28cOX55NzXDp/u+40LWotImGnFDpGizYUM/aFfLpmJvPsjQNIjNe0AiISfgr3Y7CuaC+jn8+ldVI8k8YMonVyfLhLEhEBdJ37USvcvZ8bx8+loqqKqWMG06F1YrhLEhE5QD33o7BnfwU3T5jLll2ljB89gBPbpYS7JBGRQ6jn3kD7Kyq5/YU8lm7azbM35tC/ixa1FpHIo557A1RVOe5/ZT6frCri9yP6cs7J7cJdkohInRTuAXLO8ejbS3h7wSYeGH4yI07PCndJIiKHpXAP0N8/Ws2ET9dyy5nduP3bx4e7HBGRI1K4B+CVvPX84f3lXN6vEz+/qKfuPhWRiKdwr8eHS7bw4PSFnNU9kz9cfaoWtRaRqBBQuJvZMDNbbmarzOyBOl4fbWaFZjbP/3Vr45caevnrtnPX5AJO6ZTK06NOp0VzfRaKSHSo91JIM4sDngKGAj4g18xmOOeW1Gr6snPu7iDUGBYrtuxmzIQ8OqUlMX70AFpqUWsRiSKBdEUHAqucc2ucc2XAVODy4JYVXht3lnDT+Lm0aN6MSWMGalFrEYk6gYR7Z2B9jcc+/3O1jTCzBWb2mpll17UhMxtrZnlmlldYWHgU5Qbfzn1l3Dh+LntKK5h480Cy22hRaxGJPoGEe11nEF2tx28BXZ1zfYEPgYl1bcg5N845l+Ocy2nbtm3DKg2BkrJKxkzI5euifYy7MYdenVLDXZKIyFEJJNx9QM2eeBawsWYD51yRc26//+EzwOmNU17oeItaF/Dl+p38ZWQ/hpyQEe6SRESOWiDhngt0N7NuZtYCGAnMqNnAzGquAn0ZsLTxSgw+5xwPTl/Iv5Zt5X8v783wPlrUWkSiW72XgDjnKszsbuB9IA4Y75xbbGaPAnnOuRnAvWZ2GVABbAdGB7HmRveH95fzar6Pe8/rzqjBx4W7HBGRY2bO1R4+D42cnByXl5cXln3X9PwnX/HIW0u4dmAXfnOl1j4VkchmZvnOuZz62jXpu3JmzN/Io28v4cJTtKi1iMSWJhvu/125jftfmceArm34y8jTiNO0AiISQ5pkuC/0FXP7C3mc0DaFZ27M0aLWIhJzmly4r922l9HPzyUtuQUTxwykdZIWtRaR2NOkwn3r7lJuHD+XKueYdMtA2qdqUWsRiU1NZjas3aXljB6fS+Hu/Uy+bRAntNWi1iISu5pEz31/RSVjJ+WzYstu/jGqP6dpUWsRiXEx33OvrHL88OX5fLamiMevOZWze2hRaxGJfTHdc3fO8chbi3ln4SZ+flFPruqvRa1FpGmI6XB/avYqJn22jtvO6sZtWtRaRJqQmA33qXO/5o+zVnDlaZ15cHjPcJcjIhJSMRnusxZv5mevL+Q7J7Xl91f31aLWItLkxFy4567dzj1TvqRPVhp/v74/8XEx9yuKiNQrppJv+ebd3DIhl85pSTyvRa1FpAmLmXD37djHjeO/IDE+joljBtKmZYtwlyQiEjYx0bXdsddb1HpfWSWv3D5Ei1qLSJMX9eG+r6yCmyfk4ttRwgtjBtKzoxa1FhGJ6mGZ8soq7nqpgAW+nTw58jQGHa9FrUVEIIp77s45fjptAbOXF/J/V/ZmWO8O4S5JRCRiRG3P/bH3ljG9YAM/OP8krh+kRa1FRGqKynB/9uM1/PM/a7h+UBfuPe/EcJcjIhJxoi7c35y3gV+/s5Rhp3Tg0cu1qLWISF2iLtzbpyYytFd7nhjZT4tai4gcRtSdUB18fAaDdVWMiMgRRV3PXURE6qdwFxGJQQp3EZEYpHAXEYlBAYW7mQ0zs+VmtsrMHjhCu6vNzJlZTuOVKCIiDVVvuJtZHPAUMBzoBVxrZr3qaNcKuBf4orGLFBGRhgmk5z4QWOWcW+OcKwOmApfX0e5/gd8DpY1Yn4iIHIVAwr0zsL7GY5//uQPM7DQg2zn39pE2ZGZjzSzPzPIKCwsbXKyIiAQmkJuY6roN1B140awZ8GdgdH0bcs6NA8b5f67QzNYFVuY3ZALbjvJng0l1NYzqarhIrU11Ncyx1BXQTImBhLsPyK7xOAvYWONxK6A38JF/npcOwAwzu8w5l3e4jTrn2gZSYF3MLM85F3EnbVVXw6iuhovU2lRXw4SirkCGZXKB7mbWzcxaACOBGdUvOueKnXOZzrmuzrmuwOfAEYNdRESCq95wd85VAHcD7wNLgVecc4vN7FEzuyzYBYqISMMFNHGYc24mMLPWcw8fpu3Zx15WvcaFYB9HQ3U1jOpquEitTXU1TNDrMudc/a1ERCSqaPoBEZEYpHAXEYlBER3u9c1pY2YJZvay//UvzKxrhNQ12n8d/zz/160hqmu8mW01s0WHed3M7El/3QvMrH+E1HW2mRXXOF51ns9p5JqyzWy2mS01s8Vmdl8dbUJ+vAKsKxzHK9HM5prZfH9dj9TRJuTvxwDrCsv70b/vODP70sy+cYNn0I+Xcy4iv4A4YDVwPNACmA/0qtXmTuBp//cjgZcjpK7RwN/CcMy+DfQHFh3m9YuAd/FuTBsMfBEhdZ0NvB3iY9UR6O//vhWwoo7/jyE/XgHWFY7jZUCK//t4vDmkBtdqE473YyB1heX96N/3D4HJdf3/CvbxiuSeeyBz2lwOTPR//xpwngV/xexA59oJOefcHGD7EZpcDkxyns+BNDPrGAF1hZxzbpNzrsD//W68y3w712oW8uMVYF0h5z8Ge/wP4/1fta/GCPn7McC6wsLMsoCLgWcP0ySoxyuSw73eOW1qtnHe9fjFQLAXWA2kLoAR/j/lXzOz7DpeD4dAaw+HIf4/rd81s1NCuWP/n8On8c0ZTcN6vI5QF4ThePmHGOYBW4EPnHOHPV4hfD8GUheE5/34BPAToOowrwf1eEVyuB9xTpsGtGlsgezzLaCrc64v8CEHP53DLRzHKxAFwHHOuVOBvwJvhGrHZpYCTAO+75zbVfvlOn4kJMernrrCcrycc5XOuX54U5AMNLPetZqE5XgFUFfI349mdgmw1TmXf6RmdTzXaMcrksO9vjltDmljZs2B1gT/z/9663LOFTnn9vsfPgOcHuSaAhXIMQ0559yu6j+tnXfDXLyZZQZ7v2YWjxegLznnptfRJCzHq766wnW8aux/J/ARMKzWS+F4P9ZbV5jej2cAl5nZWryh23PN7MVabYJ6vCI53I84p43fDOAm//dXA/92/rMT4ayr1rjsZXjjppFgBnCj/yqQwUCxc25TuIsysw7VY41mNhDv32VRkPdpwHPAUufc44dpFvLjFUhdYTpebc0szf99EnA+sKxWs5C/HwOpKxzvR+fcg865LOfNtzUS71iMqtUsqMcroOkHwsE5V2Fm1XPaxAHjnX9OGyDPOTcD703wgpmtwvvEGxkhdd1r3rw7Ff66Rge7LgAzm4J3JUWmmfmAX+KdYMI59zTeFBIXAauAfcDNEVLX1cAdZlYBlAAjQ/AhfQZwA7DQP14L8DOgS426wnG8AqkrHMerIzDRvJXZmuHNMfV2uN+PAdYVlvdjXUJ5vDT9gIhIDIrkYRkRETlKCncRkRikcBcRiUEKdxGRGKRwFxGJQQp3EZEYpHAXEYlB/w++6+75GhdGrgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title('Accuracy')\n",
    "plt.plot(history.history['acc'], label='train')\n",
    "plt.plot(history.history['val_acc'], label='test')\n",
    "plt.legend()\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CNN + LSTM Model - 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_8 (InputLayer)            (None, 100)          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "embedding_16 (Embedding)        (None, 100, 100)     5000000     input_8[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv1d_20 (Conv1D)              (None, 98, 200)      60200       embedding_16[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "conv1d_21 (Conv1D)              (None, 97, 200)      80200       embedding_16[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "conv1d_22 (Conv1D)              (None, 96, 200)      100200      embedding_16[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling1d_20 (MaxPooling1D) (None, 1, 200)       0           conv1d_20[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling1d_21 (MaxPooling1D) (None, 1, 200)       0           conv1d_21[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling1d_22 (MaxPooling1D) (None, 1, 200)       0           conv1d_22[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_7 (Concatenate)     (None, 1, 600)       0           max_pooling1d_20[0][0]           \n",
      "                                                                 max_pooling1d_21[0][0]           \n",
      "                                                                 max_pooling1d_22[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "dense_27 (Dense)                (None, 1, 30)        18030       concatenate_7[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "dropout_13 (Dropout)            (None, 1, 30)        0           dense_27[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "bidirectional_7 (Bidirectional) (None, 1, 200)       104800      dropout_13[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_28 (Dense)                (None, 1, 30)        6030        bidirectional_7[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "dropout_14 (Dropout)            (None, 1, 30)        0           dense_28[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "flatten_7 (Flatten)             (None, 30)           0           dropout_14[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_29 (Dense)                (None, 3)            93          flatten_7[0][0]                  \n",
      "==================================================================================================\n",
      "Total params: 5,369,553\n",
      "Trainable params: 5,369,553\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n",
      "None\n",
      "Train on 2281 samples, validate on 254 samples\n",
      "Epoch 1/3\n",
      "2281/2281 [==============================] - 18s 8ms/step - loss: 0.6605 - acc: 0.6405 - val_loss: 0.6255 - val_acc: 0.6667\n",
      "Epoch 2/3\n",
      "2281/2281 [==============================] - 11s 5ms/step - loss: 0.6352 - acc: 0.6605 - val_loss: 0.6108 - val_acc: 0.6667\n",
      "Epoch 3/3\n",
      "2281/2281 [==============================] - 11s 5ms/step - loss: 0.5881 - acc: 0.7036 - val_loss: 0.4908 - val_acc: 0.7861\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import Embedding\n",
    "from keras.models import Sequential, Model\n",
    "from keras.layers import Dense, Activation\n",
    "from keras.layers import Flatten, Conv1D, SpatialDropout1D, MaxPooling1D,AveragePooling1D, Bidirectional, merge, concatenate, Input, Dropout, LSTM\n",
    "\n",
    "y_dim=3\n",
    "num_filters=200\n",
    "filter_sizes=[3,4,5] \n",
    "pool_padding='valid' \n",
    "dropout=0.5\n",
    "\n",
    "embed_input = Input(shape=(X.shape[1],))\n",
    "x = Embedding(MAX_NB_WORDS, EMBEDDING_DIM, input_length=X.shape[1])(embed_input)\n",
    "pooled_outputs = []\n",
    "for i in range(len(filter_sizes)):\n",
    "    conv = Conv1D(num_filters, kernel_size=filter_sizes[i], padding='valid', activation='relu')(x)\n",
    "    conv = MaxPooling1D(pool_size=EMBEDDING_DIM-filter_sizes[i]+1)(conv)           \n",
    "    pooled_outputs.append(conv)\n",
    "merge = concatenate(pooled_outputs)\n",
    "    \n",
    "x = Dense(30, activation='relu')(merge)\n",
    "x = Dropout(dropout)(x)\n",
    "x = Bidirectional(LSTM(100, return_sequences=True, dropout=0.5, recurrent_dropout=0.1))(x)\n",
    "x = Dense(30, activation='relu')(x)\n",
    "x = Dropout(dropout)(x)\n",
    "x = Flatten()(x)\n",
    "x = Dense(y_dim, activation='sigmoid')(x)\n",
    "\n",
    "model = Model(inputs=embed_input,outputs=x)\n",
    "\n",
    "#   model.compile(optimizer='adam',loss = 'categorical_crossentropy', metrics = ['acc'])\n",
    "model.compile(loss='binary_crossentropy', \n",
    "             optimizer='adam', \n",
    "             metrics=['accuracy'])\n",
    "\n",
    "print(model.summary())\n",
    "\n",
    "epochs = 3\n",
    "batch_size = 64\n",
    "\n",
    "history = model.fit(X_train, Y_train, epochs=epochs, batch_size=batch_size,validation_split=0.1,callbacks=[EarlyStopping(monitor='val_loss', patience=3, min_delta=0.0001)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "282/282 [==============================] - 0s 1ms/step\n",
      "Test set\n",
      "  Loss: 0.503\n",
      "  Accuracy: 0.757\n"
     ]
    }
   ],
   "source": [
    "accr = model.evaluate(X_test,Y_test)\n",
    "print('Test set\\n  Loss: {:0.3f}\\n  Accuracy: {:0.3f}'.format(accr[0],accr[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "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": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.title('Loss')\n",
    "plt.plot(history.history['loss'], label='train')\n",
    "plt.plot(history.history['val_loss'], label='test')\n",
    "plt.legend()\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "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.title('Accuracy')\n",
    "plt.plot(history.history['acc'], label='train')\n",
    "plt.plot(history.history['val_acc'], label='test')\n",
    "plt.legend()\n",
    "plt.show();"
   ]
  }
 ],
 "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}