LSTM_Recurrent_2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "np.random.seed(42)\n",
    "import tensorflow as tf\n",
    "tf.set_random_seed(42)\n",
    "from keras.models import Sequential, load_model\n",
    "from keras.layers import Dense, Activation\n",
    "from keras.layers import LSTM, Dropout\n",
    "from keras.layers import TimeDistributed\n",
    "from keras.layers.core import Dense, Activation, Dropout, RepeatVector\n",
    "from keras.optimizers import RMSprop\n",
    "import matplotlib.pyplot as plt\n",
    "import pickle\n",
    "import sys\n",
    "import heapq\n",
    "import seaborn as sns\n",
    "from pylab import rcParams\n",
    "\n",
    "%matplotlib inline\n",
    "sns.set(style='whitegrid', palette='muted', font_scale=1.5)\n",
    "\n",
    "rcParams['figure.figsize'] = 12, 5"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Loading the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "corpus length: 11295\n"
     ]
    }
   ],
   "source": [
    "path = 'wonderland.txt'\n",
    "text = open(path, encoding=\"utf8\").read().lower()\n",
    "print('corpus length:', len(text))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "unique chars: 46\n"
     ]
    }
   ],
   "source": [
    "chars = sorted(list(set(text)))\n",
    "char_indices = dict((c, i) for i, c in enumerate(chars))\n",
    "indices_char = dict((i, c) for i, c in enumerate(chars))\n",
    "\n",
    "print(f'unique chars: {len(chars)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "num training examples: 5628\n"
     ]
    }
   ],
   "source": [
    "SEQUENCE_LENGTH = 40\n",
    "step = 2\n",
    "sentences = []\n",
    "next_chars = []\n",
    "for i in range(0, len(text) - SEQUENCE_LENGTH, step):\n",
    "    sentences.append(text[i: i + SEQUENCE_LENGTH])\n",
    "    next_chars.append(text[i + SEQUENCE_LENGTH])\n",
    "print(f'num training examples: {len(sentences)}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# generating our features and labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X = np.zeros((len(sentences), SEQUENCE_LENGTH, len(chars)), dtype=np.bool)\n",
    "y = np.zeros((len(sentences), len(chars)), dtype=np.bool)\n",
    "for i, sentence in enumerate(sentences):\n",
    "    for t, char in enumerate(sentence):\n",
    "        X[i, t, char_indices[char]] = 1\n",
    "    y[i, char_indices[next_chars[i]]] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'s saved one and two at a time by bulldoz'"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sentences[100]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# The character that needs to be predicted for it is:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'i'"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "next_chars[100]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[False False False False False False False False False False False False\n",
      " False False False False False False False False False False False False\n",
      " False False False False False False False False False False False False\n",
      " False False False False False False False False False  True]\n",
      "(5628, 40, 46)\n",
      "False\n",
      "(5628, 46)\n"
     ]
    }
   ],
   "source": [
    "print (X[0][0])\n",
    "print (X.shape)\n",
    "print (y[0][0])\n",
    "print (y.shape)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Building the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(LSTM(128, input_shape=(SEQUENCE_LENGTH, len(chars))))\n",
    "model.add(Dense(len(chars)))\n",
    "model.add(Activation('softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 5346 samples, validate on 282 samples\n",
      "Epoch 1/50\n",
      "5346/5346 [==============================] - 21s - loss: 3.1240 - acc: 0.1706 - val_loss: 2.8441 - val_acc: 0.1525\n",
      "Epoch 2/50\n",
      "5346/5346 [==============================] - 15s - loss: 2.7292 - acc: 0.2400 - val_loss: 2.5710 - val_acc: 0.2695\n",
      "Epoch 3/50\n",
      "5346/5346 [==============================] - 16s - loss: 2.4716 - acc: 0.3034 - val_loss: 2.4379 - val_acc: 0.2730\n",
      "Epoch 4/50\n",
      "5346/5346 [==============================] - 15s - loss: 2.3214 - acc: 0.3350 - val_loss: 2.3812 - val_acc: 0.2624\n",
      "Epoch 5/50\n",
      "5346/5346 [==============================] - 14s - loss: 2.2041 - acc: 0.3605 - val_loss: 2.2862 - val_acc: 0.2943\n",
      "Epoch 6/50\n",
      "5346/5346 [==============================] - 14s - loss: 2.0842 - acc: 0.3915 - val_loss: 2.2466 - val_acc: 0.3014\n",
      "Epoch 7/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.9800 - acc: 0.4160 - val_loss: 2.2543 - val_acc: 0.3085\n",
      "Epoch 8/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.8706 - acc: 0.4390 - val_loss: 2.3455 - val_acc: 0.3191\n",
      "Epoch 9/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.7681 - acc: 0.4673 - val_loss: 2.2033 - val_acc: 0.3475\n",
      "Epoch 10/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.6559 - acc: 0.5013 - val_loss: 2.3012 - val_acc: 0.3369\n",
      "Epoch 11/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.5510 - acc: 0.5264 - val_loss: 2.3095 - val_acc: 0.3085\n",
      "Epoch 12/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.4341 - acc: 0.5643 - val_loss: 2.3183 - val_acc: 0.3227\n",
      "Epoch 13/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.3292 - acc: 0.5935 - val_loss: 2.4688 - val_acc: 0.3050\n",
      "Epoch 14/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.2238 - acc: 0.6240 - val_loss: 2.5099 - val_acc: 0.3262\n",
      "Epoch 15/50\n",
      "5346/5346 [==============================] - 15s - loss: 1.1207 - acc: 0.6543 - val_loss: 2.5377 - val_acc: 0.3156\n",
      "Epoch 16/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.0057 - acc: 0.7007 - val_loss: 2.6523 - val_acc: 0.3014\n",
      "Epoch 17/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.9163 - acc: 0.7245 - val_loss: 2.6876 - val_acc: 0.3227\n",
      "Epoch 18/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.8374 - acc: 0.7445 - val_loss: 2.8032 - val_acc: 0.3050\n",
      "Epoch 19/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.7616 - acc: 0.7671 - val_loss: 2.8102 - val_acc: 0.2943\n",
      "Epoch 20/50\n",
      "5346/5346 [==============================] - 15s - loss: 0.6817 - acc: 0.7926 - val_loss: 2.9796 - val_acc: 0.2801\n",
      "Epoch 21/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.6773 - acc: 0.8111 - val_loss: 3.1272 - val_acc: 0.3014\n",
      "Epoch 22/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.1502 - acc: 0.8094 - val_loss: 3.6756 - val_acc: 0.2730\n",
      "Epoch 23/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.3411 - acc: 0.8058 - val_loss: 3.7196 - val_acc: 0.2730\n",
      "Epoch 24/50\n",
      "5346/5346 [==============================] - 14s - loss: 4.8525 - acc: 0.6203 - val_loss: 8.1803 - val_acc: 0.1596\n",
      "Epoch 25/50\n",
      "5346/5346 [==============================] - 14s - loss: 7.4809 - acc: 0.4802 - val_loss: 8.4341 - val_acc: 0.1773\n",
      "Epoch 26/50\n",
      "5346/5346 [==============================] - 14s - loss: 7.3422 - acc: 0.4903 - val_loss: 8.7163 - val_acc: 0.1348\n",
      "Epoch 27/50\n",
      "5346/5346 [==============================] - 14s - loss: 7.0912 - acc: 0.5058 - val_loss: 4.4357 - val_acc: 0.2589\n",
      "Epoch 28/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.5458 - acc: 0.7849 - val_loss: 3.9708 - val_acc: 0.2624\n",
      "Epoch 29/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.0868 - acc: 0.8412 - val_loss: 4.0065 - val_acc: 0.2589\n",
      "Epoch 30/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.0265 - acc: 0.8649 - val_loss: 4.1499 - val_acc: 0.2979\n",
      "Epoch 31/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.6222 - acc: 0.8346 - val_loss: 4.9242 - val_acc: 0.2695\n",
      "Epoch 32/50\n",
      "5346/5346 [==============================] - 14s - loss: 1.4988 - acc: 0.8389 - val_loss: 4.3190 - val_acc: 0.2589\n",
      "Epoch 33/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.9832 - acc: 0.8906 - val_loss: 4.7120 - val_acc: 0.2411\n",
      "Epoch 34/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.9902 - acc: 0.8895 - val_loss: 4.2781 - val_acc: 0.2624\n",
      "Epoch 35/50\n",
      "5346/5346 [==============================] - 15s - loss: 0.9106 - acc: 0.9025 - val_loss: 4.4011 - val_acc: 0.2801\n",
      "Epoch 36/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.9134 - acc: 0.8917 - val_loss: 4.4607 - val_acc: 0.2837\n",
      "Epoch 37/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.8609 - acc: 0.8984 - val_loss: 4.5147 - val_acc: 0.2482\n",
      "Epoch 38/50\n",
      "5346/5346 [==============================] - 15s - loss: 0.8521 - acc: 0.9009 - val_loss: 4.3979 - val_acc: 0.2872\n",
      "Epoch 39/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.8208 - acc: 0.9128 - val_loss: 4.6049 - val_acc: 0.2624\n",
      "Epoch 40/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.8289 - acc: 0.9057 - val_loss: 4.7550 - val_acc: 0.2766\n",
      "Epoch 41/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.8064 - acc: 0.9186 - val_loss: 4.6685 - val_acc: 0.2837\n",
      "Epoch 42/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.7927 - acc: 0.9095 - val_loss: 4.7228 - val_acc: 0.2837\n",
      "Epoch 43/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.7941 - acc: 0.9166 - val_loss: 4.8641 - val_acc: 0.2730\n",
      "Epoch 44/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.7825 - acc: 0.9216 - val_loss: 4.8420 - val_acc: 0.2801\n",
      "Epoch 45/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.4490 - acc: 0.9233 - val_loss: 4.4922 - val_acc: 0.3085\n",
      "Epoch 46/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.2025 - acc: 0.9454 - val_loss: 4.2977 - val_acc: 0.3014\n",
      "Epoch 47/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.1879 - acc: 0.9493 - val_loss: 4.2756 - val_acc: 0.2801\n",
      "Epoch 48/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.1931 - acc: 0.9454 - val_loss: 4.3088 - val_acc: 0.2872\n",
      "Epoch 49/50\n",
      "5346/5346 [==============================] - 14s - loss: 0.3293 - acc: 0.9435 - val_loss: 4.8617 - val_acc: 0.2943\n",
      "Epoch 50/50\n",
      "5346/5346 [==============================] - 15s - loss: 0.7704 - acc: 0.9231 - val_loss: 4.7582 - val_acc: 0.2979\n"
     ]
    }
   ],
   "source": [
    "optimizer = RMSprop(lr=0.01)\n",
    "model.compile(loss='categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])\n",
    "history = model.fit(X, y, validation_split=0.05, batch_size=128, epochs=50, shuffle=True).history"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Saving Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model.save('keras_model.h5')\n",
    "pickle.dump(history, open(\"history.p\", \"wb\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model = load_model('keras_model.h5')\n",
    "history = pickle.load(open(\"history.p\", \"rb\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluatplt.plot(history['acc'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'plt' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-1-3918201b9277>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mplot\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mhistory\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'val_acc'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      2\u001b[0m \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mtitle\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'model accuracy'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mylabel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'accuracy'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mxlabel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'epoch'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[0mplt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlegend\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'train'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'test'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mloc\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'upper left'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m;\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mNameError\u001b[0m: name 'plt' is not defined"
     ]
    }
   ],
   "source": [
    "plt.plot(history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAs4AAAFaCAYAAAAO89vkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x22ff896c240>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history['loss'])\n",
    "plt.plot(history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def prepare_input(text):\n",
    "    x = np.zeros((1, SEQUENCE_LENGTH, len(chars)))\n",
    "    for t, char in enumerate(text):\n",
    "        x[0, t, char_indices[char]] = 1.\n",
    "        \n",
    "    return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[ 0.,  0.,  0., ...,  0.,  0.,  0.],\n",
       "        [ 0.,  0.,  0., ...,  0.,  0.,  0.],\n",
       "        [ 0.,  0.,  0., ...,  0.,  0.,  0.],\n",
       "        ..., \n",
       "        [ 0.,  0.,  0., ...,  0.,  0.,  0.],\n",
       "        [ 0.,  0.,  0., ...,  0.,  0.,  0.],\n",
       "        [ 0.,  0.,  0., ...,  0.,  0.,  0.]]])"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prepare_input(\"This is an example of input for our LSTM\".lower())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def sample(preds, top_n=3):\n",
    "    preds = np.asarray(preds).astype('float64')\n",
    "    preds = np.log(preds)\n",
    "    exp_preds = np.exp(preds)\n",
    "    preds = exp_preds / np.sum(exp_preds)\n",
    "    \n",
    "    return heapq.nlargest(top_n, range(len(preds)), preds.take)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def predict_completion(text):\n",
    "    original_text = text\n",
    "    generated = text\n",
    "    completion = ''\n",
    "    while True:\n",
    "        x = prepare_input(text)\n",
    "        preds = model.predict(x, verbose=0)[0]\n",
    "        next_index = sample(preds, top_n=1)[0]\n",
    "        next_char = indices_char[next_index]\n",
    "        text = text[1:] + next_char\n",
    "        completion += next_char\n",
    "        \n",
    "        if len(original_text + completion) + 2 > len(original_text) and next_char == ' ':\n",
    "            return completion"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# predict multiple completions:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def predict_completions(text, n=3):\n",
    "    x = prepare_input(text)\n",
    "    preds = model.predict(x, verbose=0)[0]\n",
    "    next_indices = sample(preds, n)\n",
    "    return [indices_char[idx] + predict_completion(text[1:] + indices_char[idx]) for idx in next_indices]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "quotes = [\n",
    "    \"So now Della's beautiful hair fell about her rippling and shining\",\n",
    "    \"When Della reached home her intoxication gave way a little to.\",\n",
    "    \"White fingers and nimble tore at the string and paper.\",\n",
    "    \"Within forty minutes her head was covered with tiny\",\n",
    "    \"She had a habit of saying a little silent prayer\"\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "so now della's beautiful hair fell about\n",
      "[' the ', 'ou ', '. ', 's ', '\\nthe ']\n",
      "\n",
      "when della reached home her intoxication\n",
      "['. ', '-was ', ' geve ', ', ', '\\nwas ']\n",
      "\n",
      "white fingers and nimble tore at the str\n",
      "['entir ', 'ops ', 'ing ', 'ape ', 'sen ']\n",
      "\n",
      "within forty minutes her head was covere\n",
      "['d ', ' the ', 'r ', 'ntired ', '. ']\n",
      "\n",
      "she had a habit of saying a little silen\n",
      "['glap ', \"'m \", 't ', 'ed ', ', ']\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for q in quotes:\n",
    "    seq = q[:40].lower()\n",
    "    print(seq)\n",
    "    print(predict_completions(seq, 5))\n",
    "    print()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}