05_RNN_word_embeddings.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![word embeddings vs. one hot encoding](https://s3.amazonaws.com/book.keras.io/img/ch6/word_embeddings.png)\n",
    "\n",
    "word embeddings과 one hot encoding의 차이점을 보여주는 그림이다. word embeddings은 좀 더 데이터가 밀집하게 구성돼있다."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ont-hat encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "samples = ['This is test of ont hat encoding', 'ont hat encoding is very good']\n",
    "\n",
    "token_index = {}\n",
    "for sample in samples:\n",
    "    for word in sample.split():\n",
    "        if word not in token_index:\n",
    "            token_index[word] = len(token_index) + 1\n",
    "            \n",
    "max_length = 10\n",
    "\n",
    "results = np.zeros(shape = (len(samples),\n",
    "                            max_length,\n",
    "                            max(token_index.values()) + 1))\n",
    "for i, sample in enumerate(samples):\n",
    "    for j, word in list(enumerate(sample.split()))[:max_length]:\n",
    "        index = token_index.get(word)\n",
    "        results[i, j, index] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'This': 1,\n",
       " 'encoding': 7,\n",
       " 'good': 9,\n",
       " 'hat': 6,\n",
       " 'is': 2,\n",
       " 'of': 4,\n",
       " 'ont': 5,\n",
       " 'test': 3,\n",
       " 'very': 8}"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "token_index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import string\n",
    "\n",
    "samples = ['This is test of ont hat encoding', 'ont hat encoding is very good']\n",
    "characters = string.printable\n",
    "token_index = dict(zip(characters, range(1, len(characters) + 1)))\n",
    "\n",
    "max_length = 50\n",
    "results = np.zeros((len(samples), max_length, max(token_index.values()) + 1))\n",
    "for i, sample in enumerate(samples):\n",
    "    for j, character in enumerate(sample):\n",
    "        index = token_index.get(characters)\n",
    "        results[i,j, index]=1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.preprocessing.text import Tokenizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "find to num 9 token\n"
     ]
    }
   ],
   "source": [
    "samples = ['This is test of ont hat encoding', 'ont hat encoding is very good']\n",
    "\n",
    "tokenizer = Tokenizer(num_words=1000)\n",
    "tokenizer.fit_on_texts(samples)\n",
    "\n",
    "sequences = tokenizer.texts_to_sequences(samples)\n",
    "\n",
    "one_hot_results = tokenizer.texts_to_matrix(samples, mode = 'binary')\n",
    "\n",
    "word_index = tokenizer.word_index\n",
    "print(\"find to num %s token\" % len(word_index))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'encoding': 4,\n",
       " 'good': 9,\n",
       " 'hat': 3,\n",
       " 'is': 1,\n",
       " 'of': 7,\n",
       " 'ont': 2,\n",
       " 'test': 6,\n",
       " 'this': 5,\n",
       " 'very': 8}"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "word_index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0., 1., 1., ..., 0., 0., 0.],\n",
       "       [0., 1., 1., ..., 0., 0., 0.]])"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "one_hot_results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "samples = ['This is test of ont hat encoding', 'ont hat encoding is very good']\n",
    "\n",
    "dimensionality = 1000\n",
    "max_length = 10\n",
    "\n",
    "results = np.zeros((len(samples), max_length, dimensionality))\n",
    "for i, sample in enumerate(samples):\n",
    "    for j, word in list(enumerate(sample.split()))[:max_length]:\n",
    "        index = abs(hash(word)) % dimensionality\n",
    "        results[i,j,index]=1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "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.]],\n",
       "\n",
       "       [[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": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## word embedding"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### word embeddings 제작 방법\n",
    "* (문서 분류나 감성 예측과 같은) 관심 대상인 문제와 함께 단어 임베딩을 학습한다. 이런 경우에는 랜덤한 단어 벡터로 시작해서 신경망의 가중치를 학습하는 것과 같은 방식으로 단어 벡터를 학습힌다.\n",
    "* 풀려는 문제가 아니고 다른 머신 러닝 작업에서 미리 계산된 단어 임베딩을 로드한다. 이를 사전 훈련된 단어 임베딩이라고 한다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import Embedding\n",
    "\n",
    "embedding_layer = Embedding(1000, 64)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### imdb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.datasets import imdb\n",
    "from keras import preprocessing\n",
    "\n",
    "max_features = 10000\n",
    "maxlen = 20\n",
    "\n",
    "(x_train, y_train), (x_test, y_test) = imdb.load_data(num_words=max_features)\n",
    "\n",
    "x_train = preprocessing.sequence.pad_sequences(x_train, maxlen = maxlen)\n",
    "x_test = preprocessing.sequence.pad_sequences(x_test, maxlen = maxlen)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Flatten, Dense, Embedding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_2 (Embedding)      (None, 20, 8)             80000     \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 160)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 161       \n",
      "=================================================================\n",
      "Total params: 80,161\n",
      "Trainable params: 80,161\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Train on 20000 samples, validate on 5000 samples\n",
      "Epoch 1/10\n",
      "20000/20000 [==============================] - 5s 233us/step - loss: 0.6759 - acc: 0.6050 - val_loss: 0.6398 - val_acc: 0.6814\n",
      "Epoch 2/10\n",
      "20000/20000 [==============================] - 2s 109us/step - loss: 0.5657 - acc: 0.7427 - val_loss: 0.5467 - val_acc: 0.7206\n",
      "Epoch 3/10\n",
      "20000/20000 [==============================] - 2s 109us/step - loss: 0.4752 - acc: 0.7808 - val_loss: 0.5113 - val_acc: 0.7384\n",
      "Epoch 4/10\n",
      "20000/20000 [==============================] - 2s 110us/step - loss: 0.4263 - acc: 0.8077 - val_loss: 0.5008 - val_acc: 0.7452\n",
      "Epoch 5/10\n",
      "20000/20000 [==============================] - 2s 109us/step - loss: 0.3930 - acc: 0.8258 - val_loss: 0.4981 - val_acc: 0.7538\n",
      "Epoch 6/10\n",
      "20000/20000 [==============================] - 2s 115us/step - loss: 0.3668 - acc: 0.8395 - val_loss: 0.5014 - val_acc: 0.7530\n",
      "Epoch 7/10\n",
      "20000/20000 [==============================] - 2s 115us/step - loss: 0.3435 - acc: 0.8533 - val_loss: 0.5052 - val_acc: 0.7520\n",
      "Epoch 8/10\n",
      "20000/20000 [==============================] - 2s 119us/step - loss: 0.3223 - acc: 0.8657 - val_loss: 0.5132 - val_acc: 0.7486\n",
      "Epoch 9/10\n",
      "20000/20000 [==============================] - 2s 123us/step - loss: 0.3022 - acc: 0.8766 - val_loss: 0.5213 - val_acc: 0.7490\n",
      "Epoch 10/10\n",
      "20000/20000 [==============================] - 2s 116us/step - loss: 0.2839 - acc: 0.8860 - val_loss: 0.5303 - val_acc: 0.7466\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Embedding(10000,8, input_length=maxlen))\n",
    "\n",
    "model.add(Flatten())\n",
    "\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['acc'])\n",
    "model.summary()\n",
    "\n",
    "history = model.fit(x_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=32,\n",
    "                    validation_split=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "imdb_dir = './datasets/aclImdb'\n",
    "train_dir = os.path.join(imdb_dir, 'train')\n",
    "\n",
    "labels = []\n",
    "texts = []\n",
    "\n",
    "for label_type in ['neg', 'pos']:\n",
    "    dir_name = os.path.join(train_dir, label_type)\n",
    "    for fname in os.listdir(dir_name):\n",
    "        if fname[-4:] == \".txt\":\n",
    "            f = open(os.path.join(dir_name, fname), encoding=\"utf8\")\n",
    "            texts.append(f.read())\n",
    "            f.close()\n",
    "            if label_type == 'neg':\n",
    "                labels.append(0)\n",
    "            else:\n",
    "                labels.append(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "find num 88582 token\n",
      "size of data tensor: (25000, 100)\n",
      "size of label tensor: (25000,)\n"
     ]
    }
   ],
   "source": [
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "import numpy as np\n",
    "\n",
    "maxlen = 100\n",
    "training_samples = 200\n",
    "validation_samples = 10000\n",
    "max_words = 10000\n",
    "\n",
    "tokenizer = Tokenizer(num_words=max_words)\n",
    "tokenizer.fit_on_texts(texts)\n",
    "sequences = tokenizer.texts_to_sequences(texts)\n",
    "\n",
    "word_index = tokenizer.word_index\n",
    "print(\"find num %s token\" % len(word_index))\n",
    "\n",
    "data = pad_sequences(sequences, maxlen=maxlen)\n",
    "labels = np.asarray(labels)\n",
    "print(\"size of data tensor:\", data.shape)\n",
    "print(\"size of label tensor:\", labels.shape)\n",
    "\n",
    "indices = np.arange(data.shape[0])\n",
    "np.random.shuffle(indices)\n",
    "data = data[indices]\n",
    "labels = labels[indices]\n",
    "\n",
    "x_train = data[:training_samples]\n",
    "y_train = labels[:training_samples]\n",
    "x_val = data[training_samples: training_samples + validation_samples]\n",
    "y_val = labels[training_samples: training_samples + validation_samples]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "https://nlp.stanford.edu/projects/glove 에서 2014년 영문 위키피디아를 사용해 사전에 계산된 임베딩을 내려받습니다. 이 파일의 이름은 glove.6B.zip이고 압축 파일 크기는 823MB입니다. 400,000만개의 단어(또는 단어가 아닌 토큰)에 대한 100차원의 임베딩 벡터를 포함하고 있습니다. datasets 폴더 아래에 파일 압축을 해제합니다.\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "find num 400000 word vector\n"
     ]
    }
   ],
   "source": [
    "glove_dir = './datasets/'\n",
    "\n",
    "embeddings_index = {}\n",
    "f = open(os.path.join(glove_dir, 'glove.6B.100d.txt'), encoding=\"utf8\")\n",
    "for line in f:\n",
    "    values = line.split()\n",
    "    word = values[0]\n",
    "    coefs = np.asarray(values[1:], dtype='float32')\n",
    "    embeddings_index[word] = coefs\n",
    "f.close()\n",
    "\n",
    "print(\"find num %s word vector\" % len(embeddings_index))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "embedding_dim = 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "embedding_matrix = np.zeros((max_words, embedding_dim))\n",
    "for word, i in word_index.items():\n",
    "    if i < max_words:\n",
    "        embedding_vector = embeddings_index.get(word)\n",
    "        if embedding_vector is not None:\n",
    "            embedding_matrix[i] = embedding_vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Embedding, Flatten, Dense"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<bound method Network.summary of <keras.engine.sequential.Sequential object at 0x0000024C740E6B38>>"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Embedding(max_words, embedding_dim, input_length=maxlen))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(32, activation='relu'))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "model.summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.layers[0].set_weights([embedding_matrix])\n",
    "model.layers[0].trainable = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 200 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "200/200 [==============================] - 1s 3ms/step - loss: 1.3019 - acc: 0.5300 - val_loss: 0.7357 - val_acc: 0.5042\n",
      "Epoch 2/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.5259 - acc: 0.7450 - val_loss: 0.7504 - val_acc: 0.5030\n",
      "Epoch 3/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.4266 - acc: 0.7950 - val_loss: 0.6790 - val_acc: 0.5748\n",
      "Epoch 4/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.2865 - acc: 0.9500 - val_loss: 0.9250 - val_acc: 0.5122\n",
      "Epoch 5/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.3320 - acc: 0.8500 - val_loss: 0.7917 - val_acc: 0.5177\n",
      "Epoch 6/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.2014 - acc: 0.9550 - val_loss: 1.1285 - val_acc: 0.5096\n",
      "Epoch 7/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.1803 - acc: 0.9600 - val_loss: 0.7354 - val_acc: 0.5467\n",
      "Epoch 8/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.0952 - acc: 1.0000 - val_loss: 0.6977 - val_acc: 0.5789\n",
      "Epoch 9/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.0766 - acc: 1.0000 - val_loss: 0.7356 - val_acc: 0.5689\n",
      "Epoch 10/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.0772 - acc: 1.0000 - val_loss: 0.7584 - val_acc: 0.5533\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer='rmsprop',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['acc'])\n",
    "history = model.fit(x_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=32,\n",
    "                    validation_data=(x_val, y_val))\n",
    "model.save_weights('pre_trained_glove_model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\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.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_7 (Embedding)      (None, 100, 100)          1000000   \n",
      "_________________________________________________________________\n",
      "flatten_6 (Flatten)          (None, 10000)             0         \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 32)                320032    \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 1)                 33        \n",
      "=================================================================\n",
      "Total params: 1,320,065\n",
      "Trainable params: 1,320,065\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Train on 200 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "200/200 [==============================] - 1s 4ms/step - loss: 0.6913 - acc: 0.5250 - val_loss: 0.6940 - val_acc: 0.5111\n",
      "Epoch 2/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.5043 - acc: 0.9750 - val_loss: 0.6924 - val_acc: 0.5243\n",
      "Epoch 3/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.2761 - acc: 1.0000 - val_loss: 0.7035 - val_acc: 0.5208\n",
      "Epoch 4/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.1203 - acc: 1.0000 - val_loss: 0.6969 - val_acc: 0.5418\n",
      "Epoch 5/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.0558 - acc: 1.0000 - val_loss: 0.7006 - val_acc: 0.5345\n",
      "Epoch 6/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.0289 - acc: 1.0000 - val_loss: 0.7096 - val_acc: 0.5343\n",
      "Epoch 7/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.0158 - acc: 1.0000 - val_loss: 0.7184 - val_acc: 0.5367\n",
      "Epoch 8/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.0090 - acc: 1.0000 - val_loss: 0.7167 - val_acc: 0.5452\n",
      "Epoch 9/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.0054 - acc: 1.0000 - val_loss: 0.7341 - val_acc: 0.5316\n",
      "Epoch 10/10\n",
      "200/200 [==============================] - 0s 2ms/step - loss: 0.0034 - acc: 1.0000 - val_loss: 0.7303 - val_acc: 0.5425\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Embedding(max_words, embedding_dim, input_length=maxlen))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(32, activation='relu'))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "model.summary()\n",
    "\n",
    "model.compile(optimizer='rmsprop',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['acc'])\n",
    "history = model.fit(x_train, y_train,\n",
    "                    epochs=10,\n",
    "                    batch_size=32,\n",
    "                    validation_data=(x_val, y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\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.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_dir = os.path.join(imdb_dir, 'test')\n",
    "\n",
    "labels = []\n",
    "texts = []\n",
    "\n",
    "for label_type in ['neg', 'pos']:\n",
    "    dir_name = os.path.join(test_dir, label_type)\n",
    "    for fname in sorted(os.listdir(dir_name)):\n",
    "        if fname[-4:] == '.txt':\n",
    "            f = open(os.path.join(dir_name, fname), encoding='utf8')\n",
    "            texts.append(f.read())\n",
    "            f.close()\n",
    "            if label_type == 'neg':\n",
    "                labels.append(0)\n",
    "            else:\n",
    "                labels.append(1)\n",
    "                \n",
    "sequences = tokenizer.texts_to_sequences(texts)\n",
    "x_test = pad_sequences(sequences, maxlen=maxlen)\n",
    "y_test = np.asarray(labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "25000/25000 [==============================] - 1s 44us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.7520190322637558, 0.55828]"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.load_weights('pre_trained_glove_model.h5')\n",
    "model.evaluate(x_test, y_test)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tensorflowGPU",
   "language": "python",
   "name": "tensorflow"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}