19_recurrent_neural_nets/05_sentiment_analysis_imdb.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# LSTM & Word Embeddings for Sentiment Classification"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "RNNs are commonly applied to various natural language processing tasks. We've already encountered sentiment analysis using text data in part three of this book.\n",
    "\n",
    "We are now going to illustrate how to apply an RNN model to text data to detect positive or negative sentiment (which can easily be extended to a finer-grained sentiment scale). We are going to use word embeddings to represent the tokens in the documents. We covered word embeddings in Chapter 15, Word Embeddings. They are an excellent technique to convert text into a continuous vector representation such that the relative location of words in the latent space encodes useful semantic aspects based on the words' usage in context.\n",
    "\n",
    "We saw in the previous RNN example that Keras has a built-in embedding layer that allows us to train vector representations specific to the task at hand. Alternatively, we can use pretrained vectors."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Imports & Settings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:35.422202Z",
     "start_time": "2021-02-23T16:46:35.420277Z"
    }
   },
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:36.902418Z",
     "start_time": "2021-02-23T16:46:35.423915Z"
    }
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "from pathlib import Path\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.metrics import roc_auc_score\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping\n",
    "from tensorflow.keras.datasets import imdb\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, GRU, Embedding\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "import tensorflow.keras.backend as K\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:36.938942Z",
     "start_time": "2021-02-23T16:46:36.903401Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using CPU\n"
     ]
    }
   ],
   "source": [
    "gpu_devices = tf.config.experimental.list_physical_devices('GPU')\n",
    "if gpu_devices:\n",
    "    print('Using GPU')\n",
    "    tf.config.experimental.set_memory_growth(gpu_devices[0], True)\n",
    "else:\n",
    "    print('Using CPU')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:36.952017Z",
     "start_time": "2021-02-23T16:46:36.940222Z"
    }
   },
   "outputs": [],
   "source": [
    "sns.set_style('whitegrid')\n",
    "np.random.seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:36.960141Z",
     "start_time": "2021-02-23T16:46:36.953733Z"
    }
   },
   "outputs": [],
   "source": [
    "results_path = Path('results', 'sentiment_imdb')\n",
    "if not results_path.exists():\n",
    "    results_path.mkdir(parents=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load Reviews"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To keep the data manageable, we will illustrate this use case with the IMDB reviews dataset, which contains 50,000 positive and negative movie reviews evenly split into a train and a test set, and with balanced labels in each dataset. The vocabulary consists of 88,586 tokens.\n",
    "\n",
    "The dataset is bundled into Keras and can be loaded so that each review is represented as an integer-encoded sequence. We can limit the vocabulary to num_words while filtering out frequent and likely less informative words using skip_top, as well as sentences longer than maxlen. We can also choose oov_char, which represents tokens we chose to exclude from the vocabulary on frequency grounds, as follows:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:36.967985Z",
     "start_time": "2021-02-23T16:46:36.961210Z"
    }
   },
   "outputs": [],
   "source": [
    "vocab_size = 20000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:39.920943Z",
     "start_time": "2021-02-23T16:46:36.970408Z"
    }
   },
   "outputs": [],
   "source": [
    "(X_train, y_train), (X_test, y_test) = imdb.load_data(seed=42, \n",
    "                                                      skip_top=0,\n",
    "                                                      maxlen=None, \n",
    "                                                      oov_char=2, \n",
    "                                                      index_from=3,\n",
    "                                                      num_words=vocab_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:40.737546Z",
     "start_time": "2021-02-23T16:46:39.922130Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax = sns.displot([len(review) for review in X_train])\n",
    "ax.set(xscale='log');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In the second step, convert the lists of integers into fixed-size arrays that we can stack and provide as input to our RNN. The pad_sequence function produces arrays of equal length, truncated, and padded to conform to maxlen, as follows:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:40.740230Z",
     "start_time": "2021-02-23T16:46:40.738607Z"
    }
   },
   "outputs": [],
   "source": [
    "maxlen = 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:41.144907Z",
     "start_time": "2021-02-23T16:46:40.741137Z"
    }
   },
   "outputs": [],
   "source": [
    "X_train_padded = pad_sequences(X_train, \n",
    "                        truncating='pre', \n",
    "                        padding='pre', \n",
    "                        maxlen=maxlen)\n",
    "\n",
    "X_test_padded = pad_sequences(X_test, \n",
    "                       truncating='pre', \n",
    "                       padding='pre', \n",
    "                       maxlen=maxlen)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:41.148423Z",
     "start_time": "2021-02-23T16:46:41.145788Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((25000, 100), (25000, 100))"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_padded.shape, X_test_padded.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Define Model Architecture"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can define our RNN architecture. The first layer learns the word embeddings. We define the embedding dimension as previously using the input_dim keyword to set the number of tokens that we need to embed, the output_dim keyword, which defines the size of each embedding, and how long each input sequence is going to be."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:41.161746Z",
     "start_time": "2021-02-23T16:46:41.149384Z"
    }
   },
   "outputs": [],
   "source": [
    "K.clear_session()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Custom Loss Metric"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:41.168825Z",
     "start_time": "2021-02-23T16:46:41.162609Z"
    }
   },
   "outputs": [],
   "source": [
    "embedding_size = 100"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Note that we are using GRUs this time, which train faster and perform better on smaller data. We are also using dropout for regularization, as follows:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:41.630876Z",
     "start_time": "2021-02-23T16:46:41.169634Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Layer gru will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding (Embedding)        (None, 100, 100)          2000000   \n",
      "_________________________________________________________________\n",
      "gru (GRU)                    (None, 32)                12864     \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 1)                 33        \n",
      "=================================================================\n",
      "Total params: 2,012,897\n",
      "Trainable params: 2,012,897\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "rnn = Sequential([\n",
    "    Embedding(input_dim=vocab_size, \n",
    "              output_dim= embedding_size, \n",
    "              input_length=maxlen),\n",
    "    GRU(units=32,  \n",
    "        dropout=0.2, # comment out to use optimized GPU implementation\n",
    "        recurrent_dropout=0.2),\n",
    "    Dense(1, activation='sigmoid')\n",
    "])\n",
    "rnn.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The resulting model has over 2 million parameters."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We compile the model to use our custom AUC metric, which we introduced previously, and train with early stopping:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:41.650602Z",
     "start_time": "2021-02-23T16:46:41.632196Z"
    }
   },
   "outputs": [],
   "source": [
    "rnn.compile(loss='binary_crossentropy', \n",
    "            optimizer='RMSProp', \n",
    "            metrics=['accuracy', \n",
    "                     tf.keras.metrics.AUC(name='AUC')])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:41.654062Z",
     "start_time": "2021-02-23T16:46:41.651830Z"
    }
   },
   "outputs": [],
   "source": [
    "rnn_path = (results_path / 'lstm.h5').as_posix()\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath=rnn_path,\n",
    "                               verbose=1,\n",
    "                               monitor='val_AUC',\n",
    "                               mode='max',\n",
    "                               save_best_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T16:46:41.662536Z",
     "start_time": "2021-02-23T16:46:41.656135Z"
    }
   },
   "outputs": [],
   "source": [
    "early_stopping = EarlyStopping(monitor='val_AUC', \n",
    "                               mode='max',\n",
    "                              patience=5,\n",
    "                              restore_best_weights=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Training stops after eight epochs and we recover the weights for the best models to find a high test AUC of 0.9346:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T17:06:11.002545Z",
     "start_time": "2021-02-23T16:46:41.663341Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "782/782 [==============================] - ETA: 0s - loss: 0.4345 - accuracy: 0.7942 - AUC: 0.8801\n",
      "Epoch 00001: val_AUC improved from -inf to 0.93268, saving model to results/sentiment_imdb/lstm.h5\n",
      "782/782 [==============================] - 117s 150ms/step - loss: 0.4345 - accuracy: 0.7942 - AUC: 0.8801 - val_loss: 0.3455 - val_accuracy: 0.8501 - val_AUC: 0.9327\n",
      "Epoch 2/100\n",
      "782/782 [==============================] - ETA: 0s - loss: 0.2887 - accuracy: 0.8803 - AUC: 0.9492\n",
      "Epoch 00002: val_AUC improved from 0.93268 to 0.93450, saving model to results/sentiment_imdb/lstm.h5\n",
      "782/782 [==============================] - 116s 148ms/step - loss: 0.2887 - accuracy: 0.8803 - AUC: 0.9492 - val_loss: 0.3317 - val_accuracy: 0.8568 - val_AUC: 0.9345\n",
      "Epoch 3/100\n",
      "782/782 [==============================] - ETA: 0s - loss: 0.2442 - accuracy: 0.9023 - AUC: 0.9634\n",
      "Epoch 00003: val_AUC did not improve from 0.93450\n",
      "782/782 [==============================] - 117s 149ms/step - loss: 0.2442 - accuracy: 0.9023 - AUC: 0.9634 - val_loss: 0.4815 - val_accuracy: 0.8212 - val_AUC: 0.9343\n",
      "Epoch 4/100\n",
      "782/782 [==============================] - ETA: 0s - loss: 0.2143 - accuracy: 0.9160 - AUC: 0.9716\n",
      "Epoch 00004: val_AUC improved from 0.93450 to 0.94048, saving model to results/sentiment_imdb/lstm.h5\n",
      "782/782 [==============================] - 116s 148ms/step - loss: 0.2143 - accuracy: 0.9160 - AUC: 0.9716 - val_loss: 0.3312 - val_accuracy: 0.8645 - val_AUC: 0.9405\n",
      "Epoch 5/100\n",
      "782/782 [==============================] - ETA: 0s - loss: 0.1901 - accuracy: 0.9269 - AUC: 0.9774\n",
      "Epoch 00005: val_AUC improved from 0.94048 to 0.94152, saving model to results/sentiment_imdb/lstm.h5\n",
      "782/782 [==============================] - 116s 148ms/step - loss: 0.1901 - accuracy: 0.9269 - AUC: 0.9774 - val_loss: 0.3367 - val_accuracy: 0.8658 - val_AUC: 0.9415\n",
      "Epoch 6/100\n",
      "782/782 [==============================] - ETA: 0s - loss: 0.1693 - accuracy: 0.9361 - AUC: 0.9819\n",
      "Epoch 00006: val_AUC did not improve from 0.94152\n",
      "782/782 [==============================] - 116s 148ms/step - loss: 0.1693 - accuracy: 0.9361 - AUC: 0.9819 - val_loss: 0.3186 - val_accuracy: 0.8632 - val_AUC: 0.9399\n",
      "Epoch 7/100\n",
      "782/782 [==============================] - ETA: 0s - loss: 0.1519 - accuracy: 0.9426 - AUC: 0.9851\n",
      "Epoch 00007: val_AUC did not improve from 0.94152\n",
      "782/782 [==============================] - 115s 147ms/step - loss: 0.1519 - accuracy: 0.9426 - AUC: 0.9851 - val_loss: 0.5009 - val_accuracy: 0.8056 - val_AUC: 0.9354\n",
      "Epoch 8/100\n",
      "782/782 [==============================] - ETA: 0s - loss: 0.1364 - accuracy: 0.9505 - AUC: 0.9878\n",
      "Epoch 00008: val_AUC did not improve from 0.94152\n",
      "782/782 [==============================] - 117s 150ms/step - loss: 0.1364 - accuracy: 0.9505 - AUC: 0.9878 - val_loss: 0.3860 - val_accuracy: 0.8547 - val_AUC: 0.9337\n",
      "Epoch 9/100\n",
      "782/782 [==============================] - ETA: 0s - loss: 0.1206 - accuracy: 0.9564 - AUC: 0.9902\n",
      "Epoch 00009: val_AUC did not improve from 0.94152\n",
      "782/782 [==============================] - 119s 152ms/step - loss: 0.1206 - accuracy: 0.9564 - AUC: 0.9902 - val_loss: 0.3833 - val_accuracy: 0.8562 - val_AUC: 0.9343\n",
      "Epoch 10/100\n",
      "782/782 [==============================] - ETA: 0s - loss: 0.1061 - accuracy: 0.9620 - AUC: 0.9922\n",
      "Epoch 00010: val_AUC did not improve from 0.94152\n",
      "782/782 [==============================] - 118s 151ms/step - loss: 0.1061 - accuracy: 0.9620 - AUC: 0.9922 - val_loss: 0.3820 - val_accuracy: 0.8545 - val_AUC: 0.9315\n"
     ]
    }
   ],
   "source": [
    "training = rnn.fit(X_train_padded,\n",
    "                   y_train,\n",
    "                   batch_size=32,\n",
    "                   epochs=100,\n",
    "                   validation_data=(X_test_padded, y_test),\n",
    "                   callbacks=[early_stopping, checkpointer],\n",
    "                   verbose=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T17:06:11.006806Z",
     "start_time": "2021-02-23T17:06:11.003730Z"
    }
   },
   "outputs": [],
   "source": [
    "history = pd.DataFrame(training.history)\n",
    "history.index += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T17:07:49.772399Z",
     "start_time": "2021-02-23T17:07:49.119899Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axes = plt.subplots(ncols=2, figsize=(14, 4))\n",
    "df1 = (history[['accuracy', 'val_accuracy']]\n",
    "       .rename(columns={'accuracy': 'Training',\n",
    "                        'val_accuracy': 'Validation'}))\n",
    "df1.plot(ax=axes[0], title='Accuracy', xlim=(1, len(history)))\n",
    "\n",
    "axes[0].axvline(df1.Validation.idxmax(), ls='--', lw=1, c='k')\n",
    "\n",
    "df2 = (history[['AUC', 'val_AUC']]\n",
    "       .rename(columns={'AUC': 'Training',\n",
    "                        'val_AUC': 'Validation'}))\n",
    "df2.plot(ax=axes[1], title='Area under the ROC Curve', xlim=(1, len(history)))\n",
    "\n",
    "axes[1].axvline(df2.Validation.idxmax(), ls='--', lw=1, c='k')\n",
    "\n",
    "for i in [0, 1]:\n",
    "    axes[i].set_xlabel('Epoch')\n",
    "\n",
    "sns.despine()\n",
    "fig.tight_layout()\n",
    "fig.savefig(results_path / 'rnn_imdb_cv', dpi=300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T17:08:02.501972Z",
     "start_time": "2021-02-23T17:07:52.180763Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(25000, 1)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_score = rnn.predict(X_test_padded)\n",
    "y_score.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-02-23T17:08:02.513650Z",
     "start_time": "2021-02-23T17:08:02.503355Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.941730672"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_auc_score(y_score=y_score.squeeze(), y_true=y_test)"
   ]
  }
 ],
 "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": "3.8.8"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": true,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}