02 - 01 - Natural Language Processing - Words Embedding.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Natural Language Processing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Toxicity Demo - try it on your own, I warned you!\n",
    "\n",
    "https://storage.googleapis.com/tfjs-models/demos/toxicity/index.html\n",
    "\n",
    "Source code:\n",
    "\n",
    "https://github.com/tensorflow/tfjs-models/tree/master/toxicity"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualize word embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "\n",
    "import tensorflow_datasets as tfds\n",
    "tfds.disable_progress_bar()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "embedding_layer = layers.Embedding(1000, 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[-0.03378884,  0.0144037 ,  0.01714157, -0.0405741 , -0.0396964 ],\n",
       "       [ 0.04960353,  0.01850356, -0.01158099,  0.04811411,  0.02892783],\n",
       "       [ 0.04558824, -0.00311097, -0.01646082, -0.01005101, -0.00128775]],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result = embedding_layer(tf.constant([1,2,3]))\n",
    "result.numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TensorShape([2, 4, 5])"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result = embedding_layer(tf.constant([[0,1,2,3],[3,4,5,6]]))\n",
    "result.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 309 ms, sys: 14 ms, total: 323 ms\n",
      "Wall time: 329 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "(train_data, test_data), info = tfds.load('imdb_reviews/subwords8k',\n",
    "                                          split=(tfds.Split.TRAIN,\n",
    "                                                 tfds.Split.TEST),\n",
    "                                          with_info=True,\n",
    "                                          as_supervised=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['the_',\n",
       " ', ',\n",
       " '. ',\n",
       " 'a_',\n",
       " 'and_',\n",
       " 'of_',\n",
       " 'to_',\n",
       " 's_',\n",
       " 'is_',\n",
       " 'br',\n",
       " 'in_',\n",
       " 'I_',\n",
       " 'that_',\n",
       " 'this_',\n",
       " 'it_',\n",
       " ' /><',\n",
       " ' />',\n",
       " 'was_',\n",
       " 'The_',\n",
       " 'as_']"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "encoder = info.features['text'].encoder\n",
    "encoder.subwords[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "padded_shapes = ([None],())\n",
    "train_batches = train_data.shuffle(1000).padded_batch(10, padded_shapes = padded_shapes)\n",
    "test_batches = test_data.shuffle(1000).padded_batch(10, padded_shapes = padded_shapes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[7522, 4943,    9, ...,    0,    0,    0],\n",
       "       [ 398,  198,   46, ...,  160, 5748, 7975],\n",
       "       [ 274,   83,   12, ...,    0,    0,    0],\n",
       "       ...,\n",
       "       [  12,  284,   14, ...,    0,    0,    0],\n",
       "       [  12,  129,   12, ...,    0,    0,    0],\n",
       "       [  62,   27,   18, ...,    0,    0,    0]])"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_batch, train_labels = next(iter(train_batches))\n",
    "train_batch.numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_1 (Embedding)      (None, None, 16)          130960    \n",
      "_________________________________________________________________\n",
      "global_average_pooling1d (Gl (None, 16)                0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 1)                 17        \n",
      "=================================================================\n",
      "Total params: 130,977\n",
      "Trainable params: 130,977\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "embedding_dim = 16\n",
    "\n",
    "model = keras.Sequential([\n",
    "    layers.Embedding(encoder.vocab_size, embedding_dim),\n",
    "    layers.GlobalAveragePooling1D(),\n",
    "    layers.Dense(1, activation='sigmoid')\n",
    "])\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "2500/2500 [==============================] - 9s 4ms/step - loss: 0.6347 - accuracy: 0.6896 - val_loss: 0.0000e+00 - val_accuracy: 0.0000e+00\n",
      "Epoch 2/10\n",
      "2500/2500 [==============================] - 6s 2ms/step - loss: 0.4636 - accuracy: 0.8389 - val_loss: 0.4038 - val_accuracy: 0.8550\n",
      "Epoch 3/10\n",
      "2500/2500 [==============================] - 6s 2ms/step - loss: 0.3612 - accuracy: 0.8794 - val_loss: 0.3288 - val_accuracy: 0.8650\n",
      "Epoch 4/10\n",
      "2500/2500 [==============================] - 6s 2ms/step - loss: 0.3078 - accuracy: 0.8976 - val_loss: 0.3208 - val_accuracy: 0.8800\n",
      "Epoch 5/10\n",
      "2500/2500 [==============================] - 6s 2ms/step - loss: 0.2735 - accuracy: 0.9072 - val_loss: 0.3321 - val_accuracy: 0.8550\n",
      "Epoch 6/10\n",
      "2500/2500 [==============================] - 6s 2ms/step - loss: 0.2490 - accuracy: 0.9158 - val_loss: 0.3020 - val_accuracy: 0.8650\n",
      "Epoch 7/10\n",
      "2500/2500 [==============================] - 6s 2ms/step - loss: 0.2296 - accuracy: 0.9223 - val_loss: 0.2702 - val_accuracy: 0.8950\n",
      "Epoch 8/10\n",
      "2500/2500 [==============================] - 6s 2ms/step - loss: 0.2145 - accuracy: 0.9298 - val_loss: 0.3138 - val_accuracy: 0.8550\n",
      "Epoch 9/10\n",
      "2500/2500 [==============================] - 6s 2ms/step - loss: 0.2005 - accuracy: 0.9340 - val_loss: 0.3789 - val_accuracy: 0.8500\n",
      "Epoch 10/10\n",
      "2500/2500 [==============================] - 6s 2ms/step - loss: 0.1882 - accuracy: 0.9385 - val_loss: 0.2658 - val_accuracy: 0.8750\n",
      "CPU times: user 3min 1s, sys: 56.2 s, total: 3min 57s\n",
      "Wall time: 1min 3s\n"
     ]
    }
   ],
   "source": [
    "%%time \n",
    "# 1 minutes - just run it\n",
    "model.compile(optimizer='adam',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "history = model.fit(train_batches,\n",
    "                    epochs=10,\n",
    "                    validation_data=test_batches,\n",
    "                    validation_steps=20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x648 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x648 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "history_dict = history.history\n",
    "\n",
    "acc = history_dict['accuracy']\n",
    "val_acc = history_dict['val_accuracy']\n",
    "loss = history_dict['loss']\n",
    "val_loss = history_dict['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\n",
    "\n",
    "plt.figure(figsize=(12,9))\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "\n",
    "plt.figure(figsize=(12,9))\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend(loc='lower right')\n",
    "plt.ylim((0.5,1))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(8185, 16)\n"
     ]
    }
   ],
   "source": [
    "e = model.layers[0]\n",
    "weights = e.get_weights()[0]\n",
    "print(weights.shape) # shape: (vocab_size, embedding_dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "encoder = info.features['text'].encoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "import io\n",
    "\n",
    "encoder = info.features['text'].encoder\n",
    "\n",
    "out_v = io.open('vecs.tsv', 'w', encoding='utf-8')\n",
    "out_m = io.open('meta.tsv', 'w', encoding='utf-8')\n",
    "\n",
    "for num, word in enumerate(encoder.subwords):\n",
    "    vec = weights[num + 1]  # skip 0, it's padding.\n",
    "    out_m.write(word + \"\\n\")\n",
    "    out_v.write('\\t'.join([str(x) for x in vec]) + \"\\n\")\n",
    "out_v.close()\n",
    "out_m.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Open the [Embedding Projector](http://projector.tensorflow.org/)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "http://projector.tensorflow.org/"
   ]
  }
 ],
 "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": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}