01-One-hot-encoding.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img width=\"800px\" src=\"../fidle/img/00-Fidle-header-01.svg\"></img>\n",
    "\n",
    "# <!-- TITLE --> [IMDB1] - Sentiment analysis with hot-one encoding\n",
    "<!-- DESC --> A basic example of sentiment analysis with sparse encoding, using a dataset from Internet Movie Database (IMDB)\n",
    "<!-- AUTHOR : Jean-Luc Parouty (CNRS/SIMaP) -->\n",
    "\n",
    "## Objectives :\n",
    " - The objective is to guess whether film reviews are **positive or negative** based on the analysis of the text. \n",
    " - Understand the management of **textual data** and **sentiment analysis**\n",
    "\n",
    "Original dataset can be find **[there](http://ai.stanford.edu/~amaas/data/sentiment/)**  \n",
    "Note that [IMDb.com](https://imdb.com) offers several easy-to-use [datasets](https://www.imdb.com/interfaces/)  \n",
    "For simplicity's sake, we'll use the dataset directly [embedded in Keras](https://www.tensorflow.org/api_docs/python/tf/keras/datasets)\n",
    "\n",
    "## What we're going to do :\n",
    "\n",
    " - Retrieve data\n",
    " - Preparing the data\n",
    " - Build a model\n",
    " - Train the model\n",
    " - Evaluate the result\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 1 - Import and init\n",
    "### 1.1 - Python stuff"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "import tensorflow as tf\n",
    "import tensorflow.keras as keras\n",
    "import tensorflow.keras.datasets.imdb as imdb\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "\n",
    "import pandas as pd\n",
    "\n",
    "import os,sys,h5py,json\n",
    "from importlib import reload\n",
    "\n",
    "sys.path.append('..')\n",
    "import fidle.pwk as pwk\n",
    "\n",
    "run_dir = './run/IMDB1'\n",
    "datasets_dir = pwk.init('IMDB1', run_dir)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.2 - Parameters\n",
    "The words in the vocabulary are classified from the most frequent to the rarest.  \n",
    "`vocab_size` is the number of words we will remember in our vocabulary (the other words will be considered as unknown).  \n",
    "`hide_most_frequently` is the number of ignored words, among the most common ones  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "vocab_size           = 10000\n",
    "hide_most_frequently = 0\n",
    "\n",
    "epochs     = 10\n",
    "batch_size = 512"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Override parameters (batch mode) - Just forget this cell"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pwk.override('vocab_size', 'hide_most_frequently', 'batch_size', 'epochs')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 2 - Understanding hot-one encoding\n",
    "#### We have a **sentence** and a **dictionary** :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sentence = \"I've never seen a movie like this before\"\n",
    "\n",
    "dictionary  = {\"a\":0, \"before\":1, \"fantastic\":2, \"i've\":3, \"is\":4, \"like\":5, \"movie\":6, \"never\":7, \"seen\":8, \"this\":9}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### We encode our sentence as a **numerical vector** :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sentence_words = sentence.lower().split()\n",
    "\n",
    "sentence_vect  = [ dictionary[w] for w in sentence_words ]\n",
    "\n",
    "print('Words sentence are         : ', sentence_words)\n",
    "print('Our vectorized sentence is : ', sentence_vect)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Next, we **one-hot** encode our vectorized sentence as a tensor :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---- We get a (sentence length x vector size) matrix of zeros\n",
    "#\n",
    "onehot = np.zeros( (10,8) )\n",
    "\n",
    "# ---- We set some 1 for each word\n",
    "#\n",
    "for i,w in enumerate(sentence_vect):\n",
    "    onehot[w,i]=1\n",
    "\n",
    "# --- Show it\n",
    "#\n",
    "print('In a basic way :\\n\\n', onehot, '\\n\\nWith a pandas wiew :\\n')\n",
    "data={ f'{sentence_words[i]:.^10}':onehot[:,i] for i,w in enumerate(sentence_vect) }\n",
    "df=pd.DataFrame(data)\n",
    "df.index=dictionary.keys()\n",
    "df.style.set_precision(0).highlight_max(axis=0).set_properties(**{'text-align': 'center'})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 2 - Retrieve data\n",
    "\n",
    "IMDb dataset can bet get directly from Keras - see [documentation](https://www.tensorflow.org/api_docs/python/tf/keras/datasets)  \n",
    "Note : Due to their nature, textual data can be somewhat complex.\n",
    "\n",
    "### 2.1 - Data structure :  \n",
    "The dataset is composed of 2 parts: \n",
    "\n",
    " - **reviews**, this will be our **x**\n",
    " - **opinions** (positive/negative), this will be our **y**\n",
    "\n",
    "There are also a **dictionary**, because words are indexed in reviews\n",
    "\n",
    "```\n",
    "<dataset> = (<reviews>, <opinions>)\n",
    "\n",
    "with :  <reviews>  = [ <review1>, <review2>, ... ]\n",
    "        <opinions> = [ <rate1>,   <rate2>,   ... ]   where <ratei>   = integer\n",
    "\n",
    "where : <reviewi> = [ <w1>, <w2>, ...]    <wi> are the index (int) of the word in the dictionary\n",
    "        <ratei>   = int                   0 for negative opinion, 1 for positive\n",
    "\n",
    "\n",
    "<dictionary> = [ <word1>:<w1>, <word2>:<w2>, ... ]\n",
    "\n",
    "with :  <wordi>   = word\n",
    "        <wi>      = int\n",
    "\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.2 - Load dataset\n",
    "For simplicity, we will use a pre-formatted dataset - See [documentation](https://www.tensorflow.org/api_docs/python/tf/keras/datasets/imdb/load_data)  \n",
    "However, Keras offers some usefull tools for formatting textual data - See [documentation](https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/text)  \n",
    "\n",
    "By default : \n",
    " - Start of a sequence will be marked with : 1\n",
    " - Out of vocabulary word will be : 2\n",
    " - First index will be : 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ----- Retrieve x,y\n",
    "#\n",
    "(x_train, y_train), (x_test, y_test) = imdb.load_data( num_words=vocab_size, skip_top=hide_most_frequently)\n",
    "\n",
    "y_train = np.asarray(y_train).astype('float32')\n",
    "y_test  = np.asarray(y_test ).astype('float32')\n",
    "\n",
    "# ---- About\n",
    "#\n",
    "print(\"Max(x_train,x_test)  : \", pwk.rmax([x_train,x_test]) )\n",
    "print(\"Min(x_train,x_test)  : \", pwk.rmin([x_train,x_test]) )\n",
    "print(\"x_train : {}  y_train : {}\".format(x_train.shape, y_train.shape))\n",
    "print(\"x_test  : {}  y_test  : {}\".format(x_test.shape,  y_test.shape))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 3 - About our dataset\n",
    "When we loaded the dataset, we asked for using \\<start\\> as 1, \\<unknown word\\> as 2  \n",
    "So, we shifted the dataset by 3 with the parameter index_from=3\n",
    "\n",
    "### 3.1 - Sentences encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('\\nReview example (x_train[12]) :\\n\\n',x_train[12])\n",
    "print('\\nOpinions (y_train) :\\n\\n',y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.2 - Load dictionary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---- Retrieve dictionary {word:index}, and encode it in ascii\n",
    "#\n",
    "word_index = imdb.get_word_index()\n",
    "\n",
    "# ---- Shift the dictionary from +3\n",
    "#\n",
    "word_index = {w:(i+3) for w,i in word_index.items()}\n",
    "\n",
    "# ---- Add <pad>, <start> and <unknown> tags\n",
    "#\n",
    "word_index.update( {'<pad>':0, '<start>':1, '<unknown>':2, '<undef>':3,} )\n",
    "\n",
    "# ---- Create a reverse dictionary : {index:word}\n",
    "#\n",
    "index_word = {index:word for word,index in word_index.items()} \n",
    "\n",
    "# ---- About dictionary\n",
    "#\n",
    "print('\\nDictionary size     : ', len(word_index))\n",
    "print('\\nSmall extract :\\n')\n",
    "for k in range(440,455):print(f'    {k:2d} : {index_word[k]}' )\n",
    "\n",
    "# ---- Add a nice function to transpose :\n",
    "#\n",
    "def dataset2text(review):\n",
    "    return ' '.join([index_word.get(i, '?') for i in review])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.3 - Have a look, for human"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pwk.subtitle('Review example :')\n",
    "print(x_train[12])\n",
    "pwk.subtitle('After translation :')\n",
    "print(dataset2text(x_train[12]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.4 - Few statistics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sizes=[len(i) for i in x_train]\n",
    "plt.figure(figsize=(16,6))\n",
    "plt.hist(sizes, bins=400)\n",
    "plt.gca().set(title='Distribution of reviews by size - [{:5.2f}, {:5.2f}]'.format(min(sizes),max(sizes)), \n",
    "              xlabel='Size', ylabel='Density', xlim=[0,1500])\n",
    "pwk.save_fig('01-stats-sizes')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "unk=[ 100*(s.count(2)/len(s)) for s in x_train]\n",
    "plt.figure(figsize=(16,6))\n",
    "plt.hist(unk, bins=100)\n",
    "plt.gca().set(title='Percent of unknown words - [{:5.2f}, {:5.2f}]'.format(min(unk),max(unk)), \n",
    "              xlabel='# unknown', ylabel='Density', xlim=[0,30])\n",
    "pwk.save_fig('02-stats-unknown')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 3 - Basic approach with \"one-hot\" vector encoding\n",
    "Basic approach.  \n",
    "\n",
    "Each sentence is encoded with a **vector** of length equal to the **size of the dictionary**.   \n",
    "\n",
    "Each sentence will therefore be encoded with a simple vector.  \n",
    "The value of each component is 0 if the word is not present in the sentence or 1 if the word is present.\n",
    "\n",
    "For a sentence s=[3,4,7] and a dictionary of 10 words...    \n",
    "We wil have a vector v=[0,0,0,1,1,0,0,1,0,0,0]\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.1 - Our one-hot encoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def one_hot_encoder(x, vector_size=10000):\n",
    "    \n",
    "    # ---- Set all to 0\n",
    "    #\n",
    "    x_encoded = np.zeros((len(x), vector_size))\n",
    "    \n",
    "    # ---- For each sentence\n",
    "    #\n",
    "    for i,sentence in enumerate(x):\n",
    "        for word in sentence:\n",
    "            x_encoded[i, word] = 1.\n",
    "\n",
    "    return x_encoded"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.2 - Encoding.."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train = one_hot_encoder(x_train)\n",
    "x_test  = one_hot_encoder(x_test)\n",
    "\n",
    "print(\"To have a look, x_train[12] became :\", x_train[12] )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 4 - Build the model\n",
    "Few remarks :\n",
    " - We'll choose a dense vector size for the embedding output with **dense_vector_size**\n",
    " - **GlobalAveragePooling1D** do a pooling on the last dimension : (None, lx, ly) -> (None, ly)  \n",
    "   In other words: we average the set of vectors/words of a sentence\n",
    " - L'embedding de Keras fonctionne de manière supervisée. Il s'agit d'une couche de *vocab_size* neurones vers *n_neurons* permettant de maintenir une table de vecteurs (les poids constituent les vecteurs). Cette couche ne calcule pas de sortie a la façon des couches normales, mais renvois la valeur des vecteurs. n mots => n vecteurs (ensuite empilés par le pooling)  \n",
    "Voir : [Explication plus détaillée (en)](https://stats.stackexchange.com/questions/324992/how-the-embedding-layer-is-trained-in-keras-embedding-layer)  \n",
    "ainsi que : [Sentiment detection with Keras](https://www.liip.ch/en/blog/sentiment-detection-with-keras-word-embeddings-and-lstm-deep-learning-networks)  \n",
    "\n",
    "More documentation about this model functions :\n",
    " - [Embedding](https://www.tensorflow.org/api_docs/python/tf/keras/layers/Embedding)\n",
    " - [GlobalAveragePooling1D](https://www.tensorflow.org/api_docs/python/tf/keras/layers/GlobalAveragePooling1D)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_model(vector_size=10000):\n",
    "    \n",
    "    model = keras.Sequential()\n",
    "    model.add(keras.layers.Input( shape=(vector_size,) ))\n",
    "    model.add(keras.layers.Dense( 32, activation='relu'))\n",
    "    model.add(keras.layers.Dense( 32, activation='relu'))\n",
    "    model.add(keras.layers.Dense( 1, activation='sigmoid'))\n",
    "    \n",
    "    model.compile(optimizer = 'rmsprop',\n",
    "                  loss      = 'binary_crossentropy',\n",
    "                  metrics   = ['accuracy'])\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 5 - Train the model\n",
    "### 5.1 - Get it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = get_model(vector_size=vocab_size)\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.2 - Add callback"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.makedirs(f'{run_dir}/models',   mode=0o750, exist_ok=True)\n",
    "save_dir = f'{run_dir}/models/best_model.h5'\n",
    "savemodel_callback = tf.keras.callbacks.ModelCheckpoint(filepath=save_dir, verbose=0, save_best_only=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.1 - Train it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%time\n",
    "\n",
    "history = model.fit(x_train,\n",
    "                    y_train,\n",
    "                    epochs          = epochs,\n",
    "                    batch_size      = batch_size,\n",
    "                    validation_data = (x_test, y_test),\n",
    "                    verbose         = 1,\n",
    "                    callbacks       = [savemodel_callback])\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 6 - Evaluate\n",
    "### 6.1 - Training history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pwk.plot_history(history, save_as='02-history')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 6.2 - Reload and evaluate best model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.models.load_model(f'{run_dir}/models/best_model.h5')\n",
    "\n",
    "# ---- Evaluate\n",
    "score  = model.evaluate(x_test, y_test, verbose=0)\n",
    "\n",
    "print('x_test / loss      : {:5.4f}'.format(score[0]))\n",
    "print('x_test / accuracy  : {:5.4f}'.format(score[1]))\n",
    "\n",
    "values=[score[1], 1-score[1]]\n",
    "pwk.plot_donut(values,[\"Accuracy\",\"Errors\"], title=\"#### Accuracy donut is :\", save_as='03-donut')\n",
    "\n",
    "# ---- Confusion matrix\n",
    "\n",
    "y_sigmoid = model.predict(x_test)\n",
    "\n",
    "y_pred = y_sigmoid.copy()\n",
    "y_pred[ y_sigmoid< 0.5 ] = 0\n",
    "y_pred[ y_sigmoid>=0.5 ] = 1    \n",
    "\n",
    "pwk.display_confusion_matrix(y_test,y_pred,labels=range(2))\n",
    "pwk.plot_confusion_matrix(y_test,y_pred,range(2), figsize=(8, 8),normalize=False, save_as='04-confusion-matrix')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pwk.end()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "<img width=\"80px\" src=\"../fidle/img/00-Fidle-logo-01.svg\"></img>"
   ]
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}