sentiment_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# Sentiment Analysis\n",
    "\n",
    "---\n",
    "\n",
    "In this notebook I have built an end-to-end sentiment classification system from scratch. The system accepts a movie review as input and classifies it as either positive or negative. There are three main steps:\n",
    "\n",
    "* **Preprocess** - Split the data into train and test sets, tokenize, stem words, create bag-of-words features, etc.\n",
    "* **Models** - Create and experiment with different models: GaussianNB, Gradient-Boosted Decision Tree classifier and Recurrent Neural Network.\n",
    "* **Evaluation** - Compare the performances of the models and summarize steps to make the chosen model do better.\n",
    "\n",
    "## Table of Contents\n",
    "\n",
    "[Step 1: Explore the data](#step1)\n",
    "\n",
    "[Step 2: Preprocess the data](#step2)\n",
    "\n",
    "[Step 3: Extract Bag-of-words features](#step3)\n",
    "\n",
    "[Step 4: Classification using BoW features](#step4)\n",
    "\n",
    "[Step 5: Switching gears - RNNs](#step5)\n",
    "\n",
    "[Step 6: Evaluation](#step6)\n",
    "\n",
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Step 1: Exploring the data <a class=\"anchor\" id=\"step1\"></a>\n",
    "\n",
    "The [IMDb dataset](http://ai.stanford.edu/~amaas/data/sentiment/) consists of movie reviews from the website [imdb.com](http://www.imdb.com/), each labeled as either '**pos**itive', if the reviewer enjoyed the film, or '**neg**ative' otherwise.\n",
    "\n",
    "> Maas, Andrew L., et al. [Learning Word Vectors for Sentiment Analysis](http://ai.stanford.edu/~amaas/data/sentiment/). In _Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies_. Association for Computational Linguistics, 2011.\n",
    "\n",
    "### Load the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "IMDb reviews: train = 12500 pos / 12500 neg, test = 12500 pos / 12500 neg\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import glob\n",
    "import pickle\n",
    "\n",
    "def read_imdb_data(data_dir='data/imdb-reviews'):\n",
    "    \"\"\"Read IMDb movie reviews from given directory.\n",
    "    \n",
    "    Directory structure expected:\n",
    "    - data/\n",
    "        - train/\n",
    "            - pos/\n",
    "            - neg/\n",
    "        - test/\n",
    "            - pos/\n",
    "            - neg/\n",
    "    \n",
    "    \"\"\"\n",
    "\n",
    "    # Data, labels to be returned in nested dicts matching the dir. structure\n",
    "    data = {}\n",
    "    labels = {}\n",
    "\n",
    "    # Assume 2 sub-directories: train, test\n",
    "    for data_type in ['train', 'test']:\n",
    "        data[data_type] = {}\n",
    "        labels[data_type] = {}\n",
    "\n",
    "        # Assume 2 sub-directories for sentiment (label): pos, neg\n",
    "        for sentiment in ['pos', 'neg']:\n",
    "            data[data_type][sentiment] = []\n",
    "            labels[data_type][sentiment] = []\n",
    "            \n",
    "            # Fetch list of files for this sentiment\n",
    "            path = os.path.join(data_dir, data_type, sentiment, '*.txt')\n",
    "            files = glob.glob(path)\n",
    "            \n",
    "            # Read reviews data and assign labels\n",
    "            for f in files:\n",
    "                with open(f) as review:\n",
    "                    data[data_type][sentiment].append(review.read())\n",
    "                    labels[data_type][sentiment].append(sentiment)\n",
    "            \n",
    "            assert len(data[data_type][sentiment]) == len(labels[data_type][sentiment]), \\\n",
    "                    \"{}/{} data size does not match labels size\".format(data_type, sentiment)\n",
    "    \n",
    "    # Return data, labels as nested dicts\n",
    "    return data, labels\n",
    "\n",
    "# Read imdb data and write them into pickle files\n",
    "if not (os.path.isfile('data.pickle') and os.path.isfile('labels.pickle')):\n",
    "    data, labels = read_imdb_data()\n",
    "    with open(\"data.pickle\", \"wb\") as handle:\n",
    "        pickle.dump(data, handle, protocol=pickle.HIGHEST_PROTOCOL)\n",
    "    with open(\"labels.pickle\", \"wb\") as handle:\n",
    "        pickle.dump(labels, handle, protocol=pickle.HIGHEST_PROTOCOL)\n",
    "\n",
    "        \n",
    "# Open the pickle files\n",
    "if os.path.getsize(\"data.pickle\") > 0:\n",
    "    with open(\"data.pickle\", \"rb\") as handle:\n",
    "        data = pickle.load(handle)\n",
    "else:\n",
    "    print (\"You're trying to load an empty file\")\n",
    "\n",
    "if os.path.getsize(\"data.pickle\") > 0:\n",
    "    with open(\"labels.pickle\", \"rb\") as handle:\n",
    "        labels = pickle.load(handle)\n",
    "else:\n",
    "    print (\"You're trying to load an empty file\"        )\n",
    "\n",
    "print(\"IMDb reviews: train = {} pos / {} neg, test = {} pos / {} neg\".format(\n",
    "        len(data['train']['pos']), len(data['train']['neg']),\n",
    "        len(data['test']['pos']), len(data['test']['neg'])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Now that the data is loaded in, let's take a quick look at one of the positive reviews:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hood of the Living Dead had a lot to live up to even before the opening credits began. First, any play on \"...of the living dead\" invokes His Holiness Mr. Romero and instantly sets up a high standard to which many movies cannot afford to aspire. And second, my movie-watching companion professed doubt that any urban horror film would surpass the seminal Leprechaun In the Hood. Skeptical, we settled in to watch. <br /><br />We were rewarded with a surprisingly sincere and good-hearted zombie film. Oh, certainly the budget is low, and of course the directors' amateurs friends populate the cast, but Hood of the Living Dead loves zombie cinema. Cheap? Yeah. But when it's this cheap, you can clearly see where LOVE holds it together. <br /><br />Ricky works in a lab during the day and as a surrogate parent to his younger brother at night. He dreams of moving out of Oakland. Before this planned escape, however, his brother is shot to death in a drive-by. Ricky's keen scientific mind presents an option superior to CPR or 911: injections of his lab's experimental regenerative formula. Sadly, little bro wakes up in an ambulance as a bloodthirsty Oakland zombie! Chaos and mayhem! I think it's more economical to eat your enemies than take vengeance in a drive-by, but then again, I'm a poor judge of the complexities of urban life. (How poor a judge? In response to a gory scene involving four men, I opined \"Ah-ha! White t-shirts on everyone so the blood shows up. Economical! I used the same technique in my own low-budget horror film.\" Jordan replied, \"No, that's gang dress. White t-shirts were banned from New Orleans bars for a time as a result.\" Oh.)<br /><br />A lot of the movie is set in someone's living room, so there's a great deal of hanging out and waiting for the zombies. But the characters are sympathetic and the movie is sincere-- it surpasses its budget in spirit. <br /><br />Zombie explanation: When man plays God, zombies arise! Or, perhaps: Follow FDA-approved testing rules before human experimentation! <br /><br />Contribution to the zombie canon: This is the first zombie movie I've seen with a drive-by shooting. As far as the actual zombies go, infection is spread with a bite as usual, but quite unusually head shots don't work-- it's heart shots that kill. Zombies have pulses, the absence of which proves true death. And these zombies make pretty cool jaguar-growl noises. <br /><br />Gratuitous zombie movie in-joke: A mercenary named Romero. Groan. <br /><br />Favorite zombie: Jaguar-noise little brother zombie, of course!\n"
     ]
    }
   ],
   "source": [
    "print(data['train']['pos'][2])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "And one with a negative sentiment:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There was nothing about this movie that I liked. It was so obviously low-budget with bad lighting and camera work (almost like Blair Witch Project, only it wasn't supposed to be that way). There wasn't really much to the plot, and the movie just drug on and on. I actually fast-forwarded through the last 1/3 of the movies, but that did not help matters much. It looked like it might be good from the box, but I must say again: nothing about this movie even resembled good. No good actors, the special effects were so fake, the camera work was horrible, and the dialogue was painfully terrible. On my own personal scale, I give this movie a 0 of 10. Yikes!\n"
     ]
    }
   ],
   "source": [
    "print(data['train']['neg'][2])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Make a wordcloud visualization of the reviews"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYEAAADKCAYAAABDsfw/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBGUYIQICgUBQhvl/H54u1V3OfqEAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7bf67c3f28>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "from wordcloud import WordCloud, STOPWORDS\n",
    "\n",
    "def wordcloud_by_sentiment(sentiment):\n",
    "    \"\"\"Create a wordcloud by sentiment.\n",
    "    \n",
    "    Parameters:\n",
    "    sentiment: \"pos\"/\"neg\" string\n",
    "    \n",
    "    Returns: \n",
    "    A wordcloud of the reviews that have the given sentiment    \n",
    "    \"\"\"    \n",
    "    # Combine all reviews for the desired sentiment\n",
    "    combined_text = \" \".join([review for review in data['train'][sentiment]])\n",
    "\n",
    "    # Initialize wordcloud object\n",
    "    wc = WordCloud(background_color='white', max_words=50,\n",
    "            # update stopwords to include common words like film and movie\n",
    "            stopwords = STOPWORDS.update(['br','film','movie']))\n",
    "\n",
    "    # Generate and plot wordcloud\n",
    "    plt.imshow(wc.generate(combined_text))\n",
    "    plt.axis('off')\n",
    "    plt.show()\n",
    "\n",
    "# Create a wordcloud from positive reviews    \n",
    "wordcloud_by_sentiment('pos')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYEAAADKCAYAAABDsfw/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nD1YYjwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7bf67cb710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Create a wordcloud from negative reviews\n",
    "wordcloud_by_sentiment('neg')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Form training and test sets\n",
    "\n",
    "Combine the positive and negative documents to get one unified training set and one unified test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "IMDb reviews (combined): train = 25000, test = 25000\n"
     ]
    }
   ],
   "source": [
    "from sklearn.utils import shuffle\n",
    "\n",
    "def prepare_imdb_data(data):\n",
    "    \"\"\"Prepare training and test sets from IMDb movie reviews.\"\"\"\n",
    "    \n",
    "    # Combine positive and negative reviews and labels\n",
    "    data_train = data[\"train\"][\"pos\"] + data[\"train\"][\"neg\"]\n",
    "    data_test = data[\"test\"][\"pos\"] + data[\"test\"][\"neg\"]\n",
    "    labels_train = [\"pos\"] * len(data[\"train\"][\"pos\"] ) + [\"neg\"] * len(data[\"train\"][\"neg\"])\n",
    "    labels_test = [\"pos\"] * len(data[\"test\"][\"pos\"] ) + [\"neg\"] * len(data[\"test\"][\"neg\"])\n",
    "    \n",
    "    # Shuffle reviews and corresponding labels within training and test sets\n",
    "    data_train, labels_train = shuffle(data_train,labels_train)\n",
    "    data_test, labels_test = shuffle(data_test,labels_test)\n",
    "    \n",
    "    # Return a unified training data, test data, training labels, test labets\n",
    "    return data_train, data_test, labels_train, labels_test\n",
    "\n",
    "data_train, data_test, labels_train, labels_test = prepare_imdb_data(data)\n",
    "print(\"IMDb reviews (combined): train = {}, test = {}\".format(len(data_train), len(data_test)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Step 2. Preprocessing <a class=\"anchor\" id=\"step2\"></a>\n",
    "\n",
    "As seen in the sample reviews, the raw data includes HTML. Therefore there are HTML tags that need to be removed. We also need to remove non-letter characters, normalize uppercase letters by converting them to lowercase, tokenize, remove stop words, and stem the remaining words in each document.\n",
    "\n",
    "### Convert each review to words\n",
    "\n",
    "The function `review_to_words()` performs all these steps. First, let's import all the necessary libraries. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package stopwords to /home/trang/nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n"
     ]
    }
   ],
   "source": [
    "# BeautifulSoup to easily remove HTML tags\n",
    "from bs4 import BeautifulSoup \n",
    "\n",
    "# RegEx for removing non-letter characters\n",
    "import re\n",
    "\n",
    "# NLTK library for the remaining steps\n",
    "import nltk\n",
    "nltk.download(\"stopwords\")   # download list of stopwords\n",
    "from nltk.corpus import stopwords # import stopwords\n",
    "\n",
    "from nltk.stem.porter import *\n",
    "stemmer = PorterStemmer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['test', 'test', 'would', 'make', 'great', 'movi', 'review']"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def review_to_words(review):\n",
    "    \"\"\"Convert a raw review string into a sequence of words.\"\"\"\n",
    "    # Remove HTML tags using BeautifulSoup\n",
    "    clean_text = BeautifulSoup(review, \"html5lib\").get_text()\n",
    "    \n",
    "    # Remove non-letters using RegEx\n",
    "    clean_text = re.sub(r\"[^a-zA-Z]\", \" \", clean_text)\n",
    "    \n",
    "    # Convert to lowercase and split text into words\n",
    "    words = (clean_text.lower()).split()\n",
    "    \n",
    "    # Remove stopwords\n",
    "    words = [w for w in words if w not in stopwords.words(\"english\")]\n",
    "    \n",
    "    # Reduce words to their stems\n",
    "    words = [stemmer.stem(w) for w in words]\n",
    "\n",
    "    # Return final list of words\n",
    "    return words\n",
    "\n",
    "review_to_words(\"\"\"This is just a <em>test</em>.<br/><br />\n",
    "But if it wasn't a test, it would make for a <b>Great</b> movie review!\"\"\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "With the function `review_to_words()` fully implemeneted, we can apply it to all reviews in both training and test datasets. This may take a while, so let's build in a mechanism to write to a cache file and retrieve from it later."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Read preprocessed data from cache file: preprocessed_data.pkl\n",
      "\n",
      "--- Raw review ---\n",
      "don't expect much from this film. In many ways this film resembles a film that Doris Day starred in in 1956,title, Julie. In this film Doris,who was a flight attendant,stewardess,in those days,landed the air craft after her derange husband,played by Louis Jordan shot the captain. She did a far better job,more convincing,than Kim Ojah,who took control of a 747 and manage to land it without much help from the control tower. I know a little about 747 aircraft,i use to be a flight attendant myself. Like i said,do not expect much from this film,it was done on a cheap budget. The producers were to cheap to use a plane with the name of a airline on it. Oceanic is one name that several movies have used. The only writing on this plane was the name of the company that made the aircraft.\n",
      "\n",
      "--- Preprocessed words ---\n",
      "['robert', 'taylor', 'mad', 'buffalo', 'hunter', 'charli', 'gilson', 'main', 'charact', 'film', 'begin', 'think', 'charli', 'would', 'end', 'redeem', 'like', 'john', 'wayn', 'searcher', 'jame', 'stewart', 'nake', 'spur', 'film', 'goe', 'along', 'gilson', 'keep', 'atroc', 'realiz', 'hope', 'stewart', 'granger', 'sandi', 'mckenzi', 'want', 'stop', 'hunt', 'realiz', 'buffalo', 'soon', 'gone', 'becom', 'disgust', 'act', 'kill', 'gilson', 'natur', 'killer', 'make', 'distinct', 'anim', 'human', 'be', 'debra', 'paget', 'indian', 'girl', 'surpris', 'charact', 'consid', 'self', 'impos', 'censorship', 'time', 'lie', 'gilson', 'total', 'resign', 'even', 'though', 'hate', 'last', 'scene', 'frozen', 'gilson', 'unforgett']\n",
      "\n",
      "--- Label ---\n",
      "pos\n"
     ]
    }
   ],
   "source": [
    "import pickle\n",
    "\n",
    "cache_dir = os.path.join(\"cache\", \"sentiment_analysis\")  # where to store cache files\n",
    "os.makedirs(cache_dir, exist_ok=True)  # ensure cache directory exists\n",
    "\n",
    "def preprocess_data(data_train, data_test, labels_train, labels_test,\n",
    "                    cache_dir=cache_dir, cache_file=\"preprocessed_data.pkl\"):\n",
    "    \"\"\"Convert each review to words; read from cache if available.\"\"\"\n",
    "\n",
    "    # If cache_file is not None, try to read from it first\n",
    "    cache_data = None\n",
    "    if cache_file is not None:\n",
    "        try:\n",
    "            with open(os.path.join(cache_dir, cache_file), \"rb\") as f:\n",
    "                cache_data = pickle.load(f)\n",
    "            print(\"Read preprocessed data from cache file:\", cache_file)\n",
    "        except:\n",
    "            pass  # unable to read from cache, but that's okay\n",
    "    \n",
    "    # If cache is missing, then do the heavy lifting\n",
    "    if cache_data is None:\n",
    "        # Preprocess training and test data to obtain words for each review\n",
    "        words_train = list(map(review_to_words, data_train))\n",
    "        words_test = list(map(review_to_words, data_test))\n",
    "        \n",
    "        # Write to cache file for future runs\n",
    "        if cache_file is not None:\n",
    "            cache_data = dict(words_train=words_train, words_test=words_test,\n",
    "                              labels_train=labels_train, labels_test=labels_test)\n",
    "            with open(os.path.join(cache_dir, cache_file), \"wb\") as f:\n",
    "                pickle.dump(cache_data, f)\n",
    "            print(\"Wrote preprocessed data to cache file:\", cache_file)\n",
    "    else:\n",
    "        # Unpack data loaded from cache file\n",
    "        words_train, words_test, labels_train, labels_test = (cache_data['words_train'],\n",
    "                cache_data['words_test'], cache_data['labels_train'], cache_data['labels_test'])\n",
    "    \n",
    "    return words_train, words_test, labels_train, labels_test\n",
    "\n",
    "# Preprocess data\n",
    "words_train, words_test, labels_train, labels_test = preprocess_data(\n",
    "        data_train, data_test, labels_train, labels_test)\n",
    "\n",
    "# Take a look at a sample\n",
    "print(\"\\n--- Raw review ---\")\n",
    "print(data_train[1])\n",
    "print(\"\\n--- Preprocessed words ---\")\n",
    "print(words_train[1])\n",
    "print(\"\\n--- Label ---\")\n",
    "print(labels_train[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Step 3: Extracting Bag-of-Words features <a class=\"anchor\" id=\"step3\"></a>\n",
    "\n",
    "Now that each document has been preprocessed, we can transform each into a Bag-of-Words feature representation. Note that we need to create this transformation based on the training data alone, as we are not allowed to peek at the testing data at all.\n",
    "\n",
    "The dictionary or _vocabulary_ $V$ (set of words shared by documents in the training set) used here will be the one on which we train our supervised learning algorithm. Any future test data must be transformed in the same way for us to be able to apply the learned model for prediction. Hence, it is important to store the transformation / vocabulary as well.\n",
    "\n",
    "> **Note**: The set of words in the training set may not be exactly the same as the test set. What do we do if we encounter a word during testing that we haven't seen before? Unfortunately, we'll have to ignore it, or replace it with a special `<UNK>` token.\n",
    "\n",
    "### Compute Bag-of-Words features\n",
    "\n",
    "We are going to implement the `extract_BoW_features()` function, apply it to both training and test datasets, and store the results in `features_train` and `features_test` Numpy arrays, respectively. We will choose a reasonable vocabulary size, say $|V| = 5000$, and keep only the top $|V|$ occuring words and discard the rest. This number will also serve as the number of columns in the BoW matrices.\n",
    "\n",
    "We will take advantage of `CountVectorizer` from scikit-learn, and pickle the Bag-of-Words transformation so that we can use it in future."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Read features from cache file: bow_features.pkl\n",
      "Vocabulary: 5000 words\n",
      "Sample words: ['lincoln', 'notic', 'card', 'independ', 'rat', 'newer', 'everybodi', 'leon']\n",
      "\n",
      "--- Preprocessed words ---\n",
      "['much', 'look', 'forward', 'movi', 'good', 'famili', 'movi', 'howev', 'michael', 'landon', 'jr', 'edit', 'team', 'better', 'job', 'edit', 'movi', 'would', 'much', 'better', 'mani', 'scene', 'context', 'hope', 'anoth', 'movi', 'seri', 'good', 'anoth', 'one', 'made', 'beg', 'take', 'better', 'care', 'edit', 'stori', 'place', 'seem', 'center', 'unfortun', 'movi', 'seri', 'great', 'enjoy', 'stori', 'willi', 'missi', 'great', 'role', 'model', 'plu', 'romant', 'side', 'viewer', 'alway', 'enjoy', 'good', 'love', 'stori']\n",
      "\n",
      "--- Bag-of-Words features ---\n",
      "[0 0 0 ..., 0 0 0]\n",
      "\n",
      "--- Label ---\n",
      "neg\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "# joblib is an enhanced version of pickle that is more efficient for storing Numpy arrays\n",
    "from sklearn.externals import joblib\n",
    "\n",
    "def extract_BoW_features(words_train, words_test, vocabulary_size=5000,\n",
    "                         cache_dir=cache_dir, cache_file=\"bow_features.pkl\"):\n",
    "    \"\"\"Extract Bag-of-Words for a given set of documents, already preprocessed into words.\"\"\"\n",
    "    \n",
    "    # If cache_file is not None, try to read from it first\n",
    "    cache_data = None\n",
    "    if cache_file is not None:\n",
    "        try:\n",
    "            with open(os.path.join(cache_dir, cache_file), \"rb\") as f:\n",
    "                cache_data = joblib.load(f)\n",
    "            print(\"Read features from cache file:\", cache_file)\n",
    "        except:\n",
    "            pass  # unable to read from cache, but that's okay\n",
    "    \n",
    "    # If cache is missing, then do the heavy lifting\n",
    "    if cache_data is None:\n",
    "        # Fit a vectorizer to training documents and use it to transform them\n",
    "        # Training documents have already been preprocessed and tokenized into words, so\n",
    "        # pass in dummy functions to skip those steps, e.g. preprocessor=lambda x: x\n",
    "        vectorizer = CountVectorizer(max_features=vocabulary_size, preprocessor=lambda x: x, tokenizer=lambda x: x)\n",
    "        # Convert the features using .toarray() for a compact representation\n",
    "        features_train = vectorizer.fit_transform(words_train).toarray()\n",
    "\n",
    "        # Apply the same vectorizer to transform the test documents (ignore unknown words)\n",
    "        features_test = vectorizer.transform(words_test).toarray()\n",
    "        \n",
    "        # Write to cache file for future runs (store vocabulary as well)\n",
    "        if cache_file is not None:\n",
    "            vocabulary = vectorizer.vocabulary_\n",
    "            cache_data = dict(features_train=features_train, features_test=features_test,\n",
    "                             vocabulary=vocabulary)\n",
    "            with open(os.path.join(cache_dir, cache_file), \"wb\") as f:\n",
    "                joblib.dump(cache_data, f)\n",
    "            print(\"Wrote features to cache file:\", cache_file)\n",
    "    else:\n",
    "        # Unpack data loaded from cache file\n",
    "        features_train, features_test, vocabulary = (cache_data['features_train'],\n",
    "                cache_data['features_test'], cache_data['vocabulary'])\n",
    "    \n",
    "    # Return both the extracted features as well as the vocabulary\n",
    "    return features_train, features_test, vocabulary\n",
    "\n",
    "\n",
    "# Extract Bag of Words features for both training and test datasets\n",
    "features_train, features_test, vocabulary = extract_BoW_features(words_train, words_test)\n",
    "\n",
    "# Inspect the vocabulary that was computed\n",
    "print(\"Vocabulary: {} words\".format(len(vocabulary)))\n",
    "\n",
    "import random\n",
    "print(\"Sample words: {}\".format(random.sample(list(vocabulary.keys()), 8)))\n",
    "\n",
    "# Sample\n",
    "print(\"\\n--- Preprocessed words ---\")\n",
    "print(words_train[5])\n",
    "print(\"\\n--- Bag-of-Words features ---\")\n",
    "print(features_train[5])\n",
    "print(\"\\n--- Label ---\")\n",
    "print(labels_train[5])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Let's try to visualize the Bag-of-Words feature vector for one of our training documents."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAEKCAYAAAARnO4WAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n+CIikVDAFxGJhAK+iEgk/j+ONkBdBnyaygAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7bc8909c50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the BoW feature vector for a training document\n",
    "plt.plot(features_train[5,:])\n",
    "plt.xlabel('Word')\n",
    "plt.ylabel('Count')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "#### Comments: Reflecting on Bag-of-Words feature representation\n",
    "\n",
    "The average sparsity level of BoW vectors in our training set (i.e. the percentage of entries in a BoW feature vector that are zero) is ..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The average sparsity level of BoW vectors in training set is 98%\n"
     ]
    }
   ],
   "source": [
    "print (\"The average sparsity level of BoW vectors in training set is {0:.0f}%\".format((np.count_nonzero(features_train==0)*100)/features_train.size))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Zipf's law\n",
    "\n",
    "[Zipf's law](https://en.wikipedia.org/wiki/Zipf%27s_law), named after the famous American linguist George Zipf, is an empirical law stating that given a large collection of documents, the frequency of any word is inversely proportional to its rank in the frequency table. So the most frequent word will occur about twice as often as the second most frequent word, three times as often as the third most frequent word, and so on. The figure below shows the number of appearances of each word in the training set against its rank."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEOCAYAAACTqoDjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7bd0887400>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Find number of occurrences for each word in the training set\n",
    "word_freq = features_train.sum(axis=0)\n",
    "\n",
    "# Sort it in descending order\n",
    "sorted_word_freq = np.sort(word_freq)[::-1]\n",
    "\n",
    "# Plot \n",
    "plt.plot(sorted_word_freq)\n",
    "plt.gca().set_xscale('log')\n",
    "plt.gca().set_yscale('log')\n",
    "plt.xlabel('Rank')\n",
    "plt.ylabel('Number of occurrences')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "#### Comments: Zipf's law\n",
    "\n",
    "The following cell shows the total number of occurrences of the most frequent word and that of the second most frequent word. The numbers do not follow Zipf's law. It must be because of the removal of the stopwords."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of occurrences of the most frequent word: 51696\n",
      "Number of occurrences of the 2nd most frequent word: 48191\n"
     ]
    }
   ],
   "source": [
    "print (\"Number of occurrences of the most frequent word: {}\".format(sorted_word_freq[0]))\n",
    "print (\"Number of occurrences of the 2nd most frequent word: {}\".format(sorted_word_freq[1]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Normalize feature vectors\n",
    "\n",
    "Bag-of-Words features are intuitive to understand as they are simply word counts. But counts can vary a lot, and potentially throw off learning algorithms later in the pipeline. So, before we proceed further, let's normalize the BoW feature vectors to have unit length.\n",
    "\n",
    "This makes sure that each document's representation retains the unique mixture of feature components, but prevents documents with large word counts from dominating those with fewer words."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/trang/anaconda3/lib/python3.6/site-packages/sklearn/utils/validation.py:429: DataConversionWarning: Data with input dtype int64 was converted to float64 by the normalize function.\n",
      "  warnings.warn(msg, _DataConversionWarning)\n"
     ]
    }
   ],
   "source": [
    "import sklearn.preprocessing as pr\n",
    "\n",
    "# Normalize BoW features in training and test set\n",
    "features_train = pr.normalize(features_train)\n",
    "features_test = pr.normalize(features_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Step 4: Classification using BoW features <a class=\"anchor\" id=\"step4\"></a>\n",
    "\n",
    "Now that the data has all been properly transformed, we can feed it into a classifier. To get a baseline model, we train a Naive Bayes classifier from scikit-learn (specifically, [`GaussianNB`](http://scikit-learn.org/stable/modules/generated/sklearn.naive_bayes.GaussianNB.html)), and evaluate its accuracy on the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[GaussianNB] Accuracy: train = 0.81892, test = 0.72452\n"
     ]
    }
   ],
   "source": [
    "from sklearn.naive_bayes import GaussianNB\n",
    "\n",
    "# Train a Guassian Naive Bayes classifier\n",
    "clf1 = GaussianNB()\n",
    "clf1.fit(features_train, labels_train)\n",
    "\n",
    "# Calculate the mean accuracy score on training and test sets\n",
    "print(\"[{}] Accuracy: train = {}, test = {}\".format(\n",
    "        clf1.__class__.__name__,\n",
    "        clf1.score(features_train, labels_train),\n",
    "        clf1.score(features_test, labels_test)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Tree-based algorithms often work quite well on Bag-of-Words as their highly discontinuous and sparse nature is nicely matched by the structure of trees. I will try to improve on the Naive Bayes classifier's performance by using scikit-learn's Gradient-Boosted Decision Tree classifer.\n",
    "\n",
    "### Gradient-Boosted Decision Tree classifier\n",
    "\n",
    "Use [`GradientBoostingClassifier`](http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.GradientBoostingClassifier.html) from scikit-learn to classify the BoW data. This model has a number of parameters. I will use default values for some of them, except one: `n_estimators` which I will tune using grid-search."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 1 candidates, totalling 3 fits\n",
      "[CV] n_estimators=160 ................................................\n",
      "[CV] ................................. n_estimators=160, total= 2.6min\n",
      "[CV] n_estimators=160 ................................................\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:  2.6min remaining:    0.0s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV] ................................. n_estimators=160, total= 2.6min\n",
      "[CV] n_estimators=160 ................................................\n",
      "[CV] ................................. n_estimators=160, total= 2.6min\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   3 out of   3 | elapsed:  7.8min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best n_estimators:  {'n_estimators': 160}\n",
      "[GradientBoostingClassifier] Accuracy: train = 0.86256, test = 0.84244\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "# The best n_estimators is 160\n",
    "# Here is the range on which I performed GridSearchCV: params = {\"n_estimators\": range(10, 200, 10)}\n",
    "# Use only 160 to avoid rerunning gridsearch on the whole range(10,200,10)\n",
    "params = {\"n_estimators\": [160]}\n",
    "\n",
    "def classify_gboost(X_train, X_test, y_train, y_test):        \n",
    "    # Initialize classifier\n",
    "    gbc = GradientBoostingClassifier(learning_rate=1.0, max_depth=1, random_state=0)\n",
    "    \n",
    "    # Perform hyperparameter tuning\n",
    "    grid_search = GridSearchCV(gbc, params, verbose=2)\n",
    "    grid_search.fit(X_train, y_train)\n",
    "    clf = grid_search.best_estimator_\n",
    "    print (\"Best n_estimators: \", grid_search.best_params_)\n",
    "    \n",
    "    # Print final training & test accuracy\n",
    "    print (\"[{}] Accuracy: train = {}, test = {}\".format(\n",
    "    clf.__class__.__name__,\n",
    "    clf.score(X_train, y_train),\n",
    "    clf.score(X_test, labels_test)))\n",
    "    # Return best classifier model\n",
    "    return clf\n",
    "\n",
    "\n",
    "clf2 = classify_gboost(features_train, features_test, labels_train, labels_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Adverserial testing\n",
    "\n",
    "I've written a short movie review to trick the GBDT model. That is, a movie review with a clear positive or negative sentiment that the model will classify incorrectly."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['pos']\n",
      "0.0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/trang/anaconda3/envs/aind/lib/python3.6/site-packages/sklearn/utils/validation.py:429: DataConversionWarning: Data with input dtype int64 was converted to float64 by the normalize function.\n",
      "  warnings.warn(msg, _DataConversionWarning)\n"
     ]
    }
   ],
   "source": [
    "# Write a sample review and set its true sentiment\n",
    "my_review = \"The actor's initial superb performance became average later on\"\n",
    "true_sentiment = 'neg'  # sentiment must be 'pos' or 'neg'\n",
    "\n",
    "# Apply the same preprocessing and vectorizing steps done for the training data\n",
    "my_words = review_to_words(my_review)\n",
    "my_vectorizer = CountVectorizer(vocabulary=vocabulary, preprocessor=lambda x: x, tokenizer=lambda x: x)\n",
    "my_features = my_vectorizer.transform([my_words]).toarray()\n",
    "my_features = pr.normalize(my_features)\n",
    "\n",
    "# Then call the classifier to label it\n",
    "print (clf2.predict(my_features))\n",
    "print (clf2.score(my_features, [true_sentiment]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Step 5: Switching gears - RNNs <a class=\"anchor\" id=\"step5\"></a>\n",
    "\n",
    "We just saw how the task of sentiment analysis can be solved via a traditional machine learning approach: BoW + a nonlinear classifier. We now switch gears and use Recurrent Neural Networks, and in particular LSTMs, to perform sentiment analysis in Keras. Conveniently, Keras has a built-in [IMDb movie reviews dataset](https://keras.io/datasets/#imdb-movie-reviews-sentiment-classification) that we can use, with the same vocabulary size."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded dataset with 25000 training samples, 25000 test samples\n"
     ]
    }
   ],
   "source": [
    "# import the built-in imdb dataset in Keras\n",
    "from keras.datasets import imdb  \n",
    "\n",
    "# Set the vocabulary size\n",
    "vocabulary_size = 5000\n",
    "\n",
    "# Load in training and test data\n",
    "(X_train, y_train), (X_test, y_test) = imdb.load_data(num_words=vocabulary_size)\n",
    "print(\"Loaded dataset with {} training samples, {} test samples\".format(len(X_train), len(X_test)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--- Review ---\n",
      "[1, 4, 2, 716, 4, 65, 7, 4, 689, 4367, 2, 2343, 4804, 2, 2, 2, 2, 2315, 2, 2, 2, 2, 4, 2, 628, 2, 37, 9, 150, 4, 2, 4069, 11, 2909, 4, 2, 847, 313, 6, 176, 2, 9, 2, 138, 9, 4434, 19, 4, 96, 183, 26, 4, 192, 15, 27, 2, 799, 2, 2, 588, 84, 11, 4, 3231, 152, 339, 2, 42, 4869, 2, 2, 345, 4804, 2, 142, 43, 218, 208, 54, 29, 853, 659, 46, 4, 882, 183, 80, 115, 30, 4, 172, 174, 10, 10, 1001, 398, 1001, 1055, 526, 34, 3717, 2, 2, 2, 17, 4, 2, 1094, 871, 64, 85, 22, 2030, 1109, 38, 230, 9, 4, 4324, 2, 251, 2, 1034, 195, 301, 14, 16, 31, 7, 4, 2, 8, 783, 2, 33, 4, 2945, 103, 465, 2, 42, 845, 45, 446, 11, 1895, 19, 184, 76, 32, 4, 2, 207, 110, 13, 197, 4, 2, 16, 601, 964, 2152, 595, 13, 258, 4, 1730, 66, 338, 55, 2, 4, 550, 728, 65, 1196, 8, 1839, 61, 1546, 42, 2, 61, 602, 120, 45, 2, 6, 320, 786, 99, 196, 2, 786, 2, 4, 225, 4, 373, 1009, 33, 4, 130, 63, 69, 72, 1104, 46, 1292, 225, 14, 66, 194, 2, 1703, 56, 8, 803, 1004, 6, 2, 155, 11, 4, 2, 3231, 45, 853, 2029, 8, 30, 6, 117, 430, 19, 6, 2, 9, 15, 66, 424, 8, 2337, 178, 9, 15, 66, 424, 8, 1465, 178, 9, 15, 66, 142, 15, 9, 424, 8, 28, 178, 662, 44, 12, 17, 4, 130, 898, 1686, 9, 6, 2, 267, 185, 430, 4, 118, 2, 277, 15, 4, 1188, 100, 216, 56, 19, 4, 357, 114, 2, 367, 45, 115, 93, 788, 121, 4, 2, 79, 32, 68, 278, 39, 8, 818, 162, 4165, 237, 600, 7, 98, 306, 8, 157, 549, 628, 11, 6, 2, 13, 824, 15, 4104, 76, 42, 138, 36, 774, 77, 1059, 159, 150, 4, 229, 497, 8, 1493, 11, 175, 251, 453, 19, 2, 189, 12, 43, 127, 6, 394, 292, 7, 2, 4, 107, 8, 4, 2826, 15, 1082, 1251, 9, 906, 42, 1134, 6, 66, 78, 22, 15, 13, 244, 2519, 8, 135, 233, 52, 44, 10, 10, 466, 112, 398, 526, 34, 4, 1572, 4413, 2, 1094, 225, 57, 599, 133, 225, 6, 227, 7, 541, 4323, 6, 171, 139, 7, 539, 2, 56, 11, 6, 3231, 21, 164, 25, 426, 81, 33, 344, 624, 19, 6, 4617, 7, 2, 2, 6, 2, 4, 22, 9, 1082, 629, 237, 45, 188, 6, 55, 655, 707, 2, 956, 225, 1456, 841, 42, 1310, 225, 6, 2493, 1467, 2, 2828, 21, 4, 2, 9, 364, 23, 4, 2228, 2407, 225, 24, 76, 133, 18, 4, 189, 2293, 10, 10, 814, 11, 2, 11, 2642, 14, 47, 15, 682, 364, 352, 168, 44, 12, 45, 24, 913, 93, 21, 247, 2441, 4, 116, 34, 35, 1859, 8, 72, 177, 9, 164, 8, 901, 344, 44, 13, 191, 135, 13, 126, 421, 233, 18, 259, 10, 10, 4, 2, 2, 4, 2, 3074, 7, 112, 199, 753, 357, 39, 63, 12, 115, 2, 763, 8, 15, 35, 3282, 1523, 65, 57, 599, 6, 1916, 277, 1730, 37, 25, 92, 202, 6, 2, 44, 25, 28, 6, 22, 15, 122, 24, 4171, 72, 33, 32]\n",
      "--- Label ---\n",
      "0\n"
     ]
    }
   ],
   "source": [
    "# Inspect a sample review and its label\n",
    "print(\"--- Review ---\")\n",
    "print(X_train[7])\n",
    "print(\"--- Label ---\")\n",
    "print(y_train[7])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "As seen above, the label is an integer (0 for negative, 1 for positive), and the review itself is stored as a sequence of integers. These are word IDs that have been preassigned to individual words. To map them back to the original words, we can use the dictionary returned by `imdb.get_word_index()`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--- Review (with words) ---\n",
      "['the', 'of', 'and', 'local', 'of', 'their', 'br', 'of', 'attention', 'widow', 'and', 'captures', 'parties', 'and', 'and', 'and', 'and', 'excitement', 'and', 'and', 'and', 'and', 'of', 'and', 'english', 'and', 'like', 'it', 'years', 'of', 'and', 'unintentional', 'this', 'hitchcock', 'of', 'and', 'learn', 'everyone', 'is', 'quite', 'and', 'it', 'and', 'such', 'it', 'bonus', 'film', 'of', 'too', 'seems', 'he', 'of', 'enough', 'for', 'be', 'and', 'editing', 'and', 'and', 'please', 'great', 'this', 'of', 'shoots', 'thing', '3', 'and', \"it's\", 'mentioning', 'and', 'and', 'given', 'parties', 'and', 'back', 'out', 'interesting', 'times', 'no', 'all', 'average', 'talking', 'some', 'of', 'nor', 'seems', 'into', 'best', 'at', 'of', 'every', 'cast', 'i', 'i', 'inside', 'keep', 'inside', 'large', 'viewer', 'who', 'obscure', 'and', 'and', 'and', 'movie', 'of', 'and', 'entirely', \"you've\", 'see', 'because', 'you', 'deals', 'successful', 'her', 'anything', 'it', 'of', 'dedicated', 'and', 'hard', 'and', 'further', \"that's\", 'takes', 'as', 'with', 'by', 'br', 'of', 'and', 'in', 'minute', 'and', 'they', 'of', 'westerns', 'watch', 'seemed', 'and', \"it's\", 'lee', 'if', 'oh', 'this', 'japan', 'film', 'around', 'get', 'an', 'of', 'and', 'always', 'life', 'was', 'between', 'of', 'and', 'with', 'group', 'rate', 'code', \"film's\", 'was', 'although', 'of', 'arts', 'had', 'death', 'time', 'and', 'of', 'anyway', 'romantic', 'their', 'won', 'in', 'kevin', 'only', 'flying', \"it's\", 'and', 'only', 'cut', 'show', 'if', 'and', 'is', 'star', 'stay', 'movies', 'both', 'and', 'stay', 'and', 'of', 'music', 'of', 'tell', 'missing', 'they', 'of', 'here', 'really', 'me', 'we', 'value', 'some', 'silent', 'music', 'as', 'had', 'thought', 'and', 'realized', 'she', 'in', 'sorry', 'reasons', 'is', 'and', '10', 'this', 'of', 'and', 'shoots', 'if', 'average', 'remembered', 'in', 'at', 'is', 'over', 'worse', 'film', 'is', 'and', 'it', 'for', 'had', 'absolutely', 'in', 'naive', 'want', 'it', 'for', 'had', 'absolutely', 'in', 'j', 'want', 'it', 'for', 'had', 'back', 'for', 'it', 'absolutely', 'in', 'one', 'want', 'shots', 'has', 'that', 'movie', 'of', 'here', 'write', 'whatsoever', 'it', 'is', 'and', 'set', 'got', 'worse', 'of', 'where', 'and', 'once', 'for', 'of', 'accent', 'after', 'saw', 'she', 'film', 'of', 'rest', 'little', 'and', 'camera', 'if', 'best', 'way', 'elements', 'know', 'of', 'and', 'also', 'an', 'were', 'sense', 'or', 'in', 'realistic', 'actually', 'satan', \"he's\", 'score', 'br', 'any', 'himself', 'in', 'another', 'type', 'english', 'this', 'is', 'and', 'was', 'tom', 'for', 'dating', 'get', \"it's\", 'such', 'from', 'fantastic', 'will', 'pace', 'new', 'years', 'of', 'guy', 'game', 'in', 'murders', 'this', 'us', 'hard', 'lives', 'film', 'and', 'fact', 'that', 'out', 'end', 'is', 'getting', 'together', 'br', 'and', 'of', 'seen', 'in', 'of', 'jail', 'for', 'sees', 'utterly', 'it', 'meet', \"it's\", 'depth', 'is', 'had', 'do', 'you', 'for', 'was', 'rather', 'convince', 'in', 'why', 'last', 'very', 'has', 'i', 'i', 'throughout', 'never', 'keep', 'viewer', 'who', 'of', 'becoming', 'switch', 'and', 'entirely', 'music', 'even', 'interest', 'scene', 'music', 'is', 'far', 'br', 'voice', 'riveting', 'is', 'again', 'something', 'br', 'decent', 'and', 'she', 'this', 'is', 'shoots', 'not', 'director', 'have', 'against', 'people', 'they', 'line', 'cinematography', 'film', 'is', 'couples', 'br', 'and', 'and', 'is', 'and', 'of', 'you', 'it', 'sees', 'hero', \"he's\", 'if', \"can't\", 'is', 'time', 'husband', 'silly', 'and', 'result', 'music', 'image', 'sequences', \"it's\", 'chase', 'music', 'is', 'veteran', 'include', 'and', 'freeman', 'not', 'of', 'and', 'it', 'along', 'are', 'of', 'hearing', 'cutting', 'music', 'his', 'get', 'scene', 'but', 'of', 'fact', 'correct', 'i', 'i', 'means', 'this', 'and', 'this', 'blockbuster', 'as', 'there', 'for', 'disappointed', 'along', 'wrong', 'few', 'has', 'that', 'if', 'his', 'weird', 'way', 'not', 'girl', 'display', 'of', 'love', 'who', 'so', 'friendship', 'in', 'we', 'down', 'it', 'director', 'in', 'situation', 'line', 'has', 'was', 'big', 'why', 'was', 'your', 'supposed', 'last', 'but', 'especially', 'i', 'i', 'of', 'and', 'and', 'of', 'and', 'internet', 'br', 'never', 'give', 'theme', 'rest', 'or', 'really', 'that', 'best', 'and', 'release', 'in', 'for', 'so', 'multi', 'random', 'their', 'even', 'interest', 'is', 'judge', 'once', 'arts', 'like', 'have', 'then', 'own', 'is', 'and', 'has', 'have', 'one', 'is', 'you', 'for', 'off', 'his', 'dutch', 'we', 'they', 'an']\n",
      "--- Label ---\n",
      "0\n"
     ]
    }
   ],
   "source": [
    "# Map word IDs back to words\n",
    "word2id = imdb.get_word_index()\n",
    "id2word = {i: word for word, i in word2id.items()}\n",
    "print(\"--- Review (with words) ---\")\n",
    "print([id2word.get(i, \" \") for i in X_train[7]])\n",
    "print(\"--- Label ---\")\n",
    "print(y_train[7])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Unlike our Bag-of-Words approach, where we simply summarized the counts of each word in a document, this representation essentially retains the entire sequence of words (minus punctuation, stopwords, etc.). This is critical for RNNs to function. But it also means that now the features can be of different lengths!\n",
    "\n",
    "#### Comments: Variable length reviews\n",
    "\n",
    "Below are the maximum review length (in terms of number of words) and the minimum in the training set:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2494\n",
      "11\n"
     ]
    }
   ],
   "source": [
    "# The review that has the max length\n",
    "longest_review = max(X_train, key=len)\n",
    "\n",
    "# The length of the longest review\n",
    "print (len(longest_review))\n",
    "\n",
    "# The length of the shortest review\n",
    "print (len(min(X_train, key=len)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Pad sequences\n",
    "\n",
    "In order to feed this data into an RNN, all input documents must have the same length. Let's limit the maximum review length to `max_words` by truncating longer reviews and padding shorter reviews with a null value (0). We can accomplish this easily using the [`pad_sequences()`](https://keras.io/preprocessing/sequence/#pad_sequences) function in Keras. For now, set `max_words` to 500."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "from keras.preprocessing import sequence\n",
    "\n",
    "# Set the maximum number of words per document (for both training and testing)\n",
    "max_words = 500\n",
    "\n",
    "# Pad sequences in X_train and X_test\n",
    "X_train = sequence.pad_sequences(X_train, maxlen=max_words)\n",
    "X_test = sequence.pad_sequences(X_test, maxlen=max_words)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Design an RNN model for sentiment analysis\n",
    "\n",
    "I am going to build an RNN using LSTM. The input is a sequence of words (technically, integer word IDs) of maximum length = `max_words`, and the output is a binary sentiment label (0 or 1)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_3 (Embedding)      (None, 500, 300)          1500000   \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 500, 300)          0         \n",
      "_________________________________________________________________\n",
      "lstm_3 (LSTM)                (None, 512)               1665024   \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 1)                 513       \n",
      "=================================================================\n",
      "Total params: 3,165,537\n",
      "Trainable params: 3,165,537\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Embedding, LSTM, Dense, Dropout\n",
    "from keras.optimizers import Adam\n",
    "\n",
    "learning_rate = 0.001\n",
    "embed_dim = 300\n",
    "model = Sequential()\n",
    "model.add(Embedding(vocabulary_size, embed_dim, input_length=max_words))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(LSTM(512, dropout=0.2))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Train and evaluate the RNN model\n",
    "\n",
    "Now we are ready to train the model. In Keras world, first we need to _compile_ the model by specifying the loss function and optimizer we want to use while training, as well as any evaluation metrics we'd like to measure."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# Compile the model, specifying a loss function, optimizer, and metrics\n",
    "model.compile(loss='binary_crossentropy', optimizer=Adam(learning_rate), metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Once compiled, we can kick off the training process. There are two important training parameters that we have to specify - **batch size** and **number of training epochs**, which together with the model architecture determine the total training time. I will use the typical value of 32 for `batch_size`. I will select a large number of epochs, 50, but will also use `EarlyStopping` so that the training will stop if validation loss doesn't fall for a number consecutive epochs. This technique usually saves time from training the model for too long without any improvement.\n",
    "\n",
    "I will also split off a small portion of the training set to be used for validation during training. This will help monitor the training process and identify potential overfitting. I will use `validation_split` - a fraction of the training data for Keras to set aside for validation purpose. Validation metrics are evaluated once at the end of each epoch."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 20000 samples, validate on 5000 samples\n",
      "Epoch 1/50\n",
      "Epoch 00001: val_loss improved from inf to 0.60519, saving model to cache/sentiment_analysis/rnn_model.h5\n",
      " - 980s - loss: 0.5727 - acc: 0.6956 - val_loss: 0.6052 - val_acc: 0.6524\n",
      "Epoch 2/50\n",
      "Epoch 00002: val_loss improved from 0.60519 to 0.28767, saving model to cache/sentiment_analysis/rnn_model.h5\n",
      " - 972s - loss: 0.3655 - acc: 0.8397 - val_loss: 0.2877 - val_acc: 0.8814\n",
      "Epoch 3/50\n",
      "Epoch 00003: val_loss improved from 0.28767 to 0.28130, saving model to cache/sentiment_analysis/rnn_model.h5\n",
      " - 973s - loss: 0.2248 - acc: 0.9113 - val_loss: 0.2813 - val_acc: 0.8926\n",
      "Epoch 4/50\n",
      "Epoch 00004: val_loss did not improve\n",
      " - 973s - loss: 0.1671 - acc: 0.9370 - val_loss: 0.2948 - val_acc: 0.8864\n",
      "Epoch 5/50\n",
      "Epoch 00005: val_loss did not improve\n",
      " - 973s - loss: 0.1296 - acc: 0.9533 - val_loss: 0.3350 - val_acc: 0.8840\n",
      "Epoch 6/50\n",
      "Epoch 00006: val_loss did not improve\n",
      " - 973s - loss: 0.0937 - acc: 0.9677 - val_loss: 0.4817 - val_acc: 0.8638\n"
     ]
    }
   ],
   "source": [
    "from keras.callbacks import ModelCheckpoint, EarlyStopping\n",
    "\n",
    "# Specify training parameters: batch size and number of epochs\n",
    "batch_size = 32\n",
    "num_epochs = 50\n",
    "\n",
    "# Save the model, so that we can quickly load it in future (and perhaps resume training)\n",
    "model_file = \"rnn_model.h5\"  # HDF5 file\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath=os.path.join(cache_dir, model_file), verbose=1, \n",
    "                               save_best_only=True)\n",
    "# Stop training if the validation loss doesn't fall for 3 consecutive epochs\n",
    "earlystopping = EarlyStopping(monitor='val_loss', min_delta=0, patience=3, verbose=0, mode='auto')\n",
    "# Train the model, reserve some training data for validation\n",
    "hist = model.fit(X_train, y_train, batch_size=batch_size, epochs=num_epochs,\n",
    "          validation_split=0.2, callbacks=[checkpointer, earlystopping], \n",
    "          verbose=2, shuffle=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# Later we can load it using keras.models.load_model()\n",
    "from keras.models import load_model\n",
    "model = load_model(os.path.join(cache_dir, model_file))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Once we have trained the model, it's time to see how well it performs on unseen test data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test accuracy: 0.88288\n"
     ]
    }
   ],
   "source": [
    "# Evaluate the model on the test set\n",
    "scores = model.evaluate(X_test, y_test, verbose=0)  # returns loss and other metrics specified in model.compile()\n",
    "print(\"Test accuracy:\", scores[1])  # scores[1] should correspond to accuracy that I passed in metrics=['accuracy']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Step 6: Evaluation <a class=\"anchor\" id=\"step6\"></a>\n",
    "\n",
    "### Comparing RNNs and Traditional Methods\n",
    "\n",
    "The BoW + Gradient-Boosted Decision Trees (GBDT) model does a better job than the BoW + GaussianNB model. The RNN model outperforms both models. LSTM is able to remember over long sequences of inputs while the BoW models disregard word order. As seen in the adverserial testing part of BoW + GBDT, this limitation must have caused the model to classify the tricky review incorrectly.\n",
    "\n",
    "### Steps taken to improve the accuracy of the RNN model\n",
    "\n",
    "Below is the initial attempt that yielded an accuracy of 0.86792:\n",
    "\n",
    "```\n",
    "learning_rate = 0.01\n",
    "embed_dim = 128\n",
    "model = Sequential()\n",
    "model.add(Embedding(vocabulary_size, embed_dim, input_length=max_words))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(LSTM(128, dropout=0.2))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "```\n",
    "\n",
    "`Dropout` helped reduce overfitting. \n",
    "\n",
    "I then did the following experiments:\n",
    "\n",
    "* Reduced the `learning_rate`. Higher learning rate, 0.01, must have caused the optimizer to overshoot the minimum. Lower rate made training more reliable, but also slower.\n",
    "* Tried various values for `embed_dim` (64, 100, 128, and 300). 300 yielded the best validation and test accuracies, although at the cost of training time.\n",
    "* Tried a) 1 LSTM layer (128 units) and b) 3 LSTM layers (32 units each). a) showed a better performance.\n",
    "* Tried GRU instead of LSTM and the performance was similar.\n",
    "\n",
    "After the experiments, the final model (trained above) achieved an accuracy of 0.88288.\n",
    "\n",
    "```\n",
    "learning_rate = 0.001\n",
    "embed_dim = 300\n",
    "model = Sequential()\n",
    "model.add(Embedding(vocabulary_size, embed_dim, input_length=max_words))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(LSTM(512, dropout=0.2))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "```"
   ]
  }
 ],
 "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}