U5.06 - [TALLER] - Padding, Masking - Sentiment Analysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## The aim of this lab is to build a system for sentiment analysis on a dataset of tweets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!wget --no-cache -O init.py -q https://raw.githubusercontent.com/rramosp/2020.deeplearning/master/init.py\n",
    "from init import init; init(force_download=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "if 'google.colab' in sys.modules:\n",
    "    print (\"setting tensorflow version in colab\")\n",
    "    %tensorflow_version 2.x\n",
    "import tensorflow as tf\n",
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The data consist on passenger's reviews of U.S. airlines: https://www.kaggle.com/crowdflower/twitter-airline-sentiment "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv('local/data/Tweets.csv')\n",
    "# Keeping only the neccessary columns\n",
    "data = data[['text','airline_sentiment']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>airline_sentiment</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>@VirginAmerica What @dhepburn said.</td>\n",
       "      <td>neutral</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>@VirginAmerica plus you've added commercials t...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>@VirginAmerica I didn't today... Must mean I n...</td>\n",
       "      <td>neutral</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>@VirginAmerica it's really aggressive to blast...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>@VirginAmerica and it's a really big bad thing...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14635</th>\n",
       "      <td>@AmericanAir thank you we got on a different f...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14636</th>\n",
       "      <td>@AmericanAir leaving over 20 minutes Late Flig...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14637</th>\n",
       "      <td>@AmericanAir Please bring American Airlines to...</td>\n",
       "      <td>neutral</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14638</th>\n",
       "      <td>@AmericanAir you have my money, you change my ...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14639</th>\n",
       "      <td>@AmericanAir we have 8 ppl so we need 2 know h...</td>\n",
       "      <td>neutral</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>14640 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    text airline_sentiment\n",
       "0                    @VirginAmerica What @dhepburn said.           neutral\n",
       "1      @VirginAmerica plus you've added commercials t...          positive\n",
       "2      @VirginAmerica I didn't today... Must mean I n...           neutral\n",
       "3      @VirginAmerica it's really aggressive to blast...          negative\n",
       "4      @VirginAmerica and it's a really big bad thing...          negative\n",
       "...                                                  ...               ...\n",
       "14635  @AmericanAir thank you we got on a different f...          positive\n",
       "14636  @AmericanAir leaving over 20 minutes Late Flig...          negative\n",
       "14637  @AmericanAir Please bring American Airlines to...           neutral\n",
       "14638  @AmericanAir you have my money, you change my ...          negative\n",
       "14639  @AmericanAir we have 8 ppl so we need 2 know h...           neutral\n",
       "\n",
       "[14640 rows x 2 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2363\n",
      "9178\n"
     ]
    }
   ],
   "source": [
    "import re\n",
    "#Remove neutral class\n",
    "data = data[data.airline_sentiment != \"neutral\"]\n",
    "\n",
    "#text normalization\n",
    "data['text'] = data['text'].apply(lambda x: x.lower())\n",
    "data['text'] = data['text'].apply((lambda x:re.sub('@[^\\s]+','',x)))#remove the name of the airline\n",
    "data['text'] = data['text'].apply((lambda x: re.sub('[^a-zA-z0-9\\s]','',x)))\n",
    "print(np.sum(data['airline_sentiment'].values == 'positive'))\n",
    "print(np.sum(data['airline_sentiment'].values == 'negative'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>airline_sentiment</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>plus youve added commercials to the experienc...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>its really aggressive to blast obnoxious ente...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>and its a really big bad thing about it</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>seriously would pay 30 a flight for seats tha...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>yes nearly every time i fly vx this ear worm ...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14633</th>\n",
       "      <td>my flight was cancelled flightled leaving tom...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14634</th>\n",
       "      <td>right on cue with the delays</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14635</th>\n",
       "      <td>thank you we got on a different flight to chi...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14636</th>\n",
       "      <td>leaving over 20 minutes late flight no warnin...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14638</th>\n",
       "      <td>you have my money you change my flight and do...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>11541 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    text airline_sentiment\n",
       "1       plus youve added commercials to the experienc...          positive\n",
       "3       its really aggressive to blast obnoxious ente...          negative\n",
       "4                and its a really big bad thing about it          negative\n",
       "5       seriously would pay 30 a flight for seats tha...          negative\n",
       "6       yes nearly every time i fly vx this ear worm ...          positive\n",
       "...                                                  ...               ...\n",
       "14633   my flight was cancelled flightled leaving tom...          negative\n",
       "14634                       right on cue with the delays          negative\n",
       "14635   thank you we got on a different flight to chi...          positive\n",
       "14636   leaving over 20 minutes late flight no warnin...          negative\n",
       "14638   you have my money you change my flight and do...          negative\n",
       "\n",
       "[11541 rows x 2 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "for idx,row in data.iterrows():\n",
    "    row[0] = row[0].replace('rt',' ')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>airline_sentiment</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>plus youve added commercials to the experienc...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>its really aggressive to blast obnoxious ente...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>and its a really big bad thing about it</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>seriously would pay 30 a flight for seats tha...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>yes nearly every time i fly vx this ear worm ...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14633</th>\n",
       "      <td>my flight was cancelled flightled leaving tom...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14634</th>\n",
       "      <td>right on cue with the delays</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14635</th>\n",
       "      <td>thank you we got on a different flight to chi...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14636</th>\n",
       "      <td>leaving over 20 minutes late flight no warnin...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14638</th>\n",
       "      <td>you have my money you change my flight and do...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>11541 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    text airline_sentiment\n",
       "1       plus youve added commercials to the experienc...          positive\n",
       "3       its really aggressive to blast obnoxious ente...          negative\n",
       "4                and its a really big bad thing about it          negative\n",
       "5       seriously would pay 30 a flight for seats tha...          negative\n",
       "6       yes nearly every time i fly vx this ear worm ...          positive\n",
       "...                                                  ...               ...\n",
       "14633   my flight was cancelled flightled leaving tom...          negative\n",
       "14634                       right on cue with the delays          negative\n",
       "14635   thank you we got on a different flight to chi...          positive\n",
       "14636   leaving over 20 minutes late flight no warnin...          negative\n",
       "14638   you have my money you change my flight and do...          negative\n",
       "\n",
       "[11541 rows x 2 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package punkt to /home/julian/nltk_data...\n",
      "[nltk_data]   Package punkt is already up-to-date!\n",
      "[nltk_data] Downloading package stopwords to /home/julian/nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import nltk\n",
    "nltk.download('punkt')\n",
    "nltk.download('stopwords')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Removing Stop Words\n",
    "from nltk.corpus import stopwords\n",
    "all_sentences = data['text'].values\n",
    "all_words = [nltk.word_tokenize(sent) for sent in all_sentences]\n",
    "stop_words = stopwords.words('english')\n",
    "stop_words.append('')\n",
    "\n",
    "for i in range(len(all_words)):  \n",
    "    all_words[i] = [w for w in all_words[i] if w not in stop_words]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Exercise 1\n",
    "\n",
    "all_words is a list with all the tweets that are going to be used to train the model. Tokenize the tweets using a dictionary of 2000 words (remeber that position 0 is reserved). Once the sentences are tokenized, take into account that the length of every tweet is different so before they can be passed to the training step, the tweets must be **padded** in order to provide them with equal length.\n",
    "\n",
    "Review the padding function in the preprocessing module of keras and apply it to the tokenized sequences. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.preprocessing.text import Tokenizer\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Exercise 2\n",
    "\n",
    "The previous step add 0's to some tweets in order to provide them with the same length. Now it is necessary to define a model that be able to discard those 0's. Review the masking layer and masking option of the embedding layer of keras. \n",
    "\n",
    "Define a LSTM architecture to classify the tweets as \"negative\" or \"positive\". Use the Embedding layer and its masking option to discard the 0's added during padding step. Evaluate the performance of the model for embed_dim = [32,64,128] and a LSTM layer with cells = [32,64,128]. Use 20% of the data for testing purposes and 10 epochs for training. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "X_tr, X_te, y_tr, y_te = train_test_split(X, Y, test_size=0.2, random_state=2018)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You should get the following results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x216 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best accuracy= 0.8973581637072325\n"
     ]
    }
   ],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Exercise 3\n",
    "\n",
    "Train a CBOW model with the dataset using and Embedding dimension of [32,64,128]. Create a new network and transfer the pretrained weights to the Embedding layer. Use the same architecture than before (except for the Embedding layer). Compare the results.\n",
    "\n",
    "**Note**: Take care on the tokenization of the words. Keras tokenizer does not assign the zero value to any word because of padding purposes. Make sure that the order of the vectors in the CBOW embedding matrix corresponds with the indexs used to represent the words in the data. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You should get a small improvement:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x216 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best accuracy= 0.9034213945430922\n"
     ]
    }
   ],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Exercise 4\n",
    "\n",
    "Load pretrained GloVe Embeddings [glove.6B.zip](http://nlp.stanford.edu/data/glove.6B.zip). Create a new network and transfer the pretrained weights to the Embedding layer. Use the same architecture than before (except for the Embedding layer). Compare the results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!wget http://nlp.stanford.edu/data/glove.6B.zip\n",
    "!unzip glove.6B.zip     "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You should get the following results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x216 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best accuracy= 0.9086184495452577\n"
     ]
    }
   ],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Exercise 5\n",
    "\n",
    "Use a CNN architeture (Conv1D) instead of the LSTM one. Use [6,12,24] filters and define the kernel to use trigrams. You can use the GloVe embedding weights used for the former excercise. Compare the results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "#hint:\n",
    "\n",
    "def Model_Sentiment(Embeb,filters):\n",
    "    model = Sequential([\n",
    "        Embedding(input_dim=max_fatures, output_dim=Embeb, weights = [wordsmatrix], trainable = False, mask_zero=True),\n",
    "        Conv1D(filters,kernel_size=?????,activation='relu',padding='same'),\n",
    "        GlobalMaxPooling1D(),\n",
    "        #Complete the model\n",
    "    ])\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You should get the following results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x216 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best accuracy= 0.9047206582936336\n"
     ]
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}