FakeNewsClassifier_LSTM.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Importing Necessary Modules"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "import numpy as np\n",
    "import pandas as pd \n",
    "import torch\n",
    "from torch import nn\n",
    "import string\n",
    "import nltk\n",
    "from collections import Counter\n",
    "from sklearn.model_selection import train_test_split\n",
    "from torch.utils.data import DataLoader, Dataset\n",
    "from torch.autograd import Variable\n",
    "from tqdm import tqdm\n",
    "import ast\n",
    "import matplotlib.pyplot as plt\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## About Data\n",
    "\n",
    "Data Source: https://www.kaggle.com/datasets/saurabhshahane/fake-news-classification\n",
    "\n",
    "Data contains about 72,000 articles with around 35,000 real news and around 37,000 fake news.\n",
    "\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load DataSet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded 72134 rows\n"
     ]
    }
   ],
   "source": [
    "data = pd.read_csv(\"WELFake_Dataset.csv\", low_memory=False).drop(\"Unnamed: 0\", axis=1)\n",
    "print(f\"Loaded {data.shape[0]} rows\")"
   ]
  },
  {
   "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>title</th>\n",
       "      <th>text</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>LAW ENFORCEMENT ON HIGH ALERT Following Threat...</td>\n",
       "      <td>No comment is expected from Barack Obama Membe...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>NaN</td>\n",
       "      <td>Did they post their votes for Hillary already?</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>UNBELIEVABLE! OBAMA’S ATTORNEY GENERAL SAYS MO...</td>\n",
       "      <td>Now, most of the demonstrators gathered last ...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Bobby Jindal, raised Hindu, uses story of Chri...</td>\n",
       "      <td>A dozen politically active pastors came here f...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>SATAN 2: Russia unvelis an image of its terrif...</td>\n",
       "      <td>The RS-28 Sarmat missile, dubbed Satan 2, will...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               title  \\\n",
       "0  LAW ENFORCEMENT ON HIGH ALERT Following Threat...   \n",
       "1                                                NaN   \n",
       "2  UNBELIEVABLE! OBAMA’S ATTORNEY GENERAL SAYS MO...   \n",
       "3  Bobby Jindal, raised Hindu, uses story of Chri...   \n",
       "4  SATAN 2: Russia unvelis an image of its terrif...   \n",
       "\n",
       "                                                text  label  \n",
       "0  No comment is expected from Barack Obama Membe...      1  \n",
       "1     Did they post their votes for Hillary already?      1  \n",
       "2   Now, most of the demonstrators gathered last ...      1  \n",
       "3  A dozen politically active pastors came here f...      0  \n",
       "4  The RS-28 Sarmat missile, dubbed Satan 2, will...      1  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## NaN Values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "title    558\n",
       "text      39\n",
       "label      0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.isna().sum()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There is very few rows containing NaN values, so dropping all of them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Samples before removing NaNs: 72134\n",
      "Samples after removing NaNs: 71537\n"
     ]
    }
   ],
   "source": [
    "print(f\"Samples before removing NaNs: {data.shape[0]}\")\n",
    "data.dropna(inplace=True)\n",
    "print(f\"Samples after removing NaNs: {data.shape[0]}\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Preprocessing"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Creating a new columns named feature, concatenating title and text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>label</th>\n",
       "      <th>feature</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>LAW ENFORCEMENT ON HIGH ALERT Following Threat...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>UNBELIEVABLE! OBAMA’S ATTORNEY GENERAL SAYS MO...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>Bobby Jindal, raised Hindu, uses story of Chri...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>SATAN 2: Russia unvelis an image of its terrif...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1</td>\n",
       "      <td>About Time! Christian Group Sues Amazon and SP...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   label                                            feature\n",
       "0      1  LAW ENFORCEMENT ON HIGH ALERT Following Threat...\n",
       "2      1  UNBELIEVABLE! OBAMA’S ATTORNEY GENERAL SAYS MO...\n",
       "3      0  Bobby Jindal, raised Hindu, uses story of Chri...\n",
       "4      1  SATAN 2: Russia unvelis an image of its terrif...\n",
       "5      1  About Time! Christian Group Sues Amazon and SP..."
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['feature'] = data['title'] + \" \" + data['text']\n",
    "data.drop(['title', 'text'], axis=1, inplace=True)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def remove_puncts_and_censored_words(text: str) -> str:\n",
    "    \"\"\"Removes censored words and punctuations from the text\n",
    "\n",
    "    Args:\n",
    "        text (str): news feature columm\n",
    "\n",
    "    Returns:\n",
    "        str: preprocessed text\n",
    "    \"\"\"\n",
    "    # Remove censored words\n",
    "    censor_removed = \" \".join([word for word in text.split(\" \") if '*' not in word])\n",
    "    # Remove punctuations\n",
    "    punct_free = \"\".join([letter for letter in censor_removed if letter not in string.punctuation])\n",
    "\n",
    "    # Replace multiple spaces with one space and lower them\n",
    "    single_spaced = \" \".join(filter(bool, punct_free.split(\" \"))).lower()\n",
    "    return single_spaced\n",
    "\n",
    "def word_tokenize(text: str) -> list[str]:\n",
    "    \"\"\"Splits text into words\n",
    "\n",
    "    Args:\n",
    "        text (str): preprocessed text\n",
    "\n",
    "    Returns:\n",
    "        list(str): word tokens\n",
    "    \"\"\"\n",
    "    return text.split(\" \")\n",
    "\n",
    "def remove_stopwords(tokens: list[str]) -> list[str]:\n",
    "    \"\"\"Removes stopwords from the tokens\n",
    "\n",
    "    Args:\n",
    "        tokens (list[str]): word tokens\n",
    "\n",
    "    Returns:\n",
    "        list[str]: word tokens with stop words removed\n",
    "    \"\"\"\n",
    "    stopwords = nltk.corpus.stopwords.words('english')\n",
    "    return [t for t in tokens if t not in stopwords]\n",
    "\n",
    "def preprocess(text: str) ->list[str]:\n",
    "    \"\"\"Preprocesses the text\n",
    "\n",
    "    Args:\n",
    "        text (str): news feature column\n",
    "\n",
    "    Returns:\n",
    "        list[str]: preprocessed tokens\n",
    "    \"\"\"\n",
    "    preprocessed = remove_puncts_and_censored_words(text)\n",
    "    preprocessed = word_tokenize(preprocessed)\n",
    "    preprocessed = remove_stopwords(preprocessed)\n",
    "\n",
    "    return preprocessed\n",
    "\n",
    "def pad(text: list[int], pad_length: int) -> list[int]:\n",
    "    \"\"\"If the text is less than pad_length, pads the text with 0s at the beginning.\n",
    "    If the text is greater than pad_length, truncates the text to pad_length.\n",
    "    If the text is equal to pad_length, returns the text as it is.\n",
    "\n",
    "    Args:\n",
    "        text (list[int]): preprocessed tokens\n",
    "        pad_length (int): length to pad the text to\n",
    "\n",
    "    Returns:\n",
    "        list[int]: padded text\n",
    "    \"\"\"\n",
    "    if len(text) < pad_length:\n",
    "        pad_arr = [0] * (pad_length - len(text))\n",
    "        return pad_arr + text\n",
    "    elif len(text) > pad_length:\n",
    "        return text[:pad_length]\n",
    "    else:\n",
    "        return text"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Apply Preprocessing\n",
    "\n",
    "1. Removing censored words and punctuations\n",
    "2. Word Tokenizing\n",
    "3. Removing Stopwords"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>label</th>\n",
       "      <th>preprocessed</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>[law, enforcement, high, alert, following, thr...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>[unbelievable, obama’s, attorney, general, say...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>[bobby, jindal, raised, hindu, uses, story, ch...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>[satan, 2, russia, unvelis, image, terrifying,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1</td>\n",
       "      <td>[time, christian, group, sues, amazon, splc, d...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   label                                       preprocessed\n",
       "0      1  [law, enforcement, high, alert, following, thr...\n",
       "2      1  [unbelievable, obama’s, attorney, general, say...\n",
       "3      0  [bobby, jindal, raised, hindu, uses, story, ch...\n",
       "4      1  [satan, 2, russia, unvelis, image, terrifying,...\n",
       "5      1  [time, christian, group, sues, amazon, splc, d..."
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['preprocessed'] = data['feature'].apply(lambda x: preprocess(x))\n",
    "data.drop('feature', axis=1,inplace=True)\n",
    "data.head()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Converting word string tokens into integers\n",
    "\n",
    "1. Create a dictionary of all words in data with its count\n",
    "2. Filter out words with only 1 occurence\n",
    "3. Create a vocabulary of only unique words\n",
    "4. Convert word tokens into integer using index of word in vocabulary\n",
    "5. Pad work tokens into length of 300"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# List of all words in feature preprocessed column\n",
    "words = [word for row in data['preprocessed'] for word in row]\n",
    "\n",
    "# Create a dictionary of words and their counts\n",
    "word_counts = Counter(words)\n",
    "\n",
    "# Creating a vocabulary dictionary of words which occur more than once\n",
    "vocabulary = {word: index + 1 for index, (word, count) in enumerate(word_counts.most_common()) if count > 1} \n",
    "\n",
    "words = sorted(vocabulary, key=vocabulary.get, reverse=True)\n",
    "# Adding _UNK token to the vocabulary\n",
    "words = ['_UNK'] + words\n",
    "\n",
    "# Creating a dictionary of words and their indices\n",
    "w2idx = {word:i for i, word in enumerate(words)}\n",
    "idx2w = {i:word for i, word in enumerate(words)}\n",
    "\n",
    "# Encoding the preprocessed feature column\n",
    "data['encoded'] = data['preprocessed'].apply(lambda x : [w2idx[w] if w in w2idx.keys() else 0 for w in x])\n",
    "\n",
    "# Padding the encoded feature column to 300\n",
    "pad_to = 300\n",
    "data['encoded'] = data['encoded'].apply(lambda x: pad(x, pad_to))\n",
    "\n",
    "# Save preprocessed data\n",
    "data.to_csv('preprocessed.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "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>label</th>\n",
       "      <th>preprocessed</th>\n",
       "      <th>encoded</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>['law', 'enforcement', 'high', 'alert', 'follo...</td>\n",
       "      <td>[254514, 253979, 254285, 250522, 254185, 25337...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>['unbelievable', 'obama’s', 'attorney', 'gener...</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>['bobby', 'jindal', 'raised', 'hindu', 'uses',...</td>\n",
       "      <td>[245310, 242683, 253712, 246833, 252352, 25434...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>['satan', '2', 'russia', 'unvelis', 'image', '...</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>['time', 'christian', 'group', 'sues', 'amazon...</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   label                                       preprocessed  \\\n",
       "0      1  ['law', 'enforcement', 'high', 'alert', 'follo...   \n",
       "1      1  ['unbelievable', 'obama’s', 'attorney', 'gener...   \n",
       "2      0  ['bobby', 'jindal', 'raised', 'hindu', 'uses',...   \n",
       "3      1  ['satan', '2', 'russia', 'unvelis', 'image', '...   \n",
       "4      1  ['time', 'christian', 'group', 'sues', 'amazon...   \n",
       "\n",
       "                                             encoded  \n",
       "0  [254514, 253979, 254285, 250522, 254185, 25337...  \n",
       "1  [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...  \n",
       "2  [245310, 242683, 253712, 246833, 252352, 25434...  \n",
       "3  [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...  \n",
       "4  [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load Preprocessed Data\n",
    "data = pd.read_csv('preprocessed.csv', low_memory=False)\n",
    "\n",
    "# Convert encoded column to list\n",
    "data['encoded'] = data['encoded'].apply(lambda x : ast.literal_eval(x))\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get data as numpy array\n",
    "data_np = data[['encoded', 'label']].to_numpy()\n",
    "\n",
    "# Delete data to free up memory\n",
    "del(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create a list of features and labels\n",
    "X = []\n",
    "y = []\n",
    "for idx, (feat, label) in enumerate(data_np):\n",
    "    X.append(feat)\n",
    "    y.append(label)\n",
    "    "
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train, Validation, Test Split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "X_train_val, X_test, y_train_val, y_test = train_test_split(X, y, stratify=y, random_state=0, test_size=0.2)\n",
    "X_train, X_val, y_train, y_val = train_test_split(X_train_val, y_train_val, stratify=y_train_val, random_state=0,test_size=0.2)\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## DataLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "class MyDataset(Dataset):\n",
    "    def __init__(self, data, labels):\n",
    "        self.data = data\n",
    "        self.labels = labels\n",
    "\n",
    "    def __len__(self):\n",
    "        return len(self.data)\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        sample = self.data[idx]\n",
    "        label = self.labels[idx]\n",
    "        sample = torch.tensor(sample, dtype=torch.float)\n",
    "        label = torch.tensor(label, dtype=torch.long)\n",
    "        return sample, label\n",
    "    \n",
    "train_dataset = MyDataset(X_train, y_train)\n",
    "val_dataset = MyDataset(X_val, y_val)\n",
    "test_dataset = MyDataset(X_test, y_test)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size = 256\n",
    "train_dataloader = DataLoader(train_dataset, batch_size=batch_size)\n",
    "val_dataloader = DataLoader(val_dataset, batch_size=batch_size)\n",
    "test_dataloader = DataLoader(test_dataset, batch_size=batch_size)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "device(type='cuda')"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "device"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model Building"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "class NewsClassifier(nn.Module):\n",
    "\n",
    "    def __init__(self, vocab_size, embedding_dim, hidden_dim, n_layers, n_classes, drop_prob=0.5):\n",
    "\n",
    "        super(NewsClassifier, self).__init__()\n",
    "        self.hidden_dim = hidden_dim\n",
    "        self.n_layers = n_layers\n",
    "        self.embedding = nn.Embedding(vocab_size, embedding_dim)\n",
    "        self.lstm = nn.LSTM(embedding_dim, hidden_dim, batch_first=True, num_layers =n_layers)\n",
    "        self.dropout = nn.Dropout(drop_prob)\n",
    "        self.linear = nn.Linear(hidden_dim, n_classes)\n",
    "        self.sigmoid = nn.Sigmoid()\n",
    "        \n",
    "\n",
    "    def forward(self, x):\n",
    "        batch_size = x.size(0)\n",
    "        h_0 = Variable(torch.zeros(1*self.n_layers, batch_size, self.hidden_dim)).to(device)\n",
    "        c_0 = Variable(torch.zeros(1*self.n_layers, batch_size, self.hidden_dim)).to(device)\n",
    "\n",
    "\n",
    "        x = x.long()\n",
    "        embeds = self.embedding(x)\n",
    "\n",
    "        lstm_out, (h, _) = self.lstm(embeds, (h_0, c_0))\n",
    "\n",
    "        final_hidden_state = h[-1]\n",
    "\n",
    "        out = self.dropout(final_hidden_state)\n",
    "        out = self.linear(out)\n",
    "        out = self.sigmoid(out)\n",
    "        return out[:,-1]\n",
    "\n",
    "        \n",
    "\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Configurations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "vocab_size = 254575\n",
    "embedding_dim = 512\n",
    "hidden_dim = 400\n",
    "n_layers = 2\n",
    "n_classes = 2\n",
    "model = NewsClassifier(vocab_size=vocab_size, embedding_dim=embedding_dim, hidden_dim=hidden_dim\n",
    "                       , n_layers=n_layers, n_classes=n_classes).to(device)\n",
    "criterion = nn.BCELoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=0.001)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train(model, trainloader, optimizer, criterion):\n",
    "    model.train()\n",
    "    print('Training')\n",
    "    train_running_loss = 0.0\n",
    "    train_running_correct = 0\n",
    "    counter = 0\n",
    "    for i, data in tqdm(enumerate(trainloader), total=len(trainloader)):\n",
    "        counter += 1\n",
    "        image, labels = data\n",
    "        image = image.to(device)\n",
    "        labels = labels.to(device)\n",
    "        optimizer.zero_grad()\n",
    "        # forward pass\n",
    "        outputs = model(image)\n",
    "        # calculate the loss\n",
    "        loss = criterion(outputs, labels.float())\n",
    "        train_running_loss += loss.item()\n",
    "        # calculate the accuracy\n",
    "        preds = torch.round(outputs)\n",
    "        # _, preds = torch.max(outputs.data, 1)\n",
    "        train_running_correct += (preds == labels).sum().item()\n",
    "        # backpropagation\n",
    "        loss.backward()\n",
    "        # update the optimizer parameters\n",
    "        optimizer.step()\n",
    "    \n",
    "    # loss and accuracy for the complete epoch\n",
    "    epoch_loss = train_running_loss / counter\n",
    "    epoch_acc = 100. * (train_running_correct / len(trainloader.dataset))\n",
    "\n",
    "    return epoch_loss, epoch_acc\n",
    "# validation\n",
    "\n",
    "def validate(model, testloader, criterion):\n",
    "    model.eval()\n",
    "    print('Validation')\n",
    "    valid_running_loss = 0.0\n",
    "    valid_running_correct = 0\n",
    "    counter = 0\n",
    "    with torch.no_grad():\n",
    "        for i, data in tqdm(enumerate(testloader), total=len(testloader)):\n",
    "            counter += 1\n",
    "            \n",
    "            image, labels = data\n",
    "            image = image.to(device)\n",
    "            labels = labels.to(device)\n",
    "            # forward pass\n",
    "            outputs = model(image)\n",
    "            # calculate the loss\n",
    "            loss = criterion(outputs, labels.float())\n",
    "            valid_running_loss += loss.item()\n",
    "            # calculate the accuracy\n",
    "            preds = torch.round(outputs)\n",
    "\n",
    "            # _, preds = torch.max(outputs.data, 1)\n",
    "            valid_running_correct += (preds == labels).sum().item()\n",
    "        \n",
    "    # loss and accuracy for the complete epoch\n",
    "    epoch_loss = valid_running_loss / counter\n",
    "    epoch_acc = 100. * (valid_running_correct / len(testloader.dataset))\n",
    "    return epoch_loss, epoch_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO]: Epoch 1 of 10\n",
      "Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 179/179 [00:49<00:00,  3.64it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:03<00:00, 14.25it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training: Loss 0.367, Acc 83.813\n",
      "Validation loss: 0.269, validation acc: 88.878\n",
      "--------------------------------------------------\n",
      "[INFO]: Epoch 2 of 10\n",
      "Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 179/179 [00:46<00:00,  3.89it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:04<00:00,  9.61it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training: Loss 0.271, Acc 89.915\n",
      "Validation loss: 0.227, validation acc: 91.447\n",
      "--------------------------------------------------\n",
      "[INFO]: Epoch 3 of 10\n",
      "Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 179/179 [01:18<00:00,  2.29it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:04<00:00,  9.36it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training: Loss 0.177, Acc 93.192\n",
      "Validation loss: 0.173, validation acc: 93.762\n",
      "--------------------------------------------------\n",
      "[INFO]: Epoch 4 of 10\n",
      "Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 179/179 [01:16<00:00,  2.34it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:04<00:00, 10.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training: Loss 0.144, Acc 94.515\n",
      "Validation loss: 0.132, validation acc: 95.343\n",
      "--------------------------------------------------\n",
      "[INFO]: Epoch 5 of 10\n",
      "Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 179/179 [01:22<00:00,  2.18it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:04<00:00, 10.15it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training: Loss 0.075, Acc 97.425\n",
      "Validation loss: 0.111, validation acc: 95.693\n",
      "--------------------------------------------------\n",
      "[INFO]: Epoch 6 of 10\n",
      "Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 179/179 [01:10<00:00,  2.53it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:04<00:00, 10.08it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training: Loss 0.041, Acc 98.648\n",
      "Validation loss: 0.111, validation acc: 96.715\n",
      "--------------------------------------------------\n",
      "[INFO]: Epoch 7 of 10\n",
      "Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 179/179 [01:25<00:00,  2.10it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:04<00:00, 10.45it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training: Loss 0.022, Acc 99.286\n",
      "Validation loss: 0.120, validation acc: 96.715\n",
      "--------------------------------------------------\n",
      "[INFO]: Epoch 8 of 10\n",
      "Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 179/179 [01:12<00:00,  2.48it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:04<00:00,  9.58it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training: Loss 0.012, Acc 99.583\n",
      "Validation loss: 0.120, validation acc: 97.213\n",
      "--------------------------------------------------\n",
      "[INFO]: Epoch 9 of 10\n",
      "Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 179/179 [01:21<00:00,  2.19it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:04<00:00, 10.81it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training: Loss 0.015, Acc 99.526\n",
      "Validation loss: 0.110, validation acc: 97.091\n",
      "--------------------------------------------------\n",
      "[INFO]: Epoch 10 of 10\n",
      "Training\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 179/179 [01:20<00:00,  2.22it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:04<00:00,  9.48it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training: Loss 0.008, Acc 99.760\n",
      "Validation loss: 0.157, validation acc: 97.091\n",
      "--------------------------------------------------\n",
      "TRAINING COMPLETE\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "start_epoch = 0\n",
    "epochs = 10\n",
    "# lists to keep track of losses and accuracies\n",
    "train_loss, valid_loss = [], []\n",
    "train_acc, valid_acc = [], []\n",
    "# start the training\n",
    "for epoch in range(start_epoch, start_epoch+ epochs):\n",
    "    print(f\"[INFO]: Epoch {epoch+1} of {start_epoch + epochs}\")\n",
    "    train_epoch_loss, train_epoch_acc = train(model, train_dataloader, \n",
    "                                              optimizer, criterion)\n",
    "    valid_epoch_loss, valid_epoch_acc = validate(model, val_dataloader,  \n",
    "                                                 criterion)\n",
    "    train_loss.append(train_epoch_loss)\n",
    "    valid_loss.append(valid_epoch_loss)\n",
    "    train_acc.append(train_epoch_acc)\n",
    "    valid_acc.append(valid_epoch_acc)\n",
    "\n",
    "\n",
    "    print(f\"Training: Loss {train_epoch_loss:.3f}, Acc {train_epoch_acc:.3f}\")\n",
    "    print(f\"Validation loss: {valid_epoch_loss:.3f}, validation acc: {valid_epoch_acc:.3f}\")\n",
    "    print('-'*50)\n",
    "\n",
    "    \n",
    "\n",
    "\n",
    "print('TRAINING COMPLETE')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f9e82181a90>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x700 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x700 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10, 7))\n",
    "plt.plot(\n",
    "    train_acc, color='green', linestyle='-', \n",
    "    label='train accuracy'\n",
    ")\n",
    "plt.plot(\n",
    "    valid_acc, color='blue', linestyle='-', \n",
    "    label='validation accuracy'\n",
    ")\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()\n",
    "\n",
    "# loss plots\n",
    "plt.figure(figsize=(10, 7))\n",
    "plt.plot(\n",
    "    train_loss, color='orange', linestyle='-', \n",
    "    label='train loss'\n",
    ")\n",
    "plt.plot(\n",
    "    valid_loss, color='red', linestyle='-', \n",
    "    label='validation loss'\n",
    ")\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Testing\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 56/56 [00:05<00:00, 10.55it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.141, test acc: 97.302\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# Test\n",
    "\n",
    "def test(model, testloader, criterion):\n",
    "    model.eval()\n",
    "    print('Testing')\n",
    "    test_running_loss = 0.0\n",
    "    test_running_correct = 0\n",
    "    counter = 0\n",
    "    with torch.no_grad():\n",
    "        for i, data in tqdm(enumerate(testloader), total=len(testloader)):\n",
    "            counter += 1\n",
    "            \n",
    "            image, labels = data\n",
    "            image = image.to(device)\n",
    "            labels = labels.to(device)\n",
    "            # forward pass\n",
    "            outputs = model(image)\n",
    "            # calculate the loss\n",
    "            loss = criterion(outputs, labels.float())\n",
    "            test_running_loss += loss.item()\n",
    "            # calculate the accuracy\n",
    "            preds = torch.round(outputs)\n",
    "            test_running_correct += (preds == labels).sum().item()\n",
    "        \n",
    "    # loss and accuracy for the complete epoch\n",
    "    epoch_loss = test_running_loss / counter\n",
    "    epoch_acc = 100. * (test_running_correct / len(testloader.dataset))\n",
    "    return epoch_loss, epoch_acc\n",
    "\n",
    "test_loss, test_acc = test(model, test_dataloader, criterion)\n",
    "print(f\"Test loss: {test_loss:.3f}, test acc: {test_acc:.3f}\")\n",
    "\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* LSTMs are better than vanillas RNN in capturing context but as setences gets long, it suffers a lot. The impact of words gets less as the LSTM slides through later part of the sentence.\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "ml",
   "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": "3.9.16"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "eb780289cc9f5c346848791ea8ca286d18eb6e79cf93a1ba268284b000d9a250"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}