models.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "aeb9d5b1",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-12-13 17:21:51.984297: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
     ]
    }
   ],
   "source": [
    "import string\n",
    "import pickle\n",
    "\n",
    "import numpy as np \n",
    "import pandas as pd \n",
    "\n",
    "import gensim\n",
    "import gensim.downloader as api\n",
    "import nltk\n",
    "from nltk.stem import WordNetLemmatizer \n",
    "\n",
    "\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix, ConfusionMatrixDisplay, classification_report\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "\n",
    "import tensorflow as tf\n",
    "from keras.models import Sequential \n",
    "from keras.layers import LSTM, Dense, Embedding\n",
    "from keras.optimizers import Adam\n",
    "from keras.preprocessing.text import Tokenizer \n",
    "from keras.utils import pad_sequences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "98bb7ca5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are 2264 sentences in this dataset\n"
     ]
    }
   ],
   "source": [
    "fname = 'FinancialPhraseBank-v1.0/Sentences_AllAgree.txt'\n",
    "\n",
    "df = pd.read_csv(fname, sep=\"@\", encoding='latin1', header=None)\n",
    "print('There are {} sentences in this dataset'.format(len(df)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d3d067f9",
   "metadata": {},
   "source": [
    "### Preprocess the dataframe \n",
    "\n",
    "- Convert to lowercase\n",
    "- Remove punctuations\n",
    "- Apply Lemmatization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "1dc21721",
   "metadata": {},
   "outputs": [],
   "source": [
    "def preprocess(df):\n",
    "    # Remove lowercase, punctuations, lemmatize\n",
    "    lemmatizer = WordNetLemmatizer()\n",
    "    \n",
    "    df['lcase'] = df[0].apply(lambda x: x.lower())\n",
    "    df['no_punct'] = df['lcase'].apply(lambda x: x.translate(str.maketrans('', '', string.punctuation)))\n",
    "    df['cleaned'] = df['no_punct'].apply(lambda x: ' '.join(lemmatizer.lemmatize(t) for t in x.split()))\n",
    "    \n",
    "preprocess(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "f63e593c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set length - 2037, Test set length - 227\n"
     ]
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(df['cleaned'], df[1], test_size=0.1, stratify=df[1], random_state=13)\n",
    "\n",
    "print(\"Training set length - {}, Test set length - {}\".format(len(X_train), len(X_test)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "549ebc08",
   "metadata": {},
   "source": [
    "### Represent words in the sentences using tf-idf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "9df6a12a",
   "metadata": {},
   "outputs": [],
   "source": [
    "tfidf = TfidfVectorizer()\n",
    "\n",
    "X_train_tfidf = tfidf.fit_transform(X_train)\n",
    "X_test_tfidf = tfidf.transform(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "396e04d6",
   "metadata": {},
   "source": [
    "### Model 1 - Multinomial Naive Bayes' Classifier using tf-idf features\n",
    "\n",
    "The laplacian smoothing coefficient is unchanged"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "80b1eb7c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultinomialNB()"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nb = MultinomialNB()\n",
    "nb.fit(X_train_tfidf, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f0d8614c",
   "metadata": {},
   "source": [
    "### Model 1 - Prediction and evaluation metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "69dbdacc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of Naive Bayes' model - 77.974 %\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7fe350086400>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjAAAAG2CAYAAABh8Lw3AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/NK7nSAAAACXBIWXMAAA9hAAAPYQGoP6dpAABPTUlEQVR4nO3deVhUZfsH8O+wDcMyAyiL6KAYhFqoqGXgnhRomdubmqRiLmmiqZlL5b6MWiqhJmal0gtv2iK5pSHuSy64lKK4YaCCpAgICAwz5/cHP6cmUBlnYDjy/VzXud53nnPOc+5hGri9n+c8RyIIggAiIiIiEbEwdwBEREREhmICQ0RERKLDBIaIiIhEhwkMERERiQ4TGCIiIhIdJjBEREQkOkxgiIiISHSYwBAREZHoMIEhIiIi0WECQ0RERKLDBIaIiIgqbf/+/ejRowc8PT0hkUgQHx//0GNHjRoFiUSCyMhIvfbs7GyEhYVBLpfDyckJw4YNQ35+vkFxMIEhIiKiSisoKECLFi2wcuXKRx63adMm/Pbbb/D09Cy3LywsDOfOnUNCQgK2bt2K/fv3Y+TIkQbFYWXQ0URERFSrdevWDd26dXvkMTdu3MDYsWOxc+dOvPbaa3r7zp8/jx07duD48eNo06YNAGD58uXo3r07PvvsswoTnoowgamBtFotbt68CUdHR0gkEnOHQ0REBhAEAffu3YOnpycsLKpuoKOoqAglJSUm6UsQhHJ/b6RSKaRSqcF9abVaDBo0CB9++CGee+65cvuPHDkCJycnXfICAMHBwbCwsMDRo0fRu3fvSl2HCUwNdPPmTSiVSnOHQURERkhPT0eDBg2qpO+ioiJ4N3RAZpbGJP05ODiUm4Myc+ZMzJo1y+C+Fi1aBCsrK4wbN67C/ZmZmXBzc9Nrs7KygouLCzIzMyt9HSYwNZCjoyMAoIPlG7CSWJs5Gqpqlu6u5g6BqpHgaG/uEKiKlWqKse/SCt3v8qpQUlKCzCwN/kxqBLmjcVWevHtaNGx9Denp6ZDL5br2J6m+JCUl4fPPP8fJkyerfASBCUwN9OBDt5JYM4GpBSwtDP8lQeIlWPLzri2qYwqAg6MEDo7GXUeLsvPlcrleAvMkDhw4gKysLHh5eenaNBoNPvjgA0RGRuLatWvw8PBAVlaW3nmlpaXIzs6Gh4dHpa/FBIaIiEikNIIWGsH4Pkxl0KBBCA4O1msLCQnBoEGDMHToUABAYGAgcnJykJSUhNatWwMAdu/eDa1Wi7Zt21b6WkxgiIiIREoLAVoYl8EYen5+fj4uX76se52amorTp0/DxcUFXl5eqFOnjt7x1tbW8PDwgJ+fHwCgadOmCA0NxYgRIxAdHQ21Wo2IiAgMGDCg0ncgAVwHhoiIiAxw4sQJBAQEICAgAAAwceJEBAQEYMaMGZXuIzY2Fk2aNEHXrl3RvXt3tG/fHl9++aVBcbACQ0REJFJaaGHsAJChPXTu3BmCUPmqzbVr18q1ubi4IC4uzqDr/hsTGCIiIpHSCAI0BiQTD+tDjDiERERERKLDCgwREZFImWMSb03BBIaIiEiktBCgqaUJDIeQiIiISHRYgSEiIhIpDiERERGR6PAuJCIiIiIRYQWGiIhIpLT/vxnbhxgxgSEiIhIpjQnuQjL2fHNhAkNERCRSGgEmeBq1aWKpbpwDQ0RERKLDCgwREZFIcQ4MERERiY4WEmggMboPMeIQEhEREYkOKzBEREQipRXKNmP7ECMmMERERCKlMcEQkrHnmwuHkIiIiEh0WIEhIiISqdpcgWECQ0REJFJaQQKtYORdSEaeby4cQiIiIiLRYQWGiIhIpDiERERERKKjgQU0Rg6maEwUS3VjAkNERCRSggnmwAicA0NERERUPViBISIiEinOgSEiIiLR0QgW0AhGzoER6aMEOIREREREosMKDBERkUhpIYHWyFqEFuIswTCBISIiEqnaPAeGQ0hEREQkOqzAEBERiZRpJvFyCImIiIiqUdkcGCMf5sghJCIiIqLqwQoMERGRSGlN8Cwk3oVERERE1YpzYIiIiEh0tLCotevAcA4MERERiQ4rMERERCKlESTQCEYuZGfk+ebCBIaIiEikNCaYxKvhEBIRERFR9WAFhoiISKS0ggW0Rt6FpBXpXUiswBAREYnUgyEkYzdD7N+/Hz169ICnpyckEgni4+N1+9RqNaZMmQJ/f3/Y29vD09MTgwcPxs2bN/X6yM7ORlhYGORyOZycnDBs2DDk5+cbFAcTGCIiIqq0goICtGjRAitXriy3r7CwECdPnsT06dNx8uRJ/PTTT0hJScEbb7yhd1xYWBjOnTuHhIQEbN26Ffv378fIkSMNioNDSERERCKlhfF3EWkNPL5bt27o1q1bhfsUCgUSEhL02lasWIEXX3wRaWlp8PLywvnz57Fjxw4cP34cbdq0AQAsX74c3bt3x2effQZPT89KxcEKDBERkUg9WMjO2A0A8vLy9Lbi4mKTxJibmwuJRAInJycAwJEjR+Dk5KRLXgAgODgYFhYWOHr0aKX7ZQJDREREUCqVUCgUuk2lUhndZ1FREaZMmYK33noLcrkcAJCZmQk3Nze946ysrODi4oLMzMxK980hJCIiIpEyzbOQys5PT0/XJRkAIJVKjepXrVajX79+EAQBq1atMqqvijCBISIiEiktJNDC2DkwZefL5XK9BMYYD5KXP//8E7t379br18PDA1lZWXrHl5aWIjs7Gx4eHpW+BhMYMqv+YzLQLjQHDZ4pQkmRBZKT7PGNqgGuX7U1d2hkpOcCstH37avwaZKLOq7FmPthK/y27+9fTkGdM9GtTxp8muZCrlBjbFh7XL1kml+eVP2e9/8LffulwMf3LurULcLcGUE4cri+bv/2Xd9XeN7XXzbHjxv9qivMp44pKzCm8iB5uXTpEvbs2YM6dero7Q8MDEROTg6SkpLQunVrAMDu3buh1WrRtm3bSl+Hc2Aeo1GjRoiMjDR3GE8t/7b52LLeFRN6NcG0MF9YWQmY/99LkMo05g6NjGRrW4rUS45Y9elzFe6XyjRIPuOMtSuaVHNkVBVsbUuRetUJXyxvVeH+sDd76G3LPm0DrRY4dKB+hcdTzZWfn4/Tp0/j9OnTAIDU1FScPn0aaWlpUKvV+M9//oMTJ04gNjYWGo0GmZmZyMzMRElJCQCgadOmCA0NxYgRI3Ds2DEcOnQIERERGDBgQKXvQALMXIEJDw/H+vXroVKpMHXqVF17fHw8evfuDaEaVwdct24dxo8fj5ycHL3248ePw97evtriqG0+Geyr93rJB42w4fTv8PUvxNljjmaKikwh6Ygbko64PXT/nl/K/nC51SusrpCoCp04Xg8njtd76P67d/Wrqi8F3cTvp92QmeFQ1aE91UzzLCTDzj9x4gS6dOmiez1x4kQAwJAhQzBr1ixs3rwZANCyZUu98/bs2YPOnTsDAGJjYxEREYGuXbvCwsICffv2RVRUlEFxmH0IydbWFosWLcK7774LZ2dnc4dTjqurq7lDqFXsHMsqL/dyzP6fJhFVESenIrzQNgNLF79o7lBETytIoDV2HRgDz+/cufMjCwyVKT64uLggLi7OoOv+m9mHkIKDg+Hh4fHI27UOHjyIDh06QCaTQalUYty4cSgoKNDtz8jIwGuvvQaZTAZvb2/ExcWVG/pZunSpbmljpVKJ9957T7ds8d69ezF06FDdveoSiQSzZs0CoD+ENHDgQPTv318vNrVajbp16yImJgYAoNVqoVKp4O3tDZlMhhYtWuCHH34wwU/q6SeRCBg16zrOHbfHnxdl5g6HiKpI8KvXcL/QisNHZBSzJzCWlpZYsGABli9fjuvXr5fbf+XKFYSGhqJv3774/fffsWHDBhw8eBARERG6Yx48Z2Hv3r348ccf8eWXX5ab4WxhYYGoqCicO3cO69evx+7duzF58mQAQFBQECIjIyGXy5GRkYGMjAxMmjSpXCxhYWHYsmWL3vMadu7cicLCQvTu3RsAoFKpEBMTg+joaJw7dw4TJkzA22+/jX379j30Z1BcXFxuAaHaaMy8NDR69j5UYxqbOxQiqkKvhF7Dnt0NoVZbmjsU0dOa4DlIWvOnAk+kRkTdu3dvtGzZEjNnziy3T6VSISwsDOPHj4evry+CgoIQFRWFmJgYFBUV4cKFC9i1axfWrFmDtm3bolWrVvjqq69w//59vX7Gjx+PLl26oFGjRnj55Zcxb948bNy4EQBgY2MDhUIBiUQCDw8PeHh4wMGh/LhsSEgI7O3tsWnTJl1bXFwc3njjDTg6OqK4uBgLFizAN998g5CQEDRu3Bjh4eF4++23sXr16oe+f5VKpbd4kFKpfNIfpWi9NycNbbvmYvKAZ3E708bc4RBRFXnu+b+g9LqHndu9zR3KU+HB06iN3cSoxkS9aNEirF+/HufPn9drP3PmDNatWwcHBwfdFhISAq1Wi9TUVKSkpMDKygqtWv09893Hx6fcfJpdu3aha9euqF+/PhwdHTFo0CDcuXMHhYWVn0BoZWWFfv36ITY2FkDZA61+/vlnhIWFAQAuX76MwsJCvPLKK3rxxsTE4MqVKw/td9q0acjNzdVt6enplY5J/AS8NycNQaE5mDLgWdxKN27hJCKq2V7tlopLKc5Ivepk7lBI5GrMTMmOHTsiJCQE06ZNQ3h4uK49Pz8f7777LsaNG1fuHC8vL1y8ePGxfV+7dg2vv/46Ro8ejfnz58PFxQUHDx7EsGHDUFJSAjs7u0rHGRYWhk6dOiErKwsJCQmQyWQIDQ3VxQoA27ZtQ/36+mO7j1rRUCqVGr3ioViNmZeOLj2zMXv4M7hfYAlnVzUAoCDPEiXFNSa/pidgKyuFZ4O//4Hg4XkfjX3zcC/PGn/dksFBXgI39yK4uBYBAOo3LPv+3M2W4u6d2vl9EDNb21J41v97eN29XgEaP5ODe/ds8FdW2e9YmZ0aHTpex1erW5grzKeOBhJojFzIztjzzaXGJDAAsHDhQrRs2RJ+fn8vatSqVSskJyfDx8enwnP8/PxQWlqKU6dO6RbEuXz5Mu7evas7JikpCVqtFkuWLIGFRdkfxQfDRw/Y2NhAo3n82iNBQUFQKpXYsGEDfvnlF7z55puwtrYGADRr1gxSqRRpaWno1KmTYW++luox+C8AwKff6yeiSyY2RMIPdc0REpmIb9NcLIz++8FsIyaUVVd3ba2PZXNa4KUOWZgw83fd/qkLTgMAYtf4IG7Ns9UaKxnP1y8bi5b8Pddv5OgzAICEnQ2x7NOyu406dUkHJMDePV5mifFpZIohILEOIdWoBMbf3x9hYWF694JPmTIFL730EiIiIjB8+HDY29sjOTkZCQkJWLFiBZo0aYLg4GCMHDkSq1atgrW1NT744APIZDJIJGVZpY+PD9RqNZYvX44ePXrg0KFDiI6O1rt2o0aNkJ+fj8TERLRo0QJ2dnYPrcwMHDgQ0dHRuHjxIvbs2aNrd3R0xKRJkzBhwgRotVq0b98eubm5OHToEORyOYYMGVIFPzVxC/Vqbe4QqIr8cbIOXnux+0P379rWALu2NajGiKgq/XHGDd2D33zkMTu2NcaObZykT6ZR49KuOXPmQKvV6l43b94c+/btw8WLF9GhQwcEBARgxowZeqv1xcTEwN3dHR07dkTv3r0xYsQIODo6wta2bOGkFi1aYOnSpVi0aBGef/55xMbGlrttOygoCKNGjUL//v3h6uqKxYsXPzTGsLAwJCcno379+mjXrp3evrlz52L69OlQqVS61Qa3bdsGb29OWCMiItPS4O9hpCffxEkiVOdyt9Xk+vXrUCqVuom7YpOXlweFQoEuVn1hJbE2dzhUxSw93M0dAlUjQc6VvZ92pZpiJF5YgtzcXJM9HPHfHvyd+OS3V2HrYNzfiaJ8Nea99GuVxlsVatQQ0pPavXs38vPz4e/vj4yMDEyePBmNGjVCx44dzR0aERFRlamJD3OsLk9FAqNWq/HRRx/h6tWrcHR0RFBQEGJjY3WTa4mIiOjp8lQkMCEhIQgJCTF3GERERNVKgARaI2+DFngbNREREVWn2jyEJM6oiYiIqFZjBYaIiEiktIIEWsG4ISBjzzcXJjBEREQi9eCJ0sb2IUbijJqIiIhqNVZgiIiIRIpDSERERCQ6WlhAa+RgirHnm4s4oyYiIqJajRUYIiIikdIIEmiMHAIy9nxzYQJDREQkUpwDQ0RERKIjCBbQGrmSrsCVeImIiIiqByswREREIqWBBBojH8Zo7PnmwgSGiIhIpLSC8XNYtIKJgqlmHEIiIiIi0WEFhoiISKS0JpjEa+z55sIEhoiISKS0kEBr5BwWY883F3GmXURERFSrsQJDREQkUlyJl4iIiESnNs+BEWfUREREVKuxAkNERCRSWpjgWUgincTLBIaIiEikBBPchSQwgSEiIqLqVJufRs05MERERCQ6rMAQERGJVG2+C4kJDBERkUhxCImIiIhIRFiBISIiEqna/CwkJjBEREQixSEkIiIiokrYv38/evToAU9PT0gkEsTHx+vtFwQBM2bMQL169SCTyRAcHIxLly7pHZOdnY2wsDDI5XI4OTlh2LBhyM/PNygOJjBEREQi9aACY+xmiIKCArRo0QIrV66scP/ixYsRFRWF6OhoHD16FPb29ggJCUFRUZHumLCwMJw7dw4JCQnYunUr9u/fj5EjRxoUB4eQiIiIRMocQ0jdunVDt27dKtwnCAIiIyPxySefoGfPngCAmJgYuLu7Iz4+HgMGDMD58+exY8cOHD9+HG3atAEALF++HN27d8dnn30GT0/PSsXBCgwREREhLy9PbysuLja4j9TUVGRmZiI4OFjXplAo0LZtWxw5cgQAcOTIETg5OemSFwAIDg6GhYUFjh49WulrMYEhIiISKVMOISmVSigUCt2mUqkMjiczMxMA4O7urtfu7u6u25eZmQk3Nze9/VZWVnBxcdEdUxkcQiIiIhIpAcbfBi38//+mp6dDLpfr2qVSqVH9VjUmMERERCJlyjkwcrlcL4F5Eh4eHgCAW7duoV69err2W7duoWXLlrpjsrKy9M4rLS1Fdna27vzK4BASERERmYS3tzc8PDyQmJioa8vLy8PRo0cRGBgIAAgMDEROTg6SkpJ0x+zevRtarRZt27at9LVYgSEiIhIpc9yFlJ+fj8uXL+tep6am4vTp03BxcYGXlxfGjx+PefPmwdfXF97e3pg+fTo8PT3Rq1cvAEDTpk0RGhqKESNGIDo6Gmq1GhERERgwYECl70ACmMAQERGJljkSmBMnTqBLly661xMnTgQADBkyBOvWrcPkyZNRUFCAkSNHIicnB+3bt8eOHTtga2urOyc2NhYRERHo2rUrLCws0LdvX0RFRRkUBxMYIiIiqrTOnTtDEISH7pdIJJgzZw7mzJnz0GNcXFwQFxdnVBxMYIiIiESqNj8LiQkMERGRSAmCBIKRCYix55sL70IiIiIi0WEFhoiISKS0kBi9kJ2x55sLExgiIiKRqs1zYDiERERERKLDCgwREZFI1eZJvExgiIiIRKo2DyExgSEiIhKp2lyB4RwYIiIiEh1WYGowobQUgkScmTFV3rZj28wdAlWjEM+W5g6BqphGUFfbtQQTDCGJtQLDBIaIiEikBACPeCxRpfsQIw4hERERkeiwAkNERCRSWkgg4Uq8REREJCa8C4mIiIhIRFiBISIiEimtIIGEC9kRERGRmAiCCe5CEultSBxCIiIiItFhBYaIiEikavMkXiYwREREIsUEhoiIiESnNk/i5RwYIiIiEh1WYIiIiESqNt+FxASGiIhIpMoSGGPnwJgomGrGISQiIiISHVZgiIiIRIp3IREREZHoCP+/GduHGHEIiYiIiESHFRgiIiKR4hASERERiU8tHkNiAkNERCRWJqjAQKQVGM6BISIiItFhBYaIiEikuBIvERERiU5tnsTLISQiIiISHVZgiIiIxEqQGD8JV6QVGCYwREREIlWb58BwCImIiIhEhxUYIiIiseJCdo+2efPmSnf4xhtvPHEwREREVHm1+S6kSiUwvXr1qlRnEokEGo3GmHiIiIiohtJoNJg1axb++9//IjMzE56enggPD8cnn3wCiaQsERIEATNnzsSaNWuQk5ODdu3aYdWqVfD19TVpLJWaA6PVaiu1MXkhIiKqZoKRmwEWLVqEVatWYcWKFTh//jwWLVqExYsXY/ny5bpjFi9ejKioKERHR+Po0aOwt7dHSEgIioqKjH2neoyaA1NUVARbW1tTxUJEREQGqO4hpMOHD6Nnz5547bXXAACNGjXC//73Pxw7duz/+xIQGRmJTz75BD179gQAxMTEwN3dHfHx8RgwYIBRsf6TwXchaTQazJ07F/Xr14eDgwOuXr0KAJg+fTq+/vprkwVGREREj2Fs9cXAKkxQUBASExNx8eJFAMCZM2dw8OBBdOvWDQCQmpqKzMxMBAcH685RKBRo27Ytjhw5Ysw7LcfgBGb+/PlYt24dFi9eDBsbG137888/j6+++sqkwREREVH1yMvL09uKi4vLHTN16lQMGDAATZo0gbW1NQICAjB+/HiEhYUBADIzMwEA7u7ueue5u7vr9pmKwQlMTEwMvvzyS4SFhcHS0lLX3qJFC1y4cMGkwREREdGjSEy0AUqlEgqFQrepVKpyV9u4cSNiY2MRFxeHkydPYv369fjss8+wfv36Kn6f5Rk8B+bGjRvw8fEp167VaqFWq00SFBEREVWCCdeBSU9Ph1wu1zVLpdJyh3744Ye6KgwA+Pv7488//4RKpcKQIUPg4eEBALh16xbq1aunO+/WrVto2bKlkYHqM7gC06xZMxw4cKBc+w8//ICAgACTBEVERETVSy6X620VJTCFhYWwsNBPHSwtLaHVagEA3t7e8PDwQGJiom5/Xl4ejh49isDAQJPGa3AFZsaMGRgyZAhu3LgBrVaLn376CSkpKYiJicHWrVtNGhwRERE9QjWvxNujRw/Mnz8fXl5eeO6553Dq1CksXboU77zzDoCy9eDGjx+PefPmwdfXF97e3pg+fTo8PT0rvaZcZRmcwPTs2RNbtmzBnDlzYG9vjxkzZqBVq1bYsmULXnnlFZMGR0RERI9QzU+jXr58OaZPn4733nsPWVlZ8PT0xLvvvosZM2bojpk8eTIKCgowcuRI5OTkoH379tixY4fJl12RCIJYn0P59MrLy4NCoUBn9ISVxNrc4VAV23nztLlDoGoU4tnS3CFQFSsV1NiLn5Gbm6s3p8SUHvydUK6cDQuZcYmB9n4R0sfMrNJ4q8ITL2R34sQJnD9/HkDZvJjWrVubLCgiIiJ6PEEo24ztQ4wMTmCuX7+Ot956C4cOHYKTkxMAICcnB0FBQfjuu+/QoEEDU8dIREREFanFT6M2+C6k4cOHQ61W4/z588jOzkZ2djbOnz8PrVaL4cOHV0WMRERERHoMrsDs27cPhw8fhp+fn67Nz88Py5cvR4cOHUwaHBERET1CNU/irUkMTmCUSmWFC9ZpNBp4enqaJCgiIiJ6PIlQthnbhxgZPIT06aefYuzYsThx4oSu7cSJE3j//ffx2WefmTQ4IiIieoRqfphjTVKpCoyzszMkkr9LTAUFBWjbti2srMpOLy0thZWVFd555x2TL1RDRERE9G+VSmAiIyOrOAwiIiIyGOfAPNqQIUOqOg4iIiIyVC2+jfqJF7IDgKKiIpSUlOi1iWkVPyIiIhIngyfxFhQUICIiAm5ubrC3t4ezs7PeRkRERNWkFk/iNTiBmTx5Mnbv3o1Vq1ZBKpXiq6++wuzZs+Hp6YmYmJiqiJGIiIgqUosTGIOHkLZs2YKYmBh07twZQ4cORYcOHeDj44OGDRsiNjYWYWFhVREnERERkY7BFZjs7Gw0btwYQNl8l+zsbABA+/btsX//ftNGR0RERA/34C4kYzcRMjiBady4MVJTUwEATZo0wcaNGwGUVWYePNyRyFA9wm9j/dFkbLn6Oz7fegl+LQvNHRIZ6I/f7DFjsDfeCngOIZ4tcfgXxUOP/XxKA4R4tsRPa1z12vPuWmLhGC/0ftYffZr4Y+lEJe4XGPxrimoIfq+r3oOVeI3dxMjg3wxDhw7FmTNnAABTp07FypUrYWtriwkTJuDDDz80eYBit3fvXkgkEuTk5Jg7lBqr0xt3MXLmTcQu9cCYkGdxNdkW8+OuQlGn/CMrqOYqKrRA4+fuI2LB9Uced+gXBS4k2aOOR0m5fYsiGuLPFBlU313BnPVX8cdRB0R+qKyqkKkK8XtNVc3gBGbChAkYN24cACA4OBgXLlxAXFwcTp06hffff9/kAT4QHh4OiUSChQsX6rXHx8frrRJsrGvXrkEikeD06dMm65Merc/I29gR54JfN7gg7ZItoqY0QPF9CULeyjZ3aGSAF16+h/ApmWjXLfehx9zOsMYXn9THlJV/wupfM/DSLklxYo8cE5akoUmrQjzftgDvzbuOfT874U6mUSs+kBnwe11NavEkXqNrsw0bNkSfPn3QvHlzU8TzSLa2tli0aBHu3r1b5dd6nH+vf0NPxspaC9/mhTh5wFHXJggSnDrgiGatWW5+mmi1wOJxXvjP6Cw08isqt//8CXs4KErxbIv7urZWHe5BYgFcOGVfnaGSkfi9pupQqX/WREVFVbrDB9WZqhAcHIzLly9DpVJh8eLFFR5z8OBBTJs2DSdOnEDdunXRu3dvqFQq2NuX/QKUSCTYtGmT3jObnJycEBkZifDwcHh7ewMAAgICAACdOnXC3r17ER4ejpycHLzwwgtYuXIlpFIpUlNT8e233+Lzzz9HSkoK7O3t8fLLLyMyMhJubm5V9nN4mshdNLC0AnL+0v9P8e5tKyh9is0UFVWFjSvdYGkpoNew2xXuz/7LCk51SvXaLK0AR6dSZGexAiMm/F5XHwlM8DRqk0RS/Sr1W2HZsmWV6kwikVRpAmNpaYkFCxZg4MCBGDduHBo0aKC3/8qVKwgNDcW8efPwzTff4K+//kJERAQiIiKwdu3aSl3j2LFjePHFF7Fr1y4899xzsLGx0e1LTEyEXC5HQkKCrk2tVmPu3Lnw8/NDVlYWJk6ciPDwcGzfvr3S76u4uBjFxX9/qfPy8ip9LpEYXPpdhvivXLFyZwpMOOJLRLVYpRKYB3cd1QS9e/dGy5YtMXPmTHz99dd6+1QqFcLCwjB+/HgAgK+vL6KiotCpUyesWrUKtra2j+3f1bXsrog6derAw8NDb5+9vT2++uorvaTmnXfe0f3/xo0bIyoqCi+88ALy8/Ph4OBQqfekUqkwe/bsSh37tMnLtoSmFHBy1f+Xt3PdUtz9i//qflr8cdQBObet8PYLz+natBoJ1sz2RPwaV8QcS4aLayly7uh/5ppS4F6OFVzcSv/dJdVg/F5Xo1r8MEdR3p+4aNEirF+/HufPn9drP3PmDNatWwcHBwfdFhISAq1Wa5IkzN/fXy95AYCkpCT06NEDXl5ecHR0RKdOnQAAaWlple532rRpyM3N1W3p6elGxyoWpWoLXPrdDgHt7+naJBIBLdvnIznJzoyRkSkF981GdGIKViX8vdXxKMF/RmdhftwVAEDTNgXIz7XCpd9luvNOH3SEoAWaBBSYK3R6AvxeV6NaPIlXlKlwx44dERISgmnTpiE8PFzXnp+fj3fffbfCYSwvLy8AZcNcgqD/aanVlbut78E8mgcKCgoQEhKCkJAQxMbGwtXVFWlpaQgJCTFokq9UKoVUKq308U+bn76si0mR6bh4xg4pp+zQe8RfsLXT4tfvXMwdGhngfoEFbqb+/d9xZroNrpyVwdGpFG4N1JC7aPSOt7ICnN1KdXMivHyL0aZLHiInKTF20XVo1BKs/KQ+OvXMQR0PVmDEht9rqmqiTGAAYOHChWjZsiX8/Px0ba1atUJycjJ8fHweep6rqysyMjJ0ry9duoTCwr9nxT+osGg0mnLn/tuFCxdw584dLFy4EEpl2VoVJ06cMPi91Hb7NjtDUUeDwR9mwtm1FFfPyfBxmDdyblubOzQywMUzdpj8n7+/e6tn1QcAvNIvG5MiK1eRnLLiT6z8uAGm9nsGEgugffccvDfvRpXES1WL3+tqYooKCisw1cvf3x9hYWF6d0hNmTIFL730EiIiIjB8+HDY29sjOTkZCQkJWLFiBQDg5ZdfxooVKxAYGAiNRoMpU6bA2vrvL5SbmxtkMhl27NiBBg0awNbWFgpFxSuKenl5wcbGBsuXL8eoUaNw9uxZzJ07t2rf+FNq89q62Ly2rrnDICO0CMrHzpunK318zLHkcm1yZw2mffGnCaMic+L3uuqZYiXdWrMSb00yZ84caLVa3evmzZtj3759uHjxIjp06ICAgADMmDEDnp6eumOWLFkCpVKJDh06YODAgZg0aRLs7P4ek7WyskJUVBRWr14NT09P9OzZ86HXd3V1xbp16/D999+jWbNmWLhwIT777LOqebNERESkIxH+PSGkEg4cOIDVq1fjypUr+OGHH1C/fn18++238Pb2Rvv27asizlolLy8PCoUCndETVhKWW592hlQtSPxCPFuaOwSqYqWCGnvxM3JzcyGXy6vkGg/+TjSaNx8WlbjD9lG0RUW49snHVRpvVTC4AvPjjz8iJCQEMpkMp06d0q1fkpubiwULFpg8QCIiInqIWnwXksEJzLx58xAdHY01a9bozR1p164dTp48adLgiIiIiCpi8CTelJQUdOzYsVy7QqHgE5eJiIiqESfxGsDDwwOXL18u137w4EE0btzYJEERERFRJTxYidfYTYQMTmBGjBiB999/H0ePHoVEIsHNmzcRGxuLSZMmYfTo0VURIxEREVWkFs+BMXgIaerUqdBqtejatSsKCwvRsWNHSKVSTJo0CWPHjq2KGImIiIj0GJzASCQSfPzxx/jwww9x+fJl5Ofno1mzZpV+cCERERGZRm2eA/PEK/Ha2NigWbNmpoyFiIiIDMFHCVRely5dIJE8fMLP7t27jQqIiIiI6HEMTmBatmyp91qtVuP06dM4e/YshgwZYqq4iIiI6HFMMIRUayowy5Ytq7B91qxZyM/PNzogIiIiqqRaPIRksoc5vv322/jmm29M1R0RERHRQz3xJN5/O3LkCGyNfKAUERERGaAWV2AMTmD69Omj91oQBGRkZODEiROYPn26yQIjIiKiR+Nt1AZQKBR6ry0sLODn54c5c+bg1VdfNVlgRERERA9jUAKj0WgwdOhQ+Pv7w9nZuapiIiIiohrqxo0bmDJlCn755RcUFhbCx8cHa9euRZs2bQCUjczMnDkTa9asQU5ODtq1a4dVq1bB19fXpHEYNInX0tISr776Kp86TUREVBNU87OQ7t69i3bt2sHa2hq//PILkpOTsWTJEr2ixuLFixEVFYXo6GgcPXoU9vb2CAkJQVFRkfHv9x8MHkJ6/vnncfXqVXh7e5s0ECIiIjJMdc+BWbRoEZRKJdauXatr+2c+IAgCIiMj8cknn6Bnz54AgJiYGLi7uyM+Ph4DBgwwLth/MPg26nnz5mHSpEnYunUrMjIykJeXp7cRERGR+Pz773lxcXG5YzZv3ow2bdrgzTffhJubGwICArBmzRrd/tTUVGRmZiI4OFjXplAo0LZtWxw5csSk8VY6gZkzZw4KCgrQvXt3nDlzBm+88QYaNGgAZ2dnODs7w8nJifNiiIiIqpuJho+USiUUCoVuU6lU5S519epV3XyWnTt3YvTo0Rg3bhzWr18PAMjMzAQAuLu7653n7u6u22cqlR5Cmj17NkaNGoU9e/aYNAAiIiJ6QiZcByY9PR1yuVzXLJVKyx2q1WrRpk0bLFiwAAAQEBCAs2fPIjo6utofJ1TpBEYQyt5hp06dqiwYIiIiMg+5XK6XwFSkXr16aNasmV5b06ZN8eOPPwIAPDw8AAC3bt1CvXr1dMfcunWr3LMUjWXQHJhHPYWaiIiIqteDSbzGbpXVrl07pKSk6LVdvHgRDRs2BFA2odfDwwOJiYm6/Xl5eTh69CgCAwNN8p4fMOgupGefffaxSUx2drZRAREREVElVfOjBCZMmICgoCAsWLAA/fr1w7Fjx/Dll1/iyy+/BFBW6Bg/fjzmzZsHX19feHt7Y/r06fD09ESvXr2MDFSfQQnM7Nmzy63ES0RERLXDCy+8gE2bNmHatGmYM2cOvL29ERkZibCwMN0xkydPRkFBAUaOHImcnBy0b98eO3bsMPnzEg1KYAYMGAA3NzeTBkBERERPxhzPQnr99dfx+uuvP7w/iQRz5szBnDlzjAvsMSqdwHD+CxERUQ1Ti59GXelJvA/uQiIiIiIyt0pXYLRabVXGQURERIaqxRUYg5+FRERERDWDOebA1BRMYIiIiMSqFldgDH6YIxEREZG5sQJDREQkVrW4AsMEhoiISKRq8xwYDiERERGR6LACQ0REJFYcQiIiIiKx4RASERERkYiwAkNERCRWHEIiIiIi0anFCQyHkIiIiEh0WIEhIiISKcn/b8b2IUZMYIiIiMSqFg8hMYEhIiISKd5GTURERCQirMAQERGJFYeQiIiISJREmoAYi0NIREREJDqswBAREYlUbZ7EywSGiIhIrGrxHBgOIREREZHosAJDREQkUhxCIiIiIvHhEBIRERGReLACQ2RmoT0HmTsEqkaFve3MHQJVsVJ1EbDl52q5FoeQiIiISHxq8RASExgiIiKxqsUJDOfAEBERkeiwAkNERCRSnANDRERE4sMhJCIiIiLxYAWGiIhIpCSCAIlgXAnF2PPNhQkMERGRWHEIiYiIiEg8WIEhIiISKd6FREREROLDISQiIiIi8WAFhoiISKRq8xASKzBERERiJZhoe0ILFy6ERCLB+PHjdW1FRUUYM2YM6tSpAwcHB/Tt2xe3bt168os8BBMYIiIikXpQgTF2exLHjx/H6tWr0bx5c732CRMmYMuWLfj++++xb98+3Lx5E3369DHBu9XHBIaIiIgMkp+fj7CwMKxZswbOzs669tzcXHz99ddYunQpXn75ZbRu3Rpr167F4cOH8dtvv5k0BiYwREREYmXCIaS8vDy9rbi4+KGXHTNmDF577TUEBwfrtSclJUGtVuu1N2nSBF5eXjhy5Igp3rEOExgiIiIRM9XwkVKphEKh0G0qlarC63333Xc4efJkhfszMzNhY2MDJycnvXZ3d3dkZmaa8m3zLiQiIiIC0tPTIZfLda+lUmmFx7z//vtISEiAra1tdYZXDiswREREYiUIptkAyOVyva2iBCYpKQlZWVlo1aoVrKysYGVlhX379iEqKgpWVlZwd3dHSUkJcnJy9M67desWPDw8TPrWWYEhIiISqepeB6Zr1674448/9NqGDh2KJk2aYMqUKVAqlbC2tkZiYiL69u0LAEhJSUFaWhoCAwONC/RfmMAQERFRpTg6OuL555/Xa7O3t0edOnV07cOGDcPEiRPh4uICuVyOsWPHIjAwEC+99JJJY2ECQ0REJFY18FlIy5Ytg4WFBfr27Yvi4mKEhITgiy++MO1FwASGiIhItCTass3YPoyxd+9evde2trZYuXIlVq5caVzHj8FJvERERCQ6rMAQERGJVQ0cQqouTGCIiIhEqjY/jZoJDBERkVj9Yx0Xo/oQIc6BISIiItFhBYaIiEikOIRERERE4lOLJ/FyCImIiIhEhxUYIiIikeIQEhEREYkP70IiIiIiEg9WYIiIiESKQ0hEREQkPrwLiYiIiEg8WIEhIiISKQ4hERERkfhohbLN2D5EiAkMERGRWHEODBEREZF4sAJDREQkUhKYYA6MSSKpfkxgiIiIxIor8RIRERGJByswREREIsXbqImIiEh8eBcSERERkXiwAkNERCRSEkGAxMhJuMaeby5MYIiIiMRK+/+bsX2IEIeQiIiISHRYgSEiIhIpDiERERGR+NTiu5CYwBAREYkVV+IlIiIiEg9WYIiIiESKK/HWQnv37kWXLl1w9+5dODk5PfS4Ro0aYfz48Rg/fny1xVYb9Qi/jf+MzoKLaymuJsvwxSf1kXLaztxhkRH69z2LdoFpaNAgDyXFlki+4IpvYgJw/YYCAODgUIxBb/2O1gE34Vq3ELl5Uhw5qsT62BYoLLQxc/RkiF4dktGrQzI8XO4BAFIznLHul1Y4muwFAIh6fwsCns3QOyf+QFMs+a5Dtcf61KnFQ0g1PoEJDw/H+vXrAQDW1tbw8vLC4MGD8dFHH8HK6snDDwoKQkZGBhSKsl+m69atw/jx45GTk6N33PHjx2Fvb//E16HH6/TGXYyceRPLpzbAhZN26D3iL8yPu4phHfyQe8fa3OHRE/J//ha2bPfDxUt1YGEpYOigU5g/azdGRvRAcbEV6rjcRx2XQqxZ2xpp6Qq4uRZg7OijcHG5j/mLOpo7fDJA1l17RP/8Iq5nKSCRCAhtexGqd3/FOwv74FqGCwBg88Em+HpbG905RSU1/s8P1XCi+C8oNDQUa9euRXFxMbZv344xY8bA2toa06ZNe+I+bWxs4OHh8djjXF1dn/gaVDl9Rt7GjjgX/Lqh7Bdd1JQGeLFrHkLeysbGFe5mjo6e1Cezu+q9XvJ5EDZ8+wN8n7mDs8nu+DPNCfMWddLtz8h0xPr/tsSHEw/BwkILrZZT9MTi8NmGeq/XbHkRvTqcx3ONsnQJTFGJFbLzWFU1NYm2bDO2DzESxW8IqVQKDw8PNGzYEKNHj0ZwcDA2b96Mu3fvYvDgwXB2doadnR26deuGS5cu6c77888/0aNHDzg7O8Pe3h7PPfcctm/fDqBsCEkikSAnJwd79+7F0KFDkZubC4lEAolEglmzZgEoG0KKjIwEAAwcOBD9+/fXi02tVqNu3bqIiYkBAGi1WqhUKnh7e0Mmk6FFixb44Ycfqv6HJFJW1lr4Ni/EyQOOujZBkODUAUc0a11oxsjI1Ozs1ACAe/nShx5jb1+CwkJrJi8iZiHRomvry7C1UeNc6t//AHn1hcvYsmg91n/8Pd594xik1qVmjPIp8mAIydhNhERRgfk3mUyGO3fuIDw8HJcuXcLmzZshl8sxZcoUdO/eHcnJybC2tsaYMWNQUlKC/fv3w97eHsnJyXBwcCjXX1BQECIjIzFjxgykpKQAQIXHhYWF4c0330R+fr5u/86dO1FYWIjevXsDAFQqFf773/8iOjoavr6+2L9/P95++224urqiU6dO5foEgOLiYhQXF+te5+XlGf0zEgu5iwaWVkDOX/r/Kd69bQWlT/FDziKxkUgEjBp+AueSXfFnmlOFx8gdi/BWv7P45Vff6g2OTKKxZzZWTYqHjZUG94ut8fGaV3Et0xkAkHDCB7eyHXA71x7P1L+DUT2PQemeg0/WvGrmqEnMRJXACIKAxMRE7Ny5E926dUN8fDwOHTqEoKAgAEBsbCyUSiXi4+Px5ptvIi0tDX379oW/vz8AoHHjxhX2a2NjA4VCAYlE8shhpZCQENjb22PTpk0YNGgQACAuLg5vvPEGHB0dUVxcjAULFmDXrl0IDAzUXfPgwYNYvXr1QxMYlUqF2bNnP/HPhaimG/PuMTTyysEH0yr+g2UnK8GcGXuQlq7Af//XvJqjI1NIu6XAO6q+sLctQZeAVHw8aC/GRvbAtUxnbDnUVHfc1ZsuuJNrh8/f3wbPunm4eVtuxqifArV4ITtR1Gm3bt0KBwcH2Nraolu3bujfvz/Cw8NhZWWFtm3b6o6rU6cO/Pz8cP78eQDAuHHjMG/ePLRr1w4zZ87E77//blQcVlZW6NevH2JjYwEABQUF+PnnnxEWFgYAuHz5MgoLC/HKK6/AwcFBt8XExODKlSsP7XfatGnIzc3Vbenp6UbFKSZ52ZbQlAJOrvrlZOe6pbj7l6jya3qI90YeQ9sXbmDyJ6/g9p3yE+JlMjXmzdqN+/etMUfVCRqNKH4t0b+Uaixx4y8FLqa7YvXmF3H5Rh38p8sfFR6bfM0NANDANbc6Q3wqPXiUgLGbGIniN0WXLl1w+vRpXLp0Cffv38f69eshkUgee97w4cNx9epVDBo0CH/88QfatGmD5cuXGxVLWFgYEhMTkZWVhfj4eMhkMoSGhgIA8vPzAQDbtm3D6dOndVtycvIj58FIpVLI5XK9rbYoVVvg0u92CGh/T9cmkQho2T4fyUmc8CduAt4beQxBL6VjyifBuJVVfljWTlaCBbMSUaq2wKx5naFWW5ohTqoKEokAG6uKZ4f6NrgDALiTy+84PTlR/BPX3t4ePj4+em1NmzZFaWkpjh49qhtCunPnDlJSUtCsWTPdcUqlEqNGjcKoUaMwbdo0rFmzBmPHji13DRsbG2g0msfGEhQUBKVSiQ0bNuCXX37Bm2++CWvrslt9mzVrBqlUirS0tIcOF1F5P31ZF5Mi03HxjB1STpXdRm1rp8Wv37mYOzQywph3j6NLx1TMXtAZ9+9bw9npPgCgoNAaJSVWsJOVYP7s3bCVlmLxsk6ws1PrJvrm5kk5kVdE3n3jGH5LVuJWtgPsbNV4pc1lBPjexAcru8Ozbh5eaXMZR84pkVdgi2fq38HYvkdw+lI9XLlZx9yhix/XgREfX19f9OzZEyNGjMDq1avh6OiIqVOnon79+ujZsycAYPz48ejWrRueffZZ3L17F3v27EHTpk0r7K9Ro0bIz89HYmIiWrRoATs7O9jZVfyvg4EDByI6OhoXL17Enj17dO2Ojo6YNGkSJkyYAK1Wi/bt2yM3NxeHDh2CXC7HkCFDTP+DeArs2+wMRR0NBn+YCWfXUlw9J8PHYd7Iuc01YMSsR/eLAIBPFyTotS/5PBAJu5+BzzPZaOp3GwCwdvXPescMGdGrwooN1UxOjvfx8eA9qCMvREGRDa7cqIMPVnbHiQsN4OaUjzZNbuDNLn/AVlqKrLv22HfaG+t3tDJ32E8HAYCxt0GLM38RbwIDAGvXrsX777+P119/HSUlJejYsSO2b9+uq4hoNBqMGTMG169fh1wuR2hoKJYtW1ZhX0FBQRg1ahT69++PO3fuYObMmbpbqf8tLCwM8+fPR8OGDdGuXTu9fXPnzoWrqytUKhWuXr0KJycntGrVCh999JFJ3/vTZvPauti8tq65wyATCu359iP3/37W47HHkDgsin14xTkrxwFjI3tUYzS1iynmsBhyvkqlwk8//YQLFy5AJpMhKCgIixYtgp+fn+6YoqIifPDBB/juu+9QXFyMkJAQfPHFF3B3N+26XhJBEGnt6CmWl5cHhUKBzugJKwmrEE87yQv+5g6BqlFBA877eNqVqotwbMt05ObmVtmcxgd/J14OmAorS1uj+irVFGH3qYWVijc0NBQDBgzACy+8gNLSUnz00Uc4e/YskpOTdavWjx49Gtu2bcO6deugUCgQEREBCwsLHDp0yKg4/03UFRgiIqJaTYAJ5sBU/tAdO3bovV63bh3c3NyQlJSEjh07Ijc3F19//TXi4uLw8ssvAygbLWnatCl+++03vPTSS8bF+g+cJUdERCRWJlyJNy8vT2/75wKrD5ObW3YrvItL2U0XSUlJUKvVCA4O1h3TpEkTeHl54ciRIyZ960xgiIiICEqlEgqFQrepVKpHHq/VajF+/Hi0a9cOzz//PAAgMzMTNjY2cHJy0jvW3d0dmZmZJo2XQ0hERERipQXw+GXRHt8HgPT0dL05MFLpw59bBgBjxozB2bNncfDgQSMDeDJMYIiIiETKlHchGbKQakREBLZu3Yr9+/ejQYMGunYPDw+UlJQgJydHrwpz69atRz6q50lwCImIiIgqRRAEREREYNOmTdi9eze8vb319rdu3RrW1tZITEzUtaWkpCAtLU33jEBTYQWGiIhIrKp5Jd4xY8YgLi4OP//8MxwdHXXzWhQKBWQyGRQKBYYNG4aJEyfCxcUFcrkcY8eORWBgoEnvQAKYwBAREYlXNScwq1atAgB07txZr33t2rUIDw8HACxbtgwWFhbo27ev3kJ2psYEhoiIiCqlMmvf2traYuXKlVi5cmWVxsIEhoiISKz4MEciIiISHRPeRi02TGCIiIhEqrof5liT8DZqIiIiEh1WYIiIiMSKc2CIiIhIdLQCIDEyAdGKM4HhEBIRERGJDiswREREYsUhJCIiIhIfEyQwEGcCwyEkIiIiEh1WYIiIiMSKQ0hEREQkOloBRg8B8S4kIiIiourBCgwREZFYCdqyzdg+RIgJDBERkVhxDgwRERGJDufAEBEREYkHKzBERERixSEkIiIiEh0BJkhgTBJJteMQEhEREYkOKzBERERixSEkIiIiEh2tFoCR67hoxbkODIeQiIiISHRYgSEiIhIrDiERERGR6NTiBIZDSERERCQ6rMAQERGJVS1+lAATGCIiIpESBC0EI58mbez55sIEhoiISKwEwfgKCufAEBEREVUPVmCIiIjESjDBHBiRVmCYwBAREYmVVgtIjJzDItI5MBxCIiIiItFhBYaIiEisOIREREREYiNotRCMHEIS623UHEIiIiIi0WEFhoiISKw4hERERESioxUASe1MYDiERERERKLDCgwREZFYCQIAY9eBEWcFhgkMERGRSAlaAYKRQ0gCExgiIiKqVoIWxldgeBs1ERER1QIrV65Eo0aNYGtri7Zt2+LYsWPVHgMTGCIiIpEStIJJNkNs2LABEydOxMyZM3Hy5Em0aNECISEhyMrKqqJ3WTEmMERERGIlaE2zGWDp0qUYMWIEhg4dimbNmiE6Ohp2dnb45ptvquhNVoxzYGqgBxOqSqE2en0iqvkkmiJzh0DVqFTNfzc+7TTqsu90dUyONcXfiVKoAQB5eXl67VKpFFKpVK+tpKQESUlJmDZtmq7NwsICwcHBOHLkiHGBGIgJTA107949AMBBbDdzJFQtkn42dwRUnZLMHQBVl3v37kGhUFRJ3zY2NvDw8MDBTNP8nXBwcIBSqdRrmzlzJmbNmqXXdvv2bWg0Gri7u+u1u7u748KFCyaJpbKYwNRAnp6eSE9Ph6OjIyQSibnDqRZ5eXlQKpVIT0+HXC43dzhUxfh51x618bMWBAH37t2Dp6dnlV3D1tYWqampKCkpMUl/giCU+3vz7+pLTcMEpgaysLBAgwYNzB2GWcjl8lrzS474edcmte2zrqrKyz/Z2trC1ta2yq/zT3Xr1oWlpSVu3bql137r1i14eHhUaywcjCUiIqJKsbGxQevWrZGYmKhr02q1SExMRGBgYLXGwgoMERERVdrEiRMxZMgQtGnTBi+++CIiIyNRUFCAoUOHVmscTGCoRpBKpZg5c2aNH3Ml0+DnXXvws3769O/fH3/99RdmzJiBzMxMtGzZEjt27Cg3sbeqSQSxPgSBiIiIai3OgSEiIiLRYQJDREREosMEhoiIiESHCQyJVqNGjRAZGWnuMKgG2bt3LyQSCXJycswdSq1W2c+B32EyBhMYqlB4eDgkEgkWLlyo1x4fH1/tqwOvW7cOTk5O5dqPHz+OkSNHVmsstUV1ff7Xrl2DRCLB6dOnTdYnVd6Dz1kikcDGxgY+Pj6YM2cOSktLjeo3KCgIGRkZusXc+B2mqsAEhh7K1tYWixYtwt27d80dSoVcXV1hZ2dn7jCeWjXp8zfVculUXmhoKDIyMnDp0iV88MEHmDVrFj799FOj+nzwnJ7HJbv8DpMxmMDQQwUHB8PDwwMqleqhxxw8eBAdOnSATCaDUqnEuHHjUFBQoNufkZGB1157DTKZDN7e3oiLiytXNl66dCn8/f1hb28PpVKJ9957D/n5+QDKStFDhw5Fbm6u7l+KDx4u9s9+Bg4ciP79++vFplarUbduXcTExAAoWy1SpVLB29sbMpkMLVq0wA8//GCCn9TTyRSfv0QiQXx8vN45Tk5OWLduHQDA29sbABAQEACJRILOnTsDKKsM9OrVC/Pnz4enpyf8/PwAAN9++y3atGkDR0dHeHh4YODAgcjKyjLdm66FpFIpPDw80LBhQ4wePRrBwcHYvHkz7t69i8GDB8PZ2Rl2dnbo1q0bLl26pDvvzz//RI8ePeDs7Ax7e3s899xz2L697MGC/xxC4neYqgoTGHooS0tLLFiwAMuXL8f169fL7b9y5QpCQ0PRt29f/P7779iwYQMOHjyIiIgI3TGDBw/GzZs3sXfvXvz444/48ssvy/3BsbCwQFRUFM6dO4f169dj9+7dmDx5MoCyUnRkZCTkcjkyMjKQkZGBSZMmlYslLCwMW7Zs0SU+ALBz504UFhaid+/eAACVSoWYmBhER0fj3LlzmDBhAt5++23s27fPJD+vp40pPv/HOXbsGABg165dyMjIwE8//aTbl5iYiJSUFCQkJGDr1q0Ayv6gzZ07F2fOnEF8fDyuXbuG8PBw494o6ZHJZCgpKUF4eDhOnDiBzZs348iRIxAEAd27d4darQYAjBkzBsXFxdi/fz/++OMPLFq0CA4ODuX643eYqoxAVIEhQ4YIPXv2FARBEF566SXhnXfeEQRBEDZt2iQ8+M9m2LBhwsiRI/XOO3DggGBhYSHcv39fOH/+vABAOH78uG7/pUuXBADCsmXLHnrt77//XqhTp47u9dq1awWFQlHuuIYNG+r6UavVQt26dYWYmBjd/rfeekvo37+/IAiCUFRUJNjZ2QmHDx/W62PYsGHCW2+99egfRi1kis9fEAQBgLBp0ya9YxQKhbB27VpBEAQhNTVVACCcOnWq3PXd3d2F4uLiR8Z5/PhxAYBw7949QRAEYc+ePQIA4e7duwa+49rpn5+zVqsVEhISBKlUKvTq1UsAIBw6dEh37O3btwWZTCZs3LhREARB8Pf3F2bNmlVhv//+HPgdpqrARwnQYy1atAgvv/xyuX81nTlzBr///jtiY2N1bYIgQKvVIjU1FRcvXoSVlRVatWql2+/j4wNnZ2e9fnbt2gWVSoULFy4gLy8PpaWlKCoqQmFhYaXHx62srNCvXz/ExsZi0KBBKCgowM8//4zvvvsOAHD58mUUFhbilVde0TuvpKQEAQEBBv08apsn/fybNm1q1HX9/f1hY2Oj15aUlIRZs2bhzJkzuHv3LrRaLQAgLS0NzZo1M+p6tdXWrVvh4OAAtVoNrVaLgQMHok+fPti6dSvatm2rO65OnTrw8/PD+fPnAQDjxo3D6NGj8euvvyI4OBh9+/ZF8+bNnzgOfofJUExg6LE6duyIkJAQTJs2Ta9cn5+fj3fffRfjxo0rd46XlxcuXrz42L6vXbuG119/HaNHj8b8+fPh4uKCgwcPYtiwYSgpKTFogl9YWBg6deqErKwsJCQkQCaTITQ0VBcrAGzbtg3169fXO4/PaHm0J/38gbI5MMK/nlbyYAjicezt7fVeFxQUICQkBCEhIYiNjYWrqyvS0tIQEhLCSb5G6NKlC1atWgUbGxt4enrCysoKmzdvfux5w4cPR0hICLZt24Zff/0VKpUKS5YswdixY584Fn6HyRBMYKhSFi5ciJYtW+omUwJAq1atkJycDB8fnwrP8fPzQ2lpKU6dOoXWrVsDKPtX1D/vaklKSoJWq8WSJUtgYVE2JWvjxo16/djY2ECj0Tw2xqCgICiVSmzYsAG//PIL3nzzTVhbWwMAmjVrBqlUirS0NHTq1MmwN09P9PkDZXeZZGRk6F5funQJhYWFutcPKiyV+XwvXLiAO3fuYOHChVAqlQCAEydOGPxeSJ+9vX25z7Bp06YoLS3F0aNHERQUBAC4c+cOUlJS9CpdSqUSo0aNwqhRozBt2jSsWbOmwgSG32GqCkxgqFL8/f0RFhaGqKgoXduUKVPw0ksvISIiAsOHD4e9vT2Sk5ORkJCAFStWoEmTJggODsbIkSOxatUqWFtb44MPPoBMJtPdXunj4wO1Wo3ly5ejR48eOHToEKKjo/Wu3ahRI+Tn5yMxMREtWrSAnZ3dQyszAwcORHR0NC5evIg9e/bo2h0dHTFp0iRMmDABWq0W7du3R25uLg4dOgS5XI4hQ4ZUwU/t6fEknz8AvPzyy1ixYgUCAwOh0WgwZcoU3R8kAHBzc4NMJsOOHTvQoEED2Nra6tYO+TcvLy/Y2Nhg+fLlGDVqFM6ePYu5c+dW7RuvpXx9fdGzZ0+MGDECq1evhqOjI6ZOnYr69eujZ8+eAIDx48ejW7duePbZZ3H37l3s2bPnocOG/A5TlTDzHByqof45ue+B1NRUwcbGRvjnfzbHjh0TXnnlFcHBwUGwt7cXmjdvLsyfP1+3/+bNm0K3bt0EqVQqNGzYUIiLixPc3NyE6Oho3TFLly4V6tWrJ8hkMiEkJESIiYkpNxFz1KhRQp06dQQAwsyZMwVB0J8A+EBycrIAQGjYsKGg1Wr19mm1WiEyMlLw8/MTrK2tBVdXVyEkJETYt2+fcT+sp5CpPv8bN24Ir776qmBvby/4+voK27dv15vEKwiCsGbNGkGpVAoWFhZCp06dHnp9QRCEuLg4oVGjRoJUKhUCAwOFzZs3600C5iRewzzs5ywIgpCdnS0MGjRIUCgUuu/mxYsXdfsjIiKEZ555RpBKpYKrq6swaNAg4fbt24IgVPw58DtMpiYRhH8NUBNVoevXr0OpVGLXrl3o2rWrucMhIiKRYgJDVWr37t3Iz8+Hv78/MjIyMHnyZNy4cQMXL17UG0ogIiIyBOfAUJVSq9X46KOPcPXqVTg6OiIoKAixsbFMXoiIyCiswBAREZHo8FECREREJDpMYIiIiEh0mMAQERGR6DCBISIiItFhAkNEFQoPD0evXr10rzt37ozx48dXexx79+6FRCJBTk7OQ4+RSCSIj4+vdJ+zZs1Cy5YtjYrr2rVrkEgkOH36tFH9ENGTYQJDJCLh4eGQSCSQSCSwsbGBj48P5syZg9LS0iq/9k8//VTppfsrk3QQERmD68AQiUxoaCjWrl2L4uJibN++HWPGjIG1tTWmTZtW7tiSkhLdAxON5eLiYpJ+iIhMgRUYIpGRSqXw8PBAw4YNMXr0aAQHB2Pz5s0A/h72mT9/Pjw9PXVPj05PT0e/fv3g5OQEFxcX9OzZE9euXdP1qdFoMHHiRDg5OaFOnTqYPHky/r1E1L+HkIqLizFlyhQolUpIpVL4+Pjg66+/xrVr19ClSxcAgLOzMyQSCcLDwwEAWq0WKpUK3t7ekMlkaNGiBX744Qe962zfvh3PPvssZDIZunTpohdnZU2ZMgXPPvss7Ozs0LhxY0yfPh1qtbrccatXr4ZSqYSdnR369euH3Nxcvf1fffUVmjZtCltbWzRp0gRffPGFwbEQUdVgAkMkcjKZDCUlJbrXiYmJSElJQUJCArZu3Qq1Wo2QkBA4OjriwIEDOHToEBwcHBAaGqo7b8mSJVi3bh2++eYbHDx4ENnZ2di0adMjrzt48GD873//Q1RUFM6fP4/Vq1fDwcEBSqUSP/74IwAgJSUFGRkZ+PzzzwEAKpUKMTExiI6Oxrlz5zBhwgS8/fbb2LdvH4CyRKtPnz7o0aMHTp8+jeHDh2Pq1KkG/0wcHR2xbt06JCcn4/PPP8eaNWuwbNkyvWMuX76MjRs3YsuWLdixYwdOnTqF9957T7c/NjYWM2bMwPz583H+/HksWLAA06dPx/r16w2Oh4iqgBkfJElEBvrn04O1Wq2QkJAgSKVSYdKkSbr97u7uQnFxse6cb7/9VvDz89N7sm9xcbEgk8mEnTt3CoIgCPXq1RMWL16s269Wq4UGDRroPam4U6dOwvvvvy8IgiCkpKQIAISEhIQK46zoacRFRUWCnZ2dcPjwYb1jhw0bJrz11luCIAjCtGnThGbNmuntnzJlymOfMA1A2LRp00P3f/rpp0Lr1q11r2fOnClYWloK169f17X98ssvgoWFhZCRkSEIgiA888wzQlxcnF4/c+fOFQIDAwVBKHs6N/7xJGwiql6cA0MkMlu3boWDgwPUajW0Wi0GDhyIWbNm6fb7+/vrzXs5c+YMLl++DEdHR71+ioqKcOXKFeTm5iIjIwNt27bV7bOyskKbNm3KDSM9cPr0aVhaWqJTp06Vjvvy5csoLCzEK6+8otdeUlKCgIAAAMD58+f14gCAwMDASl/jgQ0bNiAqKgpXrlxBfn4+SktLIZfL9Y7x8vJC/fr19a6j1WqRkpICR0dHXLlyBcOGDcOIESN0x5SWlkKhUBgcDxGZHhMYIpHp0qULVq1aBRsbG3h6esLKSv9rbG9vr/c6Pz8frVu3RmxsbLm+XF1dnygGmUxm8Dn5+fkAgG3btuklDkDZvB5TOXLkCMLCwjB79myEhIRAoVDgu+++w5IlSwyOdc2aNeUSKktLS5PFSkRPjgkMkcjY29vDx8en0se3atUKGzZsgJubW7kqxAP16tXD0aNH0bFjRwBllYakpCS0atWqwuP9/f2h1Wqxb98+BAcHl9v/oAKk0Wh0bc2aNYNUKkVaWtpDKzdNmzbVTUh+4Lfffnv8m/yHw4cPo2HDhvj44491bX/++We549LS0nDz5k14enrqrmNhYQE/Pz+4u7vD09MTV69eRVhYmEHXJ6LqwUm8RE+5sLAw1K1bFz179sSBAweQmpqKvXv3Yty4cbh+/ToA4P3338fChQsRHx+PCxcu4L333nvkGi6NGjXCkCFD8M477yA+Pl7X58aNGwEADRs2hEQiwdatW/HXX38hPz8fjo6OmDRpEiZMmID169fjypUrOHnyJJYvX66bGDtq1ChcunQJH374IVJSUhAXF4d169YZ9H59fX2RlpaG7777DleuXEFUVFSFE5JtbW0xZMgQnDlzBgcOHMC4cePQr18/eHh4AABmz54NlUqFqKgoXLx4EX/88QfWrl2LpUuXGhQPEVUNJjBETzk7Ozvs378fXl5e6NOnD5o2bYphw4ahqKhIV5H54IMPMGjQIAwZMgSBgYFwdHRE7969H9nvqlWr8J///AfvvfcemjRpghEjRqCgoAAAUL9+fcyePRtTp06Fu7s7IiIiAABz587F9OnToVKp0LRpU4SGhmLbtm3w9vYGUDYv5ccff0R8fDxatGiB6OhoLFiwwKD3+8Ybb2DChAmIiIhAy5YtcfjwYUyfPr3ccT4+PujTpw+6d++OV199Fc2bN9e7TXr48OH46quvsHbtWvj7+6NTp05Yt26dLlYiMi+J8LBZekREREQ1FCswREREJDpMYIiIiEh0mMAQERGR6DCBISIiItFhAkNERESiwwSGiIiIRIcJDBEREYkOExgiIiISHSYwREREJDpMYIiIiEh0mMAQERGR6DCBISIiItH5P3nM0x7vxhOwAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_pred = nb.predict(X_test_tfidf)\n",
    "acc_score = accuracy_score(y_test, y_pred)\n",
    "\n",
    "print(\"Accuracy of Naive Bayes' model - %.3f\"%(acc_score*100), \"%\")\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "disp = ConfusionMatrixDisplay(cm, display_labels=['Negative', 'Neutral', 'Positive'])\n",
    "disp.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "dd87d003",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "    negative       0.07      1.00      0.12         2\n",
      "     neutral       1.00      0.81      0.89       173\n",
      "    positive       0.61      0.67      0.64        52\n",
      "\n",
      "    accuracy                           0.78       227\n",
      "   macro avg       0.56      0.83      0.55       227\n",
      "weighted avg       0.90      0.78      0.83       227\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_pred, y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3c2a31d3",
   "metadata": {},
   "source": [
    "### Word2vec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "602476ea",
   "metadata": {},
   "outputs": [],
   "source": [
    "wv = api.load('word2vec-google-news-300')\n",
    "\n",
    "tokenized_sents = [[word for word in sent.split()] for sent in df['cleaned']]\n",
    "max_length = max(len(arr) for arr in tokenized_sents)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "59fd9ae1",
   "metadata": {},
   "source": [
    "### Tokenize the dataset \n",
    "\n",
    "The text sequences is padded to match with the length of the longest sentence.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "234f230f",
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = Tokenizer()\n",
    "tokenizer.fit_on_texts(tokenized_sents)\n",
    "\n",
    "input_sents = tokenizer.texts_to_sequences(tokenized_sents)\n",
    "input_sents = pad_sequences(input_sents, maxlen=max_length, padding='post')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "94426cd0",
   "metadata": {},
   "source": [
    "### Create embeddings for words from word2vec dict\n",
    "\n",
    "The length of the word embeddings is set to be 300.\n",
    "\n",
    "Overall there are 6302 unique words in the dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "56ccf905",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6302, 300)\n"
     ]
    }
   ],
   "source": [
    "embeddings = np.zeros((len(tokenizer.word_index)+1, 300))\n",
    "for w, i in tokenizer.word_index.items():\n",
    "    if w in wv and i < embeddings.shape[0]:\n",
    "        embeddings[i] = wv[w]\n",
    "        \n",
    "print(embeddings.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "86450484",
   "metadata": {},
   "source": [
    "### Convert categorical label into one-hot-encoding & data split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "32412d4a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set length - 1811 \n",
      "Validation set length - 226 \n",
      "Test set length - 227\n"
     ]
    }
   ],
   "source": [
    "Y = pd.get_dummies(df[1]).values\n",
    "X_train_1, X_valtest_1, y_train_1, y_valtest_1 = train_test_split(input_sents, Y, test_size=0.2, stratify=df[1], random_state=13)\n",
    "X_val_1, X_test_1, y_val_1, y_test_1 = train_test_split(X_valtest_1, y_valtest_1, test_size=0.5, stratify=y_valtest_1, random_state=13)\n",
    "\n",
    "print(\"Training set length - {} \\nValidation set length - {} \\nTest set length - {}\".format(len(X_train_1), len(X_val_1), len(X_test_1)))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c2345ad9",
   "metadata": {},
   "source": [
    "### Model 2 - LSTM using word2vec features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "d35d3bf4",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-12-13 11:59:03.548819: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding (Embedding)       (None, 52, 300)           1890600   \n",
      "                                                                 \n",
      " lstm (LSTM)                 (None, 52, 32)            42624     \n",
      "                                                                 \n",
      " lstm_1 (LSTM)               (None, 52, 16)            3136      \n",
      "                                                                 \n",
      " lstm_2 (LSTM)               (None, 8)                 800       \n",
      "                                                                 \n",
      " dense (Dense)               (None, 3)                 27        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1,937,187\n",
      "Trainable params: 46,587\n",
      "Non-trainable params: 1,890,600\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "\n",
    "model.add(Embedding(embeddings.shape[0], \n",
    "                    embeddings.shape[1], \n",
    "                    weights = [embeddings],\n",
    "                    input_length=52,\n",
    "                    trainable = False))\n",
    "model.add(LSTM(32, activation='relu', dropout=0.25,\n",
    "               recurrent_dropout=0.25, return_sequences=True))\n",
    "model.add(LSTM(16, activation='relu', dropout=0.25,\n",
    "               recurrent_dropout=0.25, return_sequences=True))\n",
    "model.add(LSTM(8, activation='relu'))\n",
    "model.add(Dense(3, activation='softmax'))\n",
    "\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=Adam(learning_rate=0.0001),\n",
    "              metrics=['accuracy'])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "a30c8014",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "453/453 [==============================] - 34s 63ms/step - loss: 1.0225 - accuracy: 0.6113 - val_loss: 0.9385 - val_accuracy: 0.6150\n",
      "Epoch 2/100\n",
      "453/453 [==============================] - 35s 77ms/step - loss: 0.9079 - accuracy: 0.6151 - val_loss: 0.8448 - val_accuracy: 0.6150\n",
      "Epoch 3/100\n",
      "453/453 [==============================] - 27s 59ms/step - loss: 0.8279 - accuracy: 0.6146 - val_loss: 0.8107 - val_accuracy: 0.6150\n",
      "Epoch 4/100\n",
      "453/453 [==============================] - 25s 56ms/step - loss: 0.8011 - accuracy: 0.6146 - val_loss: 0.7778 - val_accuracy: 0.6150\n",
      "Epoch 5/100\n",
      "453/453 [==============================] - 31s 69ms/step - loss: 0.7762 - accuracy: 0.6146 - val_loss: 0.7553 - val_accuracy: 0.6150\n",
      "Epoch 6/100\n",
      "453/453 [==============================] - 40s 88ms/step - loss: 0.7585 - accuracy: 0.6173 - val_loss: 0.7238 - val_accuracy: 0.6195\n",
      "Epoch 7/100\n",
      "453/453 [==============================] - 40s 89ms/step - loss: 0.7330 - accuracy: 0.6372 - val_loss: 0.6850 - val_accuracy: 0.6327\n",
      "Epoch 8/100\n",
      "453/453 [==============================] - 31s 69ms/step - loss: 0.7088 - accuracy: 0.6466 - val_loss: 0.6597 - val_accuracy: 0.6549\n",
      "Epoch 9/100\n",
      "453/453 [==============================] - 31s 69ms/step - loss: 0.6857 - accuracy: 0.6692 - val_loss: 0.6535 - val_accuracy: 0.6726\n",
      "Epoch 10/100\n",
      "453/453 [==============================] - 29s 64ms/step - loss: 0.6641 - accuracy: 0.6853 - val_loss: 0.6390 - val_accuracy: 0.6903\n",
      "Epoch 11/100\n",
      "453/453 [==============================] - 31s 69ms/step - loss: 0.6471 - accuracy: 0.7123 - val_loss: 0.6939 - val_accuracy: 0.7345\n",
      "Epoch 12/100\n",
      "453/453 [==============================] - 29s 64ms/step - loss: 0.8610 - accuracy: 0.6880 - val_loss: 0.7825 - val_accuracy: 0.6726\n",
      "Epoch 13/100\n",
      "453/453 [==============================] - 33s 73ms/step - loss: 0.7552 - accuracy: 0.6869 - val_loss: 0.6880 - val_accuracy: 0.6593\n",
      "Epoch 14/100\n",
      "453/453 [==============================] - 34s 74ms/step - loss: 0.7210 - accuracy: 0.6858 - val_loss: 0.6829 - val_accuracy: 0.6549\n",
      "Epoch 15/100\n",
      "453/453 [==============================] - 31s 69ms/step - loss: 0.7249 - accuracy: 0.6875 - val_loss: 0.6853 - val_accuracy: 0.6549\n",
      "Epoch 16/100\n",
      "453/453 [==============================] - 30s 67ms/step - loss: 0.6959 - accuracy: 0.6875 - val_loss: 0.6772 - val_accuracy: 0.6549\n",
      "Epoch 17/100\n",
      "453/453 [==============================] - 26s 58ms/step - loss: 0.6990 - accuracy: 0.6886 - val_loss: 0.6731 - val_accuracy: 0.6726\n",
      "Epoch 18/100\n",
      "453/453 [==============================] - 29s 63ms/step - loss: 0.6915 - accuracy: 0.6902 - val_loss: 0.6664 - val_accuracy: 0.6770\n",
      "Epoch 19/100\n",
      "453/453 [==============================] - 27s 60ms/step - loss: 0.6997 - accuracy: 0.6985 - val_loss: 0.6700 - val_accuracy: 0.6726\n",
      "Epoch 20/100\n",
      "453/453 [==============================] - 27s 61ms/step - loss: 0.6958 - accuracy: 0.6941 - val_loss: 0.6680 - val_accuracy: 0.6726\n",
      "Epoch 21/100\n",
      "453/453 [==============================] - 26s 58ms/step - loss: 0.6828 - accuracy: 0.7007 - val_loss: 0.6606 - val_accuracy: 0.6726\n",
      "Epoch 22/100\n",
      "453/453 [==============================] - 27s 59ms/step - loss: 0.6791 - accuracy: 0.7024 - val_loss: 0.6488 - val_accuracy: 0.6858\n",
      "Epoch 23/100\n",
      "453/453 [==============================] - 28s 63ms/step - loss: 0.6818 - accuracy: 0.7029 - val_loss: 0.6528 - val_accuracy: 0.7124\n",
      "Epoch 24/100\n",
      "453/453 [==============================] - 27s 59ms/step - loss: 0.6682 - accuracy: 0.7035 - val_loss: 0.6400 - val_accuracy: 0.7124\n",
      "Epoch 25/100\n",
      "453/453 [==============================] - 27s 59ms/step - loss: 0.6551 - accuracy: 0.7134 - val_loss: 0.6253 - val_accuracy: 0.7212\n",
      "Epoch 26/100\n",
      "453/453 [==============================] - 26s 58ms/step - loss: 0.6728 - accuracy: 0.7211 - val_loss: 0.6320 - val_accuracy: 0.7257\n",
      "Epoch 27/100\n",
      "453/453 [==============================] - 31s 68ms/step - loss: 0.6433 - accuracy: 0.7173 - val_loss: 0.6263 - val_accuracy: 0.7345\n",
      "Epoch 28/100\n",
      "453/453 [==============================] - 29s 64ms/step - loss: 0.6319 - accuracy: 0.7245 - val_loss: 0.6110 - val_accuracy: 0.7345\n",
      "Epoch 29/100\n",
      "453/453 [==============================] - 27s 60ms/step - loss: 0.6187 - accuracy: 0.7272 - val_loss: 0.6007 - val_accuracy: 0.7345\n",
      "Epoch 30/100\n",
      "453/453 [==============================] - 32s 71ms/step - loss: 0.6213 - accuracy: 0.7372 - val_loss: 0.5970 - val_accuracy: 0.7345\n",
      "Epoch 31/100\n",
      "453/453 [==============================] - 29s 63ms/step - loss: 0.6171 - accuracy: 0.7394 - val_loss: 0.5937 - val_accuracy: 0.7345\n",
      "Epoch 32/100\n",
      "453/453 [==============================] - 39s 86ms/step - loss: 0.5978 - accuracy: 0.7421 - val_loss: 0.5809 - val_accuracy: 0.7389\n",
      "Epoch 33/100\n",
      "453/453 [==============================] - 52s 115ms/step - loss: 0.6077 - accuracy: 0.7443 - val_loss: 0.5814 - val_accuracy: 0.7389\n",
      "Epoch 34/100\n",
      "453/453 [==============================] - 49s 108ms/step - loss: 0.5866 - accuracy: 0.7488 - val_loss: 0.5691 - val_accuracy: 0.7389\n",
      "Epoch 35/100\n",
      "453/453 [==============================] - 51s 113ms/step - loss: 0.5864 - accuracy: 0.7515 - val_loss: 0.5620 - val_accuracy: 0.7478\n",
      "Epoch 36/100\n",
      "453/453 [==============================] - 51s 113ms/step - loss: 0.5720 - accuracy: 0.7543 - val_loss: 0.5619 - val_accuracy: 0.7478\n",
      "Epoch 37/100\n",
      "453/453 [==============================] - 28s 63ms/step - loss: 0.5599 - accuracy: 0.7581 - val_loss: 0.5480 - val_accuracy: 0.7478\n",
      "Epoch 38/100\n",
      "453/453 [==============================] - 33s 73ms/step - loss: 0.5438 - accuracy: 0.7631 - val_loss: 0.5351 - val_accuracy: 0.7345\n",
      "Epoch 39/100\n",
      "453/453 [==============================] - 36s 80ms/step - loss: 0.5709 - accuracy: 0.7559 - val_loss: 0.5252 - val_accuracy: 0.7345\n",
      "Epoch 40/100\n",
      "453/453 [==============================] - 37s 82ms/step - loss: 0.5296 - accuracy: 0.7554 - val_loss: 0.5183 - val_accuracy: 0.7257\n",
      "Epoch 41/100\n",
      "453/453 [==============================] - 42s 92ms/step - loss: 0.5233 - accuracy: 0.7604 - val_loss: 0.5082 - val_accuracy: 0.7345\n",
      "Epoch 42/100\n",
      "453/453 [==============================] - 39s 86ms/step - loss: 0.5071 - accuracy: 0.7615 - val_loss: 0.5009 - val_accuracy: 0.7522\n",
      "Epoch 43/100\n",
      "453/453 [==============================] - 34s 75ms/step - loss: 0.4847 - accuracy: 0.7692 - val_loss: 0.4874 - val_accuracy: 0.7434\n",
      "Epoch 44/100\n",
      "453/453 [==============================] - 41s 90ms/step - loss: 0.4781 - accuracy: 0.7780 - val_loss: 0.4878 - val_accuracy: 0.7478\n",
      "Epoch 45/100\n",
      "453/453 [==============================] - 35s 77ms/step - loss: 0.4672 - accuracy: 0.7813 - val_loss: 0.4768 - val_accuracy: 0.7832\n",
      "Epoch 46/100\n",
      "453/453 [==============================] - 33s 72ms/step - loss: 0.4640 - accuracy: 0.7841 - val_loss: 0.4580 - val_accuracy: 0.7832\n",
      "Epoch 47/100\n",
      "453/453 [==============================] - 33s 73ms/step - loss: 0.4473 - accuracy: 0.7907 - val_loss: 0.4485 - val_accuracy: 0.7832\n",
      "Epoch 48/100\n",
      "453/453 [==============================] - 33s 74ms/step - loss: 0.4276 - accuracy: 0.7951 - val_loss: 0.4382 - val_accuracy: 0.7920\n",
      "Epoch 49/100\n",
      "453/453 [==============================] - 29s 65ms/step - loss: 0.4287 - accuracy: 0.7962 - val_loss: 0.4362 - val_accuracy: 0.7920\n",
      "Epoch 50/100\n",
      "453/453 [==============================] - 32s 70ms/step - loss: 0.4160 - accuracy: 0.7973 - val_loss: 0.4250 - val_accuracy: 0.7920\n",
      "Epoch 51/100\n",
      "453/453 [==============================] - 32s 70ms/step - loss: 0.4060 - accuracy: 0.8007 - val_loss: 0.4056 - val_accuracy: 0.8009\n",
      "Epoch 52/100\n",
      "453/453 [==============================] - 31s 67ms/step - loss: 0.3874 - accuracy: 0.8123 - val_loss: 0.3877 - val_accuracy: 0.8097\n",
      "Epoch 53/100\n",
      "453/453 [==============================] - 31s 68ms/step - loss: 0.3832 - accuracy: 0.8216 - val_loss: 0.3823 - val_accuracy: 0.8142\n",
      "Epoch 54/100\n",
      "453/453 [==============================] - 31s 68ms/step - loss: 0.3837 - accuracy: 0.8216 - val_loss: 0.3771 - val_accuracy: 0.8186\n",
      "Epoch 55/100\n",
      "453/453 [==============================] - 31s 68ms/step - loss: 0.3868 - accuracy: 0.8244 - val_loss: 0.3771 - val_accuracy: 0.8186\n",
      "Epoch 56/100\n",
      "453/453 [==============================] - 32s 70ms/step - loss: 0.3608 - accuracy: 0.8255 - val_loss: 0.3880 - val_accuracy: 0.8274\n",
      "Epoch 57/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "453/453 [==============================] - 32s 71ms/step - loss: 0.3526 - accuracy: 0.8332 - val_loss: 0.3741 - val_accuracy: 0.8186\n",
      "Epoch 58/100\n",
      "453/453 [==============================] - 33s 72ms/step - loss: 0.3570 - accuracy: 0.8327 - val_loss: 0.3767 - val_accuracy: 0.8274\n",
      "Epoch 59/100\n",
      "453/453 [==============================] - 32s 72ms/step - loss: 0.3426 - accuracy: 0.8327 - val_loss: 0.3619 - val_accuracy: 0.8363\n",
      "Epoch 60/100\n",
      "453/453 [==============================] - 34s 75ms/step - loss: 0.3420 - accuracy: 0.8371 - val_loss: 0.3623 - val_accuracy: 0.8230\n",
      "Epoch 61/100\n",
      "453/453 [==============================] - 41s 90ms/step - loss: 0.3471 - accuracy: 0.8310 - val_loss: 0.3514 - val_accuracy: 0.8363\n",
      "Epoch 62/100\n",
      "453/453 [==============================] - 43s 94ms/step - loss: 0.3312 - accuracy: 0.8377 - val_loss: 0.3746 - val_accuracy: 0.8274\n",
      "Epoch 63/100\n",
      "453/453 [==============================] - 40s 89ms/step - loss: 0.3331 - accuracy: 0.8338 - val_loss: 0.3513 - val_accuracy: 0.8363\n",
      "Epoch 64/100\n",
      "453/453 [==============================] - 41s 91ms/step - loss: 0.3193 - accuracy: 0.8426 - val_loss: 0.3492 - val_accuracy: 0.8319\n",
      "Epoch 65/100\n",
      "453/453 [==============================] - 38s 84ms/step - loss: 0.3141 - accuracy: 0.8476 - val_loss: 0.3783 - val_accuracy: 0.8230\n",
      "Epoch 66/100\n",
      "453/453 [==============================] - 39s 87ms/step - loss: 0.3180 - accuracy: 0.8426 - val_loss: 0.3487 - val_accuracy: 0.8230\n",
      "Epoch 67/100\n",
      "453/453 [==============================] - 35s 78ms/step - loss: 0.3033 - accuracy: 0.8493 - val_loss: 0.4193 - val_accuracy: 0.8009\n",
      "Epoch 68/100\n",
      "453/453 [==============================] - 39s 85ms/step - loss: 0.3295 - accuracy: 0.8366 - val_loss: 0.3553 - val_accuracy: 0.8319\n",
      "Epoch 69/100\n",
      "453/453 [==============================] - 44s 96ms/step - loss: 0.3082 - accuracy: 0.8470 - val_loss: 0.3436 - val_accuracy: 0.8274\n",
      "Epoch 70/100\n",
      "453/453 [==============================] - 38s 85ms/step - loss: 0.3045 - accuracy: 0.8487 - val_loss: 0.3578 - val_accuracy: 0.8274\n",
      "Epoch 71/100\n",
      "453/453 [==============================] - 34s 74ms/step - loss: 0.2998 - accuracy: 0.8575 - val_loss: 0.3511 - val_accuracy: 0.8274\n",
      "Epoch 72/100\n",
      "453/453 [==============================] - 34s 74ms/step - loss: 0.3192 - accuracy: 0.8437 - val_loss: 0.3446 - val_accuracy: 0.8363\n",
      "Epoch 73/100\n",
      "453/453 [==============================] - 32s 70ms/step - loss: 0.3003 - accuracy: 0.8498 - val_loss: 0.3581 - val_accuracy: 0.8319\n",
      "Epoch 74/100\n",
      "453/453 [==============================] - 34s 76ms/step - loss: 0.3032 - accuracy: 0.8542 - val_loss: 0.3363 - val_accuracy: 0.8230\n",
      "Epoch 75/100\n",
      "453/453 [==============================] - 32s 71ms/step - loss: 0.2964 - accuracy: 0.8603 - val_loss: 0.3342 - val_accuracy: 0.8407\n",
      "Epoch 76/100\n",
      "453/453 [==============================] - 31s 67ms/step - loss: 0.2864 - accuracy: 0.8620 - val_loss: 0.3426 - val_accuracy: 0.8761\n",
      "Epoch 77/100\n",
      "453/453 [==============================] - 31s 69ms/step - loss: 0.2723 - accuracy: 0.8780 - val_loss: 0.3381 - val_accuracy: 0.8319\n",
      "Epoch 78/100\n",
      "453/453 [==============================] - 36s 79ms/step - loss: 0.2898 - accuracy: 0.8570 - val_loss: 0.3228 - val_accuracy: 0.8407\n",
      "Epoch 79/100\n",
      "453/453 [==============================] - 34s 74ms/step - loss: 0.2712 - accuracy: 0.8691 - val_loss: 0.3194 - val_accuracy: 0.8717\n",
      "Epoch 80/100\n",
      "453/453 [==============================] - 33s 73ms/step - loss: 0.2776 - accuracy: 0.8708 - val_loss: 0.3247 - val_accuracy: 0.8363\n",
      "Epoch 81/100\n",
      "453/453 [==============================] - 34s 74ms/step - loss: 0.7709 - accuracy: 0.8564 - val_loss: 0.3189 - val_accuracy: 0.8628\n",
      "Epoch 82/100\n",
      "453/453 [==============================] - 34s 75ms/step - loss: 0.2742 - accuracy: 0.8730 - val_loss: 0.3197 - val_accuracy: 0.8540\n",
      "Epoch 83/100\n",
      "453/453 [==============================] - 35s 76ms/step - loss: 0.2689 - accuracy: 0.8730 - val_loss: 0.3250 - val_accuracy: 0.8628\n",
      "Epoch 84/100\n",
      "453/453 [==============================] - 36s 80ms/step - loss: 0.2812 - accuracy: 0.8758 - val_loss: 0.3218 - val_accuracy: 0.8451\n",
      "Epoch 85/100\n",
      "453/453 [==============================] - 38s 83ms/step - loss: 0.2552 - accuracy: 0.8791 - val_loss: 0.3289 - val_accuracy: 0.8451\n",
      "Epoch 86/100\n",
      "453/453 [==============================] - 39s 86ms/step - loss: 0.2452 - accuracy: 0.8918 - val_loss: 0.3301 - val_accuracy: 0.8761\n",
      "Epoch 87/100\n",
      "453/453 [==============================] - 38s 83ms/step - loss: 0.2568 - accuracy: 0.8807 - val_loss: 0.3193 - val_accuracy: 0.8628\n",
      "Epoch 88/100\n",
      "453/453 [==============================] - 35s 77ms/step - loss: 0.2596 - accuracy: 0.8785 - val_loss: 0.3258 - val_accuracy: 0.8628\n",
      "Epoch 89/100\n",
      "453/453 [==============================] - 35s 77ms/step - loss: 0.2549 - accuracy: 0.8835 - val_loss: 0.3142 - val_accuracy: 0.8628\n",
      "Epoch 90/100\n",
      "453/453 [==============================] - 36s 80ms/step - loss: 0.3260 - accuracy: 0.8713 - val_loss: 0.3364 - val_accuracy: 0.8274\n",
      "Epoch 91/100\n",
      "453/453 [==============================] - 39s 86ms/step - loss: 0.2719 - accuracy: 0.8697 - val_loss: 0.3201 - val_accuracy: 0.8451\n",
      "Epoch 92/100\n",
      "453/453 [==============================] - 43s 94ms/step - loss: 0.2552 - accuracy: 0.8791 - val_loss: 0.3185 - val_accuracy: 0.8584\n",
      "Epoch 93/100\n",
      "453/453 [==============================] - 36s 80ms/step - loss: 0.2408 - accuracy: 0.8896 - val_loss: 0.3111 - val_accuracy: 0.8628\n",
      "Epoch 94/100\n",
      "453/453 [==============================] - 35s 78ms/step - loss: 0.2581 - accuracy: 0.8791 - val_loss: 0.3154 - val_accuracy: 0.8628\n",
      "Epoch 95/100\n",
      "453/453 [==============================] - 38s 84ms/step - loss: 0.2532 - accuracy: 0.8857 - val_loss: 0.3548 - val_accuracy: 0.8540\n",
      "Epoch 96/100\n",
      "453/453 [==============================] - 52s 115ms/step - loss: 0.2358 - accuracy: 0.8945 - val_loss: 0.3141 - val_accuracy: 0.8673\n",
      "Epoch 97/100\n",
      "453/453 [==============================] - 46s 101ms/step - loss: 0.2419 - accuracy: 0.8890 - val_loss: 0.3139 - val_accuracy: 0.8982\n",
      "Epoch 98/100\n",
      "453/453 [==============================] - 44s 97ms/step - loss: 0.2468 - accuracy: 0.8890 - val_loss: 0.2896 - val_accuracy: 0.8982\n",
      "Epoch 99/100\n",
      "453/453 [==============================] - 39s 85ms/step - loss: 0.2291 - accuracy: 0.8995 - val_loss: 0.2975 - val_accuracy: 0.9115\n",
      "Epoch 100/100\n",
      "453/453 [==============================] - 39s 86ms/step - loss: 0.2251 - accuracy: 0.8984 - val_loss: 0.2982 - val_accuracy: 0.9071\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train_1, \n",
    "                    y_train_1, \n",
    "                    validation_data=(X_val_1, y_val_1),\n",
    "                    batch_size=4, \n",
    "                    epochs=100, \n",
    "                    verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "08d5a53d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8/8 [==============================] - 0s 20ms/step - loss: 0.5791 - accuracy: 0.8194\n",
      "Accuracy of LSTM model - 81.938 %\n"
     ]
    }
   ],
   "source": [
    "test_accuracy = model.evaluate(X_test_1, y_test_1)\n",
    "print(\"Accuracy of LSTM model - %.3f\"%(test_accuracy[1]*100), \"%\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "6f1ac0d2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8/8 [==============================] - 0s 17ms/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7feea44a6a20>"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_pred_1 = model.predict(X_test_1)\n",
    "y_pred_1 = np.apply_along_axis(lambda x: np.argmax(x), axis=1, arr=y_pred_1)\n",
    "\n",
    "y_test_class = np.apply_along_axis(lambda x: np.argmax(x), axis=1, arr=y_test_1)\n",
    "cm = confusion_matrix(y_test_class, y_pred_1)\n",
    "disp = ConfusionMatrixDisplay(cm, display_labels=['Negative', 'Neutral', 'Positive'])\n",
    "disp.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "103147b9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.39      0.57      0.46        21\n",
      "           1       0.95      0.95      0.95       139\n",
      "           2       0.74      0.63      0.68        67\n",
      "\n",
      "    accuracy                           0.82       227\n",
      "   macro avg       0.69      0.72      0.70       227\n",
      "weighted avg       0.83      0.82      0.82       227\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_pred_1, y_test_class))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "86d8d6bd",
   "metadata": {},
   "outputs": [],
   "source": [
    "# model.save('model.keras')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "16288f21",
   "metadata": {},
   "outputs": [],
   "source": [
    "# m1 = tf.keras.models.load_model('model.keras')\n",
    "# ta = m1.evaluate(X_test_1[9:], y_test_1[9:])\n",
    "# res = m1.predict(X_test_1[9:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "68abd8fe",
   "metadata": {},
   "outputs": [],
   "source": [
    "# with open('tokenizer.pkl', 'wb') as f:\n",
    "#     pickle.dump(tokenizer, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "3cf236d3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3\n"
     ]
    }
   ],
   "source": [
    "print('3')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "dbaba3f7",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.7.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}