Intent_Detection_using_Sequence_Models.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.preprocessing.text import Tokenizer\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Embedding, LSTM, Dense\n",
    "from tensorflow.keras.metrics import Precision, Recall\n",
    "from tensorflow.keras import utils\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn import metrics\n",
    "\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Read data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>intent</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Add another song to the Cita Romántica playlist.</td>\n",
       "      <td>add_to_playlist</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>add clem burke in my playlist Pre-Party R&amp;B Jams</td>\n",
       "      <td>add_to_playlist</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Add Live from Aragon Ballroom to Trapeo</td>\n",
       "      <td>add_to_playlist</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>add Unite and Win to my night out</td>\n",
       "      <td>add_to_playlist</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Add track to my Digster Future Hits</td>\n",
       "      <td>add_to_playlist</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                text           intent\n",
       "0  Add another song to the Cita Romántica playlist.   add_to_playlist\n",
       "1   add clem burke in my playlist Pre-Party R&B Jams  add_to_playlist\n",
       "2            Add Live from Aragon Ballroom to Trapeo  add_to_playlist\n",
       "3                  add Unite and Win to my night out  add_to_playlist\n",
       "4                Add track to my Digster Future Hits  add_to_playlist"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "snips_df = pd.read_json(\"data/snips.json\")\n",
    "snips_df.columns = [\"text\", \"intent\"]\n",
    "snips_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "get_weather               2100\n",
       "play_music                2100\n",
       "book_restaurant           2073\n",
       "search_screening_event    2059\n",
       "rate_book                 2056\n",
       "search_creative_work      2054\n",
       "add_to_playlist           2042\n",
       "Name: intent, dtype: int64"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "snips_df['intent'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 1080x1080 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3EAAAEGCAYAAADLzBwcAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/d3fzzAAAACXBIWXMAAAsTAAALEwEAmpwYAABt8UlEQVR4nO3dd5wV1f3/8dd7+y5VEBABXRQUUBQFCyoW1JioiZoiMRpRkxgNxpio32x+RoOmYaLRxKhoYsESY2xRIXYFV5DelqqUFRZQeodt9/P7Y2bxsmzn7s6Wz/PxuA9mZ86c87l3713mc8+Zc2RmOOecc84555xrGpKiDsA555xzzjnnXM15Euecc84555xzTYgncc4555xzzjnXhHgS55xzzjnnnHNNiCdxzjnnnHPOOdeEpEQdgGu8DjzwQMvOzo46DOecS4gZM2asN7NOUcfhnHPO7S9P4lylsrOzmT59etRhOOdcQkj6LOoYnHPOuUTw4ZTOOeecc84514R4Euecc84555xzTYgncc4555xzzjnXhHgS55xzzjnnnHNNiCdxzjnnnHPOOdeEeBLnnHPOuUZB0lWS/l7Jse1VnJct6Xv1EE+lbVZzXrakeeH2IEl/q6ZswmN3jZOkMyWNjTqO2pB0naQro46jNsLX+ZSo46hPnsQ555xzrqnLBhplImRm083sxiqKZNNIY3d1JymyZbwUSNg1vpmNNrOnElVfAzkT8CTOOeecc25/SfqvpBmS5ku6Ntx3taRPJE0ATo0r21PSx5KmSfptNVWPAoZImi3p55IyJD0hKU/SLElnVRHTVZJelfSmpMWSflNBmdaS3pM0M6zzonD/byX9LK7c7yXdWO7cPT0vks4IY5wdxtWmfOzVvoiuXkhqJWmcpDmS5kkaJmmgpAnhe/YtSV3Dsj8K35dzJL0kKSvc/6Skv0j6ALhbUi9J74blZko6PGyutaQXJS2S9KwkVRHXKEkLJM2VdE+4r4ukV8J650g6JezRXSjpIWAm0EPSrWGccyXdGVfnFZKmhu+5RyQlh/u3h+/hOZImS+oS7h8p6ZZwe7yku8PzP5E0JNyfJek/YVvPS5oiaVAVz+sr4ed7pqQXws/Y1yT9J67MmZJer6x8uD9f0p1xn80+krKB64Cfh89xSK3eDE2EL/btKpW3agvZOeOiDsM51wLlj7og6hBc/bjGzDZKygSmSRoH3AkMBLYAHwCzwrJ/BR42s6ckjaim3hzgFjO7EEDSzQBm1l9SH+BtSUeY2e5Kzj8ROBrYWRaXmU2PO74buMTMtko6EJgs6TXgMeBl4K8Kej6+G9bVppJ2bgFGmNnE8CJ0d/nYG5PsnHGtgB7ho3vcvwcQXEOmho+UO8hc/hVSewLFcY8SYBOwMnwUlG13HzVkV8M+m2p9FVhtZhcASGoHvAFcZGbrJA0Dfg9cA7xsZv8Iy/0O+AHwQFjPEcA5ZlYqaQowysxekZRB0HnSAzgOOApYDUwk+PLio/IBSeoAXAL0MTOT1D489DdggpldEiZgrQl+J0cCV5vZTyR9BehN8H4U8Jqk04F1wDDgVDMrDpO+y4GngFbAZDO7TdKfgB8Bv6vgtUoxsxMlnQ/8BjgH+AmwycyOkXQ0MLuyFzr8DP06fJ12SPol8AvgD8AjklqZ2Y4wzuerKH9XWOV6Mzte0k8IPks/lDQa2G5m91QWR1PnSZxzzjnnGsqNki4Jt3sA3wfGm9k6AEnPE1wEQ3Bh+61w+2ng7lq0cxrhRbWZLZL0WVjv3ErKv2NmG8IYXg7Pj0/iBPwhvAiOAd2ALmaWL2mDpOOALsAsM9ugoIetIhOBv0h6liARKKiiE6ZBZOeM6wQMBo5h34StfU3ryYIJwBk1LV+Qk7uBuKQufMwDJnUfNWRDTetJoDzgHkl3A2MJks+jgXfC31EysCYse3SYvLUnSKDeiqvnhTCBawN0M7NXAMq+QAjrmmpmBeHPswmG1O6TxAFbCRL9f4ZfeJTdSzcUuDKstxTYIukA4DMzmxyW+Ur4KPtSpDVBUncMwZcm08JYMoG1YZmiuDZmAOdW8lq9HFcmO9w+jeCLF8xsnqTKPmsAJwP9gIlhDGnAx2ZWIulN4OuSXgQuAP6P4H21T/lK4vlmFe02K57EOeecc67eSTqT4Bv7wWa2U9J4YBHQt4rTrK7N1bJ8+XbK/3w50AkYGPZe5AMZ4bF/AlcBBwGPV9mI2ajwYvx8gt68c2oZ537JzhmXRJCYnEKQuJ0C9GrIGOJ0DB/Hlj9QkJO7GJgU91jYfdSQur4XasTMPpE0kOB380fgHWC+mQ2uoPiTwMVmNkfSVQT3X5XZEf5b1XuwMG67lEqux8Ok5kTgbIJe3hsIErjK7IjbFvBHM3skvoCknwJjzOxXFZxfbGZlr3OlccXFH1+mNp85EXxxclkFx54HRgAbgWlmtk1B5lZZ+criafZazBN1zjnnXKTaEQy32hkOcTyZoBfgTEkdCXodvgPMCctPJLhwfYYgiarKNvYewvhheM77ko4ADgEWV3H+ueHQtV3AxQRD5srHvjZM4M4CDo079grBsK5UqpmgRNLhZpYH5EkaDPQh6IGqrOduv2TnjGvLl8naYOAkoG19tJVgR4aPq8OfNxXk5E4mSOgmAlO7jxqyo7KT60LSwcBGM3tGwayk1wKdJA02s48lpQJHmNl8gt/XmnDf5cCq8vWFQ28LJF1sZv+VlE7Qm1ebmFoDWWb2P0mTgSXhofeA64H7w+GUrSo4/S3gt5KeNbPtkroRDHF9D3hV0n1mtjZ837cxs89qE1sFPgIuBT6Q1A/oX0XZycCDknqZ2RIF9xR2N7NPgPEEw5R/RJDQVVe+MttoGu/1OvMkzjnnnHMN4U3gunCY1WKCC7M1wEiCoVFrCCZkKLvQ/RnwLwUTh7xUTd1zgRJJcwh6SR4CRkvKI7gv6yozK6zi/I8Ihmz2Av5V7n44gGeB1yVNJ7jXZ1HZATMrUjCRxeZwaFtVbgqTwFJgAcE9V7H42M3svmrqqFJ2zrg2BPdRfZeg5zN1f+prJA4AvhY+AIoLcnLfBp4D/pughK4/8GdJMYJk53qC987fwvvjUoD7gfnA7cAU4DOCYZiVJeHfJ7jH666wzu/UMqY2BAlXBkHvVdnENz8DHpX0A4L30vV8OdQTADN7W1Jf4ONwCOJ24AozWyDp1wT3iSaFcY0In8v+eAgYE36+ZxF8JrdUVDC8x/Aq4LkwuYXgnrdPwqGoYwl6t4dXV76KeF4HXlQwCdFPzSx3f55cY6Qve02d21t6197Wdfj9UYfhnGuB6mNiE0kzzKzS2dJcyxReHA4ysxvqeH4SQfL5HTP7NJGx1VR2zrhM4EKCxO18vhzq2WBGkTnhNFJrfE9cAu0kuGB/Dnij+6ghRRHE0OKFPYKpZrZbwSyc7xH0XPrvo554T5xzzjnnXB2Ew8bGAq80dAKXnTMuFTiPIHG7iGDiipYoi2AWw2HA5oKc3JcJEroPuo8aUl3PqEucLIKhlKkEvYbXewJXvzyJc84551yTIKk/wbDHeIVmdlINzj2PfWe4XG5mlxAMwaw1M1sAHFaXc+sqO2fccQTTuX+LYJih+1J7gvsZrwG+KMjJ/Q/wz+6jhlQ1U2KjIOkVoGe53b80s7cqKt/YmNk2YJ+RDuEyC+nldn8/vDfU7QdP4pxzzjnXJIQXfgPqeO5b7D0VfJORnTNOBPeD3QJUunC520sX4KfATwtyct8B/tx91JB3Io6pUuGXCc1OTb5gcXXjSZxzzjnnXCOUnTMuDbgCuJlgnSxXN+cC5xbk5M4G7gX+3X3UkJJoQ3Ju/yRFHUAZSWeGs9E0GZKuk3Rl1HHURvg6nxJ1HM4555yrWHbOuLTsnHHXA58STLfuCVxiDCAYjru4ICf3moKcXO/McE1Wg795JaWYWSTffoSLBcrMYomoz8xGJ6KeBnYmwTSzkyKOwznnnHNxsnPGpQM/AHKAHhGH05wdRpAc/7+CnNzfA097z5xraqrtiZPUStI4SXMkzZM0TNJASRMkzZD0lqSuYdkfSZoWln0pXIwPSU9K+ku4jsrdknpJejcsNzOcihSgtaQXJS2S9GyYdFUW1yhJCyTNlXRPuK+LpFfCeudIOkVStqSFkh4imAK4h6RbwzjnSrozrs4rJE2VNFvSI+F0qUjaLun3YZ2TJXUJ94+UdEu4PV7S3eH5n0gaEu7PkvSfsK3nJU2RVOkU15K+Iunj8HV5QVJrSV+T9J+4MmdKer2y8uH+fEl3hvvzJPWRlA1cB/w8fI5DKmj/WknTJU0v3Vnh8h7OOeecS7DsnHHfJuh5exBP4BrK4cDjwKKCnNxmeU+aa75qMpzyq8BqMzvWzI4mWKzzAeDbZjaQ4M3/+7Dsy2Z2gpkdCywk+DapzBHAOWZ2M8GimQ+G5U7hywUKjwNuIhg2cBhwakUBKVhd/hLgKDM7BvhdeOhvwISw3uMJFmQEOBJ4ysyOC7d7AycSdKsPlHS6ggURhwGnmtkAgsUTLw/PbwVMDuv9kGAV+YqkmNmJ4XP4TbjvJ8CmMM7fAgMrORdJBxIsXniOmR0PTAd+AbwDnCypVVh0GPB8FeXLrA/3PwzcYmb5wGjgPjMbUNHCh2b2qJkNMrNByVntKgvVOeeccwmQnTOuZ3bOuHHAC3jyFpXDgZcLcnJfK8jJPSTqYJyriZoMp8wD7pF0N8FaKJuAo4F3wo6yZL5Mwo6W9DuCKV5bs/csUC+Eq7C3AbqZ2SsAZrYbIKxrqpkVhD/PBrKBjyqIaSuwG/inpHFhXABDgSvDekuBLZIOAD4zs8lhma+Ej1nhz60JkrpjCBKsaWEsmcDasExRXBszCG6QrcjLcWWyw+3TgL+GMc1TsJJ9ZU4mSGAnhjGkAR+bWYmkN4GvS3oRuAD4P+CMispXEs83q2jXOeeccw0oXOftVoIvYzMjDscFvg4MLcjJ/Q1wv68z5xqzapM4M/tE0kDgfOCPBL1C881scAXFnwQuNrM5kq4iuP+qzI7w30qHSAKFcdullcUXJjUnAmcTLHJ5A0ECV5kdcdsC/mhmj8QXkPRTYIyZ/aqC84vNzKqLKy7++DJVPd/yBLxjZpdVcOx5YASwEZhmZtvC4aaVla8sHuecc85FKDtn3BCCkTE+YUnj0wq4B7i8ICf3x91HDZkWdUDOVaQm98QdDOw0s2cI3tQnAZ0kDQ6Pp0o6KizeBlijYLX2yyuqz8y2AgWSLg7PTy+7d66mwvu+2pnZ/wiGLg4ID70HXB+WSZbUtoLT3wKuibt3rJukzuG53w63kdRB0qG1iasSHwGXhnX2A/pXUXYycKqkXmH5LElHhMfGEwwR/RFBQldd+cpsI/g9Oeecc64BZeeMOzA7Z9wTwAQ8gWvsjgMmF+TkPlCQk1vR9aRzkarJPXH9ganh8MbbgDuAbxNMUDIHmE1wXxvA7cAUgt66RVXU+X3gxnBo4STgoFrG3QYYG54/Afh5uP9nwFmS8giGEB5V/kQzexv4F/BxWO5FoI2ZLSAY0vB2WO87QNdaxlWRhwiS3rnAL4G5QIUzhpjZOuAq4Lmw/GSgT3islGBI59fCf6ssX4XXgUsqm9jEOeecc4mVnTNO2TnjriG4NrqK2o3ScdFJIhjttaAgJ/dbUQfjXDx9OUrQ1YdwhstUM9utYBbO94AjzKwo4tCqld61t3Udfn/UYTjnWqD8URckvE5JM8ys0tmBnasP2Tnj2gJjgIsjDqVejSJzwmmknhF1HPXsJeDq7qOGbIs6EOcazWLfzVgW8FHYa/kKcH1TSOCcc845t3+yc8YdRTBz9MURh+IS41vA1IKc3OpGPTlX7xr9ZBeSXgF6ltv9SzN7q6LyjY2ZbQP2+eZX0hQgvdzu75tZXoME5pxzzrl6k50z7rvAPwkmynDNRx+CRO7K7qOG/DfqYFzL1eiTODNrlosvmtlJUcfgnHPOucTKzhmXAvyZYOI11zy1IVhX7g/AHd1HDYlFHZBreXw4pXPOOedcAmTnjOtCcO/7TRGH4uqfCCb8G1eQk3tA1MG4lqfR98S56PTv1o7p9TC5gHPOOdfcZOeMOwV4ATg46lhcg/oqML0gJ/eS7qOGzI06GNdyeE+cc84559x+yM4Zdz3Beq6ewLVMhwEfF+TkXhZ1IK7l8CTOOeecc66OsnPG/T+CNWFTo47FRSoL+FdBTu7/RR2Iaxk8iXPOOeecq4PsnHEjgd9HHYdrVO4uyMnNiToI1/x5Euecc845V0vZOeN+B/wm6jhco/THgpzc26IOwjVvnsQ555xzztVCds64PxHMTOhcZX5XkJN7e9RBuObLZ6d0lcpbtYXsnHFRh+Gca4HyfWZc10hl54y7D19CwNXMXQU5uUndRw25M+pAXPPjSZxzzjnnXDWyc8YJeAAYEXUsrkkZGSZyPvTWJZQPp3TOOeecq0KYwI3GEzhXN3cU5OTeFXUQrnnxJM4555xzrmoPAtdGHYRr0m4vyMn9XdRBuObDkzjnnHPOuUpk54z7BXB91HG4ZuG2gpzcn0QdhGsePIlzzjnnnKtAds64bwB/jjoO16z8tSAn95yog3BNnydxzjnnnHPlZOeMOw74F36t5BIrBXihICf3yKgDcU2b/2FyzjnnnIs3st1BvbVyNNAq6lBcs9QeGFuQk9sh6kBc0+VJnHPOOedcmZHtUoEX3krL6XVe0tRZUYfjmqfiWFHR/wr+8dC9wy70a3FXJ/7Gcc4555z70j3AaUmyDqNT7+9/R8pTH0YdkGteNhetm/jfFX87bFvxxmGALwTu6qRJJ3GSzpQ0NuIYsiV9L+7nQZL+FmVMFWkMr5VzzjnXqI1sdzlwY9mPEinXpLx5+mtpt+WmUlIUYWSuGTCzkk+3zvzwrVWPnxqz0oxw9233Drvw/EgDc01Sk0jiJKU04naygT1JnJlNN7MbKy/e8Brq9XPOOeearJHt+gOPVnTomKTlQ6anX7foQDava+CoXDNhZhsnrv1v3swN75xe7pCAZ+4ddmF2BGG5JqxekjhJrSSNkzRH0jxJwyQNlDRB0gxJb0nqGpb9kaRpYdmXJGWF+5+U9BdJHwB3S+ol6d2w3ExJh4fNtZb0oqRFkp6VpCriOkHSpLCOqZLaSLpK0guSXgfeDmN/PIxplqSLwnOzJeWGbc+UdEpY7ShgiKTZkn5e1uMlKUlSvqT2ce0vkdRFUqfwuU4LH6dWEXOepPYKbJB0Zbj/aUnnSMqQ9ERYbpaks8Ljez2vCl6HWZIOq6C9ayVNlzS9dOeWan7TzjnnXDMQ3Af3DJBVWZF22nnM5PQbigdp8cKGC+xL6/93PysfuJzVj+27zNiWKS/z2d0XUtn/25Wdu2n8E6x+/AbWj713z77t895n6/RXExt8OTf/bxQDHvgGZz82fJ9jo6c8R4+7T2fjzs37HFu99Qsufe5nnPWPKzj7n1fy2PQX9hz7w/iHOffxq7hp7O/37Htp3lt7lYlKSaxo8f8KHt25aucnx1VS5ADghXuHXehfursaq6+euK8Cq83sWDM7GngTeAD4tpkNBB4Hyj5lL5vZCWZ2LLAQ+EFcPUcA55jZzcCzwINhuVOANWGZ44CbgH7AYUCFCZGkNOB54GdhHecAu8LDg4HhZjYUuA1438xOAM4C/iypFbAWONfMjgeGAWVDJnOAXDMbYGb3lbVnZjHgVeCSsP2TgHwz+wL4K3Bf2Ma3gH9W8VpODJ/TUcAyYEi4/2RgMjAibK8/cBkwRlJZF3388yp7HU4BRgMXmdmy8o2Z2aNmNsjMBiVntasiLOecc67Z+D/gmOoKpSh28Atpd2Zfk/zGpAaIaS+t+59D5+/se/tUydZ17M6fRXLbTrU6N1a4g8JVCzn4mr9jFqNoXT6x4kJ2zHuXNsddkPD4432n/1d5+jv7Lr+3eusX5OZPp1vbLhWel5yUzO1n/YQPfvQMr35/NGNmvsIn6/PZWrid6avm8841T1JqpSxct5RdxYW8MO8Nrjzuknp9LtXZWrRh0n9XPNBje8nm7tUUHQT8oiFics1DfSVxecA5ku6WNAToARwNvCNpNvBroOzNfHTYw5UHXE6QrJR5wcxKJbUBupnZKwBmttvMdoZlpppZQZg0zSYY3liRI4E1ZjYtrGOrmZWEx94xs43h9leAnDDO8UAGcAiQCvwjjPMFgqSxOs8TJHwA3w1/hiCB/HvYxmtA2/A5ViQXOD18PAz0l9QN2Ghm24HTgKfD57QI+Iwg+S3/vAD6EgwV+bqZrahB/M4551zzNrJdH+D2mhaXyLwj9elT/pl6zwQwq8fI9pLR42iSM/e9VNj03j844KyrCUbl1eZcYaUlmBlWUoSSktk69WXaDPwGSq7fDqGTewygfWbbffbf+d7fue2s61Elz6VL6wPpf1CwvFrr9Cx6dTyUz7etI4kkikuLMTN2lxSSmpTCI1Of4+qB3yK1np9LZcwstmzbnAlvrPrnKaVWUmkPbzkj7x124T6jpJyrSL0kcWb2CTCQIJn7I0Fv0/ywt2qAmfU3s6+ExZ8Ebgh7ku4kSJrK7Aj/rfwvExTGbZcSLKJYEQGV/bHdUa7ct+JiPcTMFgI/B74AjiX4tiStipjKfAz0ktQJuBh4OdyfBAyOa6ObmW2rpI4PCXrfhhAkleuAbxMkd2XxVmZHuZ/XALsJei+dc865lm1kOwH/ANJre+o5yTPPmJh+47RW7Krs/+96t/PTKSS36Uha59pf9yelZ5F15CmsefJGUtp1QemtKFrzCVm9T66HSKv39qcfcVCbA+nXuVeNyq/csob5X3zKcQf3o3V6FucfeQZfffIH9GjXlTbprZizZhHn9R5SfUX1wMw2T173+sxp6988o5anZgKP1EdMrvmpr3viDgZ2mtkzBFP1ngR0kjQ4PJ4qqazHrQ2wRlIqQU/cPsxsK1Ag6eLw/PSye+dqYRFwsKQTwjraVDLhx1vAT8vurZNUlvC0I+jJiwHfB5LD/dvC51BR3Aa8AvwFWGhmG8JDbwM3lJWTNKCyoM1sJXAg0Dsc/vgRcAtfJnEfEr5uko4g6DVcXEl1m4ELgD9IOrOyNp1zzrkW4nqCES110k0bTpyRft3aw7XqswTGVCOx4t1s+fh52g+5os51tDvp2xx89QN0GPpDtuQ+Q/shV7Btzlus++8oNk/6dwKjrdqu4t088PHT3DzkB9UXBnYU7eTHr9zOyLN/Spv0YD3260/6Hm9d/Th3DL2Be3If4+Yh1/DcnLFc/9/f8NdJY+oz/L2UxIo/fWPVP7es2LFwUB2rOOfeYRdemdCgXLNUX8Mp+wNTw+GCtwF3EPQe3S1pDsGwx7KJQW4HpgDvECRalfk+cKOkucAk4KDaBGRmRQRDGx8IY3iHvXv9yvyWYOjkXEnzwp8BHgKGS5pMMFyxrJdrLlASTpby8wrqex64gi+HUkIwffEgSXMlLQCuqyb8KcAn4XYu0I0gmSuLKzkc5vk8cJWZFe5bRSC8J+/rwIPhfXrOOedcyzOyXQ+Cycn2S4aKD38n7dZ2FyRNnpGAqGqsZPPnlGz5gtWP/5SCh6+hdNt61jx5E6XbN9W6rqIvlgKQckA3dsx7n04X51C87jOKN65KdNgVyt+8ipVb1nDe49cw+OFLWbNtHV978oes3b5hn7LFpSVc+8rtXNzvXL525L4dXfO+CC6XDjugBy/Oe5OHL76TxeuWs3zjynp/HtuLN09+dcUDXbcVbzx0P6v6y73DLjwwIUG5ZqteBgqb2VsEPVrllZ9WFTN7mOBer/L7ryr386fA0HLFlhEMMSwrcwNVCO+HKz9O4MnwUVZmF/DjCs79lL1vev5VuL8YOLtc8fiYplNuyKOZrefLe+WqZWbfj9ueRFzybWa7gasqOOdJ9n5e48viCu+HO6r8Oc4551wL8jCVjKSprSTR/u+pfxtwUunCCXeUXF3bIXR1ktYpmx4/fXbPzwUPX0PX4fdRl0nJNuc+Q4fzboBYCVgs2KkkrKTS74QTqm+nw5n909f2/Dz44UsZN/xROmS136ucmXHrG3fTu+OhXHtixZdR9+Q+xqjzbqU4VkIsfC5JSmJXPT4XM7MVOxZ8OHnd2NOp+jaXmuoI3EfQgeFchZrEOnHOOeeccwkzst1lBLcXJIxE8pUp75wxLu1XH9XHwuDrXvsTnz99C8UbV1Hw4HC2zXm70rIl2zbwxQu/qdG5Oz/5mLSDepPSpiNJGa1JP7gPqx8bAaJO99rVxIjX7uTip69n2cYVnPDgt/j3nLGVlv1823qufOFWAKatyuOl+W8xccVMznviGs574hreX/rxnrJvfpLLsQf14aA2B9Iuow3HH3wU5zw2HIka32tXW2a2der6/02bvG7sGSQmgStzxb3DLvxK9cVcSyVruImVGoykV4Ce5Xb/MuwhbJQkXQ38rNzuiWY2Iop4ANK79rauw++PqnnnXAuWPyrxU5xLmmFmdb1PxTUXI9t1JFjSqPI5+ffTVsvKO7fwT12+oEPn+mqjsRlF5oTTSG2QXsjGotRKlr2zaoy2FK8vf82ZKMuBo29+fuzOaku6FqdZLipoZtEuClIHZvYE8ETUcTjnnHPN3O+oxwQOoK129p+UfuOay4puWzDV+tZkSSLXxOwo2Tr1rYLH+hRb0b5rJSROT2AkwTqGzu3Fh1M655xzrmUIJjO5piGaSlas6/Npvz3s2uSxExuiPdcwzMxW7lg8YezKh0+o5wSuzE/vHXZh1wZoxzUxnsQ555xzrqX4FTVb5zUhJDL+X+q/Tn0i9e7xIhZrqHZd/TCz7TM2vD1l0tr/Jvr+t6pk4D1xrgLN8p44lxiDBg2y6dOnRx2Gc84lhN8T18KNbNcdWEoDJnHxPrcDpp1b+KcjttGq9tNHNgHN/Z64Uiv97L3VTxdvKvqifmZIqdouIPvm58eujaBt10h5T5xzzjnnWoIcIkrgAA7SphOmp1+/4QitXB5VDK5udpVsm/7qigfaR5TAAWQCN0fUtmukPIlzzjnnXPM2sl034IdRh5GuksPeSvtlh4uSJvowlyZi9c6lE15b+dDxxbHCqHtQf3LvsAs7RhyDa0Q8iXPOOedcc/dLID3qIAAk2t2f+uBxf0j554SoY3GVM7Odsza8Nyn3ixfPoHFcL7cGfh51EK7xaAxvSuecc865+jGyXVfgR1GHEU8i+Xsp75/xVtr/TUynaHfU8bi9xax05burny74ZOv0U6KOpZyf3jvswvZRB+Eah2a5TpxLjLxVW8jOGRd1GM65Fqg+Fvt2LdYvCWb4a3SOTCo4dXr69fO/UvinjmvoeFDU8TjYXbpj5hsFj2UXxXb1iDqWCrQFbgTuijoQFz3viXPOOedc8zSy3UHAj6MOoypttOuoj9Jv1MlJ8+dHHUtL98Wu/AmvrXjw2KLYrg5Rx1KFm+4ddmGbqINw0fMkzjnnnHPN1fU00l64eMmyLs+l/r7XiOT/fhR1LC2Rme2es3H8xPGfP3+GYclRx1ONAwje166F8yTOOeecc83PyHYCvh91GDUlkX5r6n9Oezr1DxOSiJVGHU9LEbPY6vfX/Gv5oi1TTo06llq4OuoAXPQ8iXPOOedcc3Qq0DPqIGprSPK8Myan3zCrLdu3RB1Lc1dYumvO6ysfSl1fWNA36lhqqc+9wy4cGHUQLlqexDnnnHOuOboi6gDqqrM2D5qe/pNNfbRiWdSxNFfrdq/88NUVf++3u3RHp6hjqaPLow7ARcuTOOecc841LyPbpQOXRh3G/khTSfYbaTkHXpKUOy3qWJoTMyuct+mj3PfX/Ot0I5YadTz74bv3Druwsd+/5+qRJ3HOOeeca24uIJgAokmTaPuX1IcH/ill9PioY2kOYhb7fPzn/14yf/PEIVHHkgBdgbOjDsJFx5M455xzzjU3TWZCk+pIJF2a8uGZ76TdMjGDwl1Rx9NUFZXuznt95cNJa3evOCrqWBLIh1S2YJ7EOeecc675GNmuA3B+1GEkWu+k1adOT79++cGsXxN1LE3NhsI1ua+ueODI3aXbO0cdS4Jdcu+wC7OiDsJFw5O4BiDpJkl1+pBJypd0YAJieFLSt/e3Huecc66RuxRIizqI+tBau/vlpv8s+bSkvLyoY2kKzKx44ebJue+ufmpIjFhzfE+0Ab4RdRAuGp7EJYgClb2eNwH+TYlzzjlX/5rNUMqKJMs6P536xyN/lvySLwxeBbPYug+/eHHh3E0TmsP9b1VpsrOwuv3jSdx+kJQtaaGkh4CZwGOSpkuaL+nOsMyNwMHAB5I+CPd9RdLHkmZKekFS62qaulXS1PDRK6zjUEnvSZob/ntIVfvLxf3bsGdun9+/pGvD5zC9dKcvUeOcc4kQ/n8xLwH1JGR0Rg3aqfMIkkhjGNkuGzilXgJqRCTSfp760mnPpf7WFwavQHGscMHYgkdKPt+17JioY2kA59077MJ6/5tQF5LOlDQ24hiyJX0v7udBkv4WZUwVqctr5Unc/jsSeMrMjgNuNrNBwDHAGZKOMbO/AauBs8zsrPA/318D55jZ8cB04BfVtLHVzE4E/g7cH+77e9juMcCzwN+q2Q+ApD8BnYGrzSxWviEze9TMBpnZoOSsdrV7JZxzzjU61YwUqcxNJGgEiaSUOp5alxjOq2NbTdLg5IVnTE3/yZz2bNsUdSyNxebCtR/9d8UDh+0s2do16lgaSArwtSgD2I/PeEO0kw3sSeLMbLqZ3ZiwoBKgrq+fJ3H77zMzmxxuXyppJjALOAroV0H5k8P9EyXNBoYDh1bTxnNx/w4OtwcD/wq3nwZOq2Y/wO1AezP7sZlZNW0655xLrBRJY8KREi9KypJ0tqRZkvIkPS4pHaCy/WUkZUp6U9KPKmqogpEiPSTdKmla2H7ZaJFWksZJmiNpnqRhlYwgebj8SJNw/56ewfAb7vHh9khJj0p6G3gqjCc3HIEyU9IpYbkzJY0PX49Fkp4Nk859YqihobUo2ywcqK3HT0v/ydZ+yl8adSxRMrOST7ZMn/DW6idOi1lpRtTxNLCzalO4ks/9QEkTJM2Q9JakrmHZH4V/N+ZIeqmsdzwc0fWX8PN5t6Rekt4Ny82UdHjYXOvyn+8q4jpB0qSwjqmS2ki6SsGotdeBt8PYHw9jmiXpovDcCv/GAKOAIZJmS/p5+DdnrKSk8O9X+7j2l0jqIqlT+FynhY9Tq4g5T1L78O/WBklXhvuflnSOpAxJT4TlZkk6Kzy+1/Oq4HWYJemwqn6PnsTtvx0AknoCtwBnh71g44CK/ogIeMfMBoSPfmb2g2rasEq2KytT2f5pwEBJHappzznnXOIdCTwa/h+xlWAUxpPAMDPrT/CN+vWSMiraH1dPa+B14F9m9o9q2isbKXIk0Bs4ERhA8H/B6cBXgdVmdqyZHQ28WX4ESVjXbeVHmtTg+Q4ELjKz7wFrgXPDESjD2HuUyHEEvW79gMOAUyuJoWoj24kWmMQBpKr00HFp/6/zpckfTI06liiY2YaP1r48b9bG986IOpaI1CqJo4LPPfAA8G0zGwg8Dvw+LPuymZ1gZscCC4H4a9YjCEaW3Uww+uvBsNwpQNksqvt8visKSFIa8Dzws7COc4CyJTUGA8PNbChwG/C+mZ0QPu8/S2pF5X9jcoDc8Jr7vrL2wtForwKXhO2fBOSb2RfAX4H7wja+BfyzitdyYvicjgKWAWX3YJ4MTAZGhO31By4DxoR/48s/r7LX4RRgNMHfzmVVtOtJXAK1JUjotkjqwt5d29sIZhCC4Bd6qr68ty1L0hHV1D0s7t+Pw+1JwHfD7cuBj6rZD8GHdBQwTlIbnHPONaSVZjYx3H6GYKHe5Wb2SbhvDHA6QcJV0f4yrwJPmNlT1bQXP1LkK+FjFkHPXB+CpC4POEfS3ZKGmFllN0PXZKRJea+ZWdlFWCrwD0l5wAvlzp9qZgXhRdVsguFPdXEM0CjvDWoIEm3uTvnHCfelPjgh6lgaUkmsaNG4gkd3rd65ZEDUsUQo+95hF/asRfm9PvdAD+Bo4B0Fo8R+DXQPyx4d9nDlEVxXxq+z94KZlYbXlN3M7BUAM9ttZjvDMjX9fB8JrDGzaWEdW82sJDz2jpltDLe/AuSEcY4n6DA5hKr/xlTmeb68xv5u+DMECeTfwzZeA9pWcd2cS/D3+XTgYaC/pG7ARjPbTjAi7unwOS0CPiNIfss/L4C+wKPA181sRXXBN8gY1pbAzOZImgXMJ8jEJ8YdfhR4Q9Ka8L64q4Dn9OXwmF8Dn1C5dElTCJLuy8J9NwKPS7oVWAdcXc3+sjhfCN+Ir0k6P+4/WOecc/WrpsPYKx1uFJoIfE3Sv6oZGr+jXJ1/NLNH9mlMGkiwrtofJb1tZneVO1420uQEM9sk6Um+HGlSwpdfCJcffRLf/s+BL4Bjw/K7444Vxm2XUvdrkxbZCxdPQpckTzzjWC2bdEHRHwbsIr1Zz4y9pWj9xHdWjzm+1Eoyo46lETgLWF6Tgmb2SfznHngHmG9mgyso/iRwcXidexVwZtyxss94VX+zavr5FpX/jSz/t+xbZrZ4r5OlkVT+N6YyHwO9JHUCLgZ+F+5PAgbX8Br5Q4LetkMIegkvAb5NkNyVxVuZHeV+XkPwd/Q4gpEIVfKeuP1gZvlhN3TZz1eZWV8zu8DMvmlmT4b7HzCzPmVDQszs/bBr+pjw8VoVbWSb2Z1mdlJ4zpK4toeG559dlrFXsf8qM3sx3H7czM7yBM455xrUIZLKLpIuA94FsstGZhBMjT8BWFTJ/jJ3ABuAh2rR9lvANQpnQ5bUTVJnSQcDO83sGeAe4PiwfPwIkqpGmuQTDJuEYNhRZdoRfMseC59Pcg1ijo+hJk6vvkjLcFjSmlNmpF+3orvWVXsh2BSZWenSrbMnvLnqsVM9gdujxkNJK/jcnwR0Kvv7JClVUlmPWxtgjaRUgp64fZjZVqBA0sXh+emq/ey2i4CDJZ0Q1tFGFU/48RbwUym4t07SceH+yv7GVPp3JPwS7BXgL8BCM9sQHnobuKGsnKQBlQVtZisJRgD0Doc/fkTwpVdZEvch4esWjrw7BFhcQVUAm4ELgD9IOrOyNst4Euecc841jIXAcElzgQ7AfQSjJV4IhwDFgNFmtrui/eXqugnIUDDjcLXM7G2CSa8+Dut8keDCpj8wNRw2dBtffhNdNoLkAzObQzCMcj7BvTLxI03uBP4qKZfgW/bKPBQ+98kEQ4nKfwNdkT0x1OQ5ElyIulCWCvt8mHZT6plJs+dGHUsimdmmj9e9Nnv6hrda6v1vlamoF60y5T/3dxD0Ht0taQ7BsMeyiUFuB6YQ9NYtqqLO7wM3hn/fJgEH1SZ4MysiGNr4QBjDO1Q8t8RvCYZOzlWwbMtvw/2V/Y2ZC5QomCzl5xXU9zzBWnvPx+27ERikYBKoBcB11YQ/hS9H1OUC3fjydqaHgOTw7+7zwFVmVrhvFYHwnryvAw+G9+lVSlWPxHANRdIrQPnxzL80s7eiiAcgvWtv6zr8/qiad861YPmjLkh4nZJmhJNzuOZmZLtDCXoFXTlmFD1YetGUe0qG1eui16PInHAaqfWaWJXEij99e9UT6dtKNu2zBq4DoOPNz4/dWH0x1xz4PXGNhJldEnUMzjnnXBNVm16IFkUi7YaUV4eclLTww+8W3X5KKclN8tpvW/Gmj99a9cQxpVbcKupYGrGTgDeiDsI1DB9O6ZxzzjVRkjoqWP+o/KNj1LE1sJOjDqCxOyHpk9Onpv8k7wC2NqmeGjOL5W+bN+F/BY8O9gSuWk1iSLGkVyr4m3Ve1HFVRdLVFcT8YJQxNclvY5xzzjkXrI9FsO5bSzew+iL755pXdzH2kxI6txLzftJ6r2P3TCrk1ncKWXdraw7M2vv78d0lxulP7KCwFEpi8O2+Kdx5VnCrzy/f2c0bS0oYcFAyT10SzM/x9JwiNu4yfnbyXuu7J0RHbTtuavqIld8sunNDnh3WO+ENJJiZbZmyftwnn22fn9Bhms9PncOCNWtpnZ7GrV/du+rxi5Yydu4i7rzoXFqlp+11bPPOXTw3ZTbbdhciiZMPO4QhRwR3woyds5DFn6/j4PZtueykAQDMyC9gZ1HxnjINoEkkcU1x9JmZPQE8EXUc8TyJc5Xq360d0+vhvhTnnHMuwQ6r7wauGpDKDSemceUre0/svHJLjHeWlXBIu4pnEk9PhveHt6J1miguNU57Ygdf611C3wOTmVRQytzrW3P5yzvJ+6KUXh2SeHJOMW9eXn8rA6SqtMdrab/eflvJNZP/VXpOo+3BLI2VLH179ZikrcXrT0h03YN6dufU3tk8N2X2Xvs379zFJ1+sp31WxRNeJkl8fUA/uh/Qjt3FJdz/zkf07nIg7TIz+GzDJm4+73SenTyLNZu3cmDrVkzLL+BHp5+Y6PCr0qv6Iq658OGUzjnnnGu6RrZLA7rWdzOnH5pCh8x9E7Wfv7WbP52TUeliUJJonRYcLY5BcWmwcFSSoKjUMDN2FUNqMvx5UhE3nphGanJ1SwXuH4nWv095/KQHUv82HhrfDHc7irdM+e+KBzpvLV5fL11Yh3fqSFZa6j77X529gAuP7Vvp77JtZgbdD2gHQEZqCl3atmbrrt1IoiQW/C6LS0tJTkrig8XLOK13NslJDXqp3ePeYRfW75vHNRqexDnnnHOuKTuU6hdIrxevLS6mW5skjj2o6mXvSmPGgNHb6fznbZx7WAondU+hTbr4Vt9UjntkBz3bJ9EuXUxbXcpFffZNLuqDhL6ePPnM8Wm/mJzF7pos+VDvzMxW7lg0YWzB6BNLrKg2awTut/mrvqBdZgYHt29bo/Ibd+xk1eYtHNKxPRmpKRzT/SDue+cjOrTKIiM1hZUbN3N0t1rNsp8I6dRyan/XdPlwSuecc841ZYdG0ejOYuP3uYW8fUX1c20kJ4nZ17Vm827jkud3Mm9tKUd3Tub/Tk3n/04N7n374Wu7uOvMdP45s4i3l5ZwTJdkfn164u+LKy876YvB09OvW/zVortbrbAu3eu9wUqY2bbp699csGz73AZf/62opJR3Fy7h2hoOfSwsLmHMpBlcNKAfGalB0n1Wn8M5q8/hAPxn2ly+evQRTFm2gsWfr+fg9m04p1+D3YJ4KLCmoRpz0fGeOOecc841ZdlRNLp0Y4zlm4xjR28n+/5tFGw1jn9kB59vj1V6TvsMceahKby5pGSv/bPWBOukH9ExiafmFPOf72Qxb20pn26oav30xMlS0ZHj036ReXbSjNkN0mA5pVaa//bqJ9cu2z43kok5NmzfwcYdO/nL27n8fuz7bNm1m/veyWXrrt37lC2NxRgzaQbHH9KN/t33HcW7atMWAA5s04rp+au48pTj+XzLNtZta7DOzki+1HANz3viXKXyVm0hO2dc1GE459we9bEIuGvysqNotH+XZNbe+uWIv+z7tzH92lb7zE65bkeM1GTRPkPsKjbeXV7CL0/de9bD2z8o5NGvZ1Acg9LwDrUkwc7ien8aeyTJOv4z9d62j5ZemPvHku/V68Lg8XaWbJv25qrHjiiOFbZrqDbL69q+LXdedO6en38/9n1uOve0fWanNDP+M20uXdq25owjK55L5815n/DtQf2JhffIQXBfZHFpwyTkeBLXYnhPnHPOOeeasga5aL3spZ0MfmwHizfE6P6XbTw2s6jSsqu3xTj/2Z0ArNlunDVmB8c8vJ0T/rGDcw9L4cIjvrzv7b+Lijnh4GQObpNE+wwxuHsy/R/ejkS199olmkTqj1PGDnk57Y4PUyip9xRy1c5PJ7y+8qGBDZ3APfPxLB54bxLrtu3gt6+/x5RlKyotu2XXbv754VQA8tdvYsZnq1iydgN/eTuXv7ydy8I1a/eUnbfqc3p0aEe7zAwy01I5tGN77nnrQ4Aa32uXAJ7EtRCyxjcpkWsk0rv2tq7D7486DOec22N/euIkzTCzQQkMxzUGI9vlAqdFHUZzs8lazzmn8M/dNtDuwJqUH0XmhNNIrdH9bGa2Y9bGd+d+unXm4P2L0lVg3M3Pj70w6iBc/fOeOOecc841Zd7zUA8O0PZjp6SP2D1ASxYnst6Yla54Z/VTqz2Bqzf+eWghPIlzzjnnXNM0sl0qcHDUYTRXKYp1fyXtju5XJr/1cSLq2126Y+arK/7eZlPR5w02VWML5ElcC+FJnHPOOeeaqi5Aw9441sJItLozZczJD6feN2F/Fgb/fOfyCa+tePDYotjuAxIZn9tHm3uHXRjZJDGu4XgS55xzzrmmKq36Im5/SehrydPOyE372dRW7Npem3PNbNecjeMnTvjiP2cY5gl3w6j/BQZd5DyJc84551xT5UlBA+qRtP6kGenXremp1ZVP5xgnZrFV7695Nn/Rlimn1ndsbi/+uWgBPIlzzjnnXFPl1zENLEPFvd9Lu7X1eUlTZ1VVrrB05+zXVjyYvr5wVd+Gis3t4Z+LFsB/yc4555xrqrzHIQJJsg6jU+/vf3vK0xMqOr5214oPX13x96MLYztrtDyBSzj/XLQAnsTtB0njJTXqNYck3SXpnKjjcM455+qBX6xGRCLlBylvnPFa2m25qZQUAZhZYd6m3I8++Py50w1LiTrGFsyv71sA/yU3c2Z2h5m9G3UczjnnXD3wJC5ixyQtHzI9/bpFSaXrto///N9LFmye5AuvR8+v71sA/yXXgKRsSYskjZE0V9KLkrLKlXlY0nRJ8yXdGe47W9IrcWXOlfRyFe1sl3S3pBmS3pV0Ytjbt0zSN8IyV0n6e9w5YyWdKSlZ0pOS5knKk/Tz8PiTkr4dbp8gaZKkOZKmSmpTQQzXhs9jeunOLfv70jnnnHP1ya9jIvJZSkrBmLZtJl1zUOcJX83ukGyLXrNNKYNiyWnHTFBS+49BBVHH2IL5lxstgHd119yRwA/MbKKkx4GflDt+m5ltlJQMvCfpGOB94EFJncxsHXA18EQVbbQCxpvZL8Pk73fAuUA/YAzwWhXnDgC6mdnRAJLaxx+UlAY8Dwwzs2mS2gK7yldiZo8CjwKkd+1d5/VgnHPOuQbgF6sNYIe0/ePMjCXvZ2VtmZaZnvlFcnJPk7oD3QGOWxKbe8ynMy/8sMtleSWtss8oO89i29eWFi9fHitesjtW8nlbbFdvoHVUz6MF8S83WgBP4mpupZlNDLefAW4sd/xSSdcSvKZdgX5mNlfS08AVkp4ABgNXVtFGEfBmuJ0HFJpZsaQ8ILua+JYBh0l6ABgHvF3u+JHAGjObBmBmW6upzznnnGvsPIlLsBjEFqWlLvsgK+vz3KwMW5KaelChdDjSgMrO+fEbsSKAIxc/Vzz/qB/s2a+k1p1T0vt3Jr0/AGYWs9IvPi0tXvJ5rDgfK93QFUoPw5OORPPXswXwJK7myvdK7flZUk/gFuAEM9sk6UkgIzz8BPA6sBt4wcxKqmij2MzK6o0BhRD80ZNU9rsqYe8PZ0ZYZpOkY4HzgBHApcA1ceVUwXNwzjnnmjK/WN1PG5KS1n+Ylbn8g6zMHXMy0tttTEo6HKkX0Ksm5x+50hZ22M4ggC7rZh6/uPiyvJLUrP4VlZWUpJSDeielHNSbzODWObPCrbHiFUtLi5dsiZUUZBHb1hPolKCn11L5lxstgCdxNXeIpMFm9jFwGfAR8PXwWFtgB7BFUhfga8B4ADNbLWk18GuCoZH7Kx/4iaQkoBtwIoCkA4EiM3tJ0lLgyXLnLQIOlnRCOJyyDbCrmqTSOeeca8z8y8laKIbiWRnpn76XlbV+cmZG8orUlB4l0iFAnZcCGDG2dK8b6I/8ZO/euOpI6W2T03ofl5zWe8++WOmmgljxshWx4qUlsdK1HbCi3kB6XWNsgfxz0QJ4EldzC4Hhkh4BPgUeJkzizGyOpFnAfIJhjRPLnfss0MnMFiQgjonAcoLhlvOAmeH+bsATYXIH8Kv4k8ysSNIw4AFJmQT3w50DbE9ATM4551wU1kYdQGO2MiWlYHxW5srxWZlFC9LSOmxPUm+kfomq/5C1tqzL5uDL5DJhb9zcktSsY+pab1LyAd2Tkgd2J2MgAGalRbGS1fNjxUs2xIo/S7bY5h4QO2Q/w2/O1kUdgKt/nsTVXMzMriu378yyDTO7qopzTwP+UV0DZtY6bntkRcfC4ZaXV1LF8RXUeVXc9jTg5OricM4555qI1VEH0FjskLZPzsxY8n5W5ubpGRlZn6ckZ8fiJh+pDze8XrpGcFj5/X0WP1sy7+gfJawdKTktObXHUcmpPfbss9iODaXF+ctixUt3xEpWt8V2Hg60S1ijTVcxnsS1CJ7E1TNJMwiGWt4cdSzOOedcszJyyy5GttsMtI84kgZlYOHkI2vCyUe67K5m8pFE67LJCg5dy0kVHeu8fvbxKcU75pSktjq2vtpXUquOKelHdST9KCD4kttK1y2NFS9dU1q8PGal6w+CksNpefeHfX7z82N9OGUL4ElcDZhZPnB0Hc8dWH6fpCnsO7b7+2aWV5c2nHPOuRZsNc08iduUlLQxNytz6ftZmTtnp6e33pCc1AvpcODwqGIa8XrpMlXRy9dn8bOxeUdf22DxSJJSOh+elNL58JTMwQCYFe2IFa9cEitesilWsjLTYtsOBTuowYKKhvdOtxCexEXAzCr85so555xztbaaYD3VZqEYiudkpH/6flbm+kmZGSmfpaZ2Dycf6RB1bGXab7d1R67a+1648jqvn3NcfffGVUdKa5WcdvixyWlf5rqx0i1rYsXL8mPFy4pipZ8fgBX2BjKjirEeeBLXQngS55xzzrmmrElftBakJK8en5X52YSsrKL5aWkdtiV48pH6cP242ALBGdWV67v4mVje0T9uiJBqLCm5Xdek5OO6knEcAGaxEitds7C0aMm6WMlnSVa6qRuUZhMszdQUrYo6ANcwPIlzzjnnXFPWZC5ad0k7p2RkfPpeq8zNUzMyMsLJRw4GDo46tppqvcs2D1hm+0ykVpFO6+cel1q8fXZxausB9RxWnUlJKUrp1jcppVvfsn0W27U5VpK/tLRo6bZYyarW2I7DgQMiDLM2mvSXGq7mPIlzlerfrR3TR10QdRjOOedcVRrlRauBfZKauvyDVplrPszMjH2atmfykciGFybCD96KzalJL1yZPoueJa9/4+qNq46SMtsnp/UdmJy2J68jVrohP1a0dFVp8bJSK13XCYp7AanRRVmpRvl5cInnSZxzzjnnmrJGcdG6OSlpU25WxtL3s7K2z05Pb7M+mHzkMCqYgr+pSi+yHYMXWv/anNNpw9wBqUXbZxenNd7euJpISu6YnZTZMTslM7gV0Kxkd6ykYFGsaMnGWMmKNIttOQSsW8RhQiP5PLj650mcc84555qyBr9oLYGSOenpS95vlbl2UmZG8mepqd2KpWxgUEPH0pC+/35selIteuHK9Fn8DHn9yy+127RJKRnJqdn9k1Oz9+yz2LYvSouX58eKlu6Ola5ph+3uBbSutJL64UlcC+FJnHPOOeeaspX13cDqlOQ1EzIzP/sgK7NwfnraAVuTknoj9QH61HfbjUVKiRWePduOrMu5nTbkDUgt2jarOK3NcYmOqzFRUpsuKenHdCH9GADMYqVW+sUnpcVLv4gV52OlGw6G0p5AUj2GUe+fB9c4eBLnnHPOuaZr5JZVjGy3AeiYiOp2STunZqQveb9V1sapGekZq1NSDo1JXYGuiai/qbo0NzY12RhS1/P7Ln4maW7/6xMZUqMnJSUrpesRSSldjyDzNADMCrfGij9bWlq8ZEuseFUrbFtP4MAENbn85ufHbk1QXa6R8yTOOeecc03dDOArtT3JwJakpua/n5W5Ojcrs/STtNTOu6ReSMfUQ4xNVlLMSi+caofuTx0Hbph3bEvojauOlN42Oe2I45LTjtizL1a6aWWseOnKWPGy4ljJ2gOhqDeQVofqZyYsUNfoeRLnKpW3agvZOeOiDsM55/bI9xlzXcVqlMRtSUra/FFmxtL3W2Vtn5We1mp9cvLhJvUEetZ/iE3XhVNtSkqMU/a3nn6Lnkqec8yIRITUrCQlH9AjKXlQDzKCWyrNSotiJavmx4qXrI8Vr0i12ObuEDukBlV5EteCeBLnnHPOuaZun4vXEiiZl5625L2srHUTszKS8lNTDy6GbKSBUQTYZJnZpbmxzomoquPGBcekFm2dVZzWtkX3xlVHSk5LTj3kqOTUL/M2i+1YX1qcvyxWvGRXrGRNG2xnL6BtuVNnNGigLlKexDnnnHOuqZvxeXLy5+OzMvPHZ2Xunpee1mFLUlKvljb5SH04e7ZNSyvhxETV12/hU8lzjr0hUdW1GEpqdWBK+lEHkn4UAGYWs9J1S0uLl6yJFeebla7vAiXeE9eCeBLnnHPOuaZt5Jbl547pL+DkqENpbr7/fqxVIuvruGnhMalFW2cWp7U9PpH1tjSSkpTS+fCklM6Hk3kKwNIRo4euizou13Dqc4pT55xzzrmG8lHUATQ3Jy+Mzcwq4qhE19tv4VOpia7T+fu/pfEkzjnnnHPNgV/EJtgP34rVy3Vix00L+6cVbvH7txIrN+oAXMPyJM4555xzzYFfxCZQ/+WxvLa7GFBf9fdb9FRdptB3lfP3fwvjSZxzzjnnmoPZwLaog2gurh8X212f9XfYtKh/WuGW6fXZRguydsTooZ9EHYRrWJ7EOeecc67JyxueV4oPqUyIw1fbJx23Mai+2+m3cExGfbfRQrwfdQCu4XkSV0eSLpbUr4HbHCDp/LifR0q6pSFjcM455xqxF6MOoDn46eulGwSq73Y6bF58dFrhZu+N238vRB2Aa3iexNXdxUCDJnHAAOD86grVlKTkRNXlnHPONQKvAEVRB9GUdV9n+V03clJDtddv4ZjMhmqrmdoO/C/qIFzD8yQujqTbJS2S9I6k5yTdIulwSW9KmiEpV1IfSacA3wD+LGm2pMMrqKuzpBnh9rGSTNIh4c9LJWVJ6iTpJUnTwsep4fETJU2SNCv890hJacBdwLCwzWFhU/0kjZe0TNKNce1fIWlqWPaRsoRN0nZJd0maAgyuIO5rJU2XNL1055bEvsDOOedcPcobnrcJeDvqOJqyEWNLC9SA14cdNn9ylPfG7ZfXRoweWq/3L7rGyZO4kKRBwLeA44Bvwp6x4I8CPzWzgcAtwENmNgl4DbjVzAaY2dLy9ZnZWiBDUltgCDAdGCLpUGCtme0E/grcZ2YnhG3/Mzx9EXC6mR0H3AH8wcyKwu3nwzafD8v2Ac4DTgR+IylVUl9gGHCqmQ0ASoHLw/KtgHlmdpKZ7XPvgJk9amaDzGxQcla72r+QzjnnXLSer76Iq8iBW2zNYZ9zYkO3e9SCJ7Maus1m5D9RB+CikRJ1AI3IacCrZrYLQNLrQAZwCvCCtGdoeHot6pwEnAqcDvwB+CrBGPOyaWDPIehJKyvfVlIboB0wRlJvwICqFsUcZ2aFQKGktUAX4GxgIDAtrDsTWBuWLwVeqsVzcM4555qSV4HdBP+Hu1oYMbb0U0HXhm73gC2f9ksv3DStMP2AExq67SZuK/Bm1EG4aHgS96WKbuBNAjaHvVl1kUvQC3cowX8qvyRIysbG1T+4LHHcE4j0APCBmV0iKRsYX0UbhXHbpQS/UwFjzOxXFZTfbWaltX8qzjnnXOOXNzxvW/8x/d8ALok6lqak3Q5b329F/c9IWZl+C8a0mnXcTftVxzPj/8y8zybTJrM9t1362F7H3p3zH/47+RFGXfkyrTP3Hmm0aftanvpgFFt3bkISp/a9gLP6fwuA/05+lAUrp9K9Yy+uHJoDwNRP3mFH4dY9ZSL06ojRQwurL+aaIx9O+aWPgK9LypDUGrgA2Aksl/QdAAWODctvA9pUU+eHwBXAp2YWAzYSTEwyMTz+NnBDWWFJA8LNdsCqcPuquPpq0ibAe8C3JXUO6+0QDuN0zjnnWoJIh1QWPFbAwp8u5NPbPt3n2Po31jPvqnmUbCup8NzFNy/m019/ypLbl7Bk5JI9+z//z+d8+utPKXi0YM++TRM3sf7t9QmJ+do3YvMFkQ1rPGDLp/3Sd2+atj91nHzEeYw4/4/77N+0fS2LCmZwQOvOFZ6XpGS+efJ13D7sCW65+O98OP9V1mzKZ1fhdpZ/MZ//951/ErMYqzYso6ikkMmL3+L0fhftT6iJ4kMpWzBP4kJmNo3gPrc5wMsE97BtIbiX7AeS5gDzgbJP7b+BW8PJR/aZ2CSsMz/c/DD89yOCnr1N4c83AoMkzZW0ALgu3P8n4I+SJgLxM0h+QDD8Mn5ik4raXQD8Gnhb0lzgHSIYHuGcc85F5HVgR1SNH3DaAWTfnL3P/qINRWyfv53UjlXdJQE9f9mTXr/tRa+RvQAo3VnKziU76f273ljM2L1yN7GiGJs/2kzHoR33O96s3bZl0Kd1HnWUMP0WPtlqf87vdfAxZGW03Wf/S5Me4uKTr0WVrJrQrlVHenQ6AoCMtCwOan8om3esR0qiJFaCmVFcUkhyUgrvzXmeM/tfQnJy5IPZNuOT+LRokb8DG5l7zGykpCyCxOteM1tOcC/bXsxsIjVYYsDMDonb/gPBvXFlP68nmICk/DkfA0fE7bo93L8RqHS8uJkdHbf9PBV8E2lmrauL2TnnnGvK8obn7ew/pv9YKvg/tiG0OrIVRev2Xeng8+c+p8ulXVjx1xW1q1BgJYaZYcWGksX6N9bT8dyOKGX/l3O7+p3YbMEZ+13Rfjpgy5J+6bs3Ti3M6JCwyVXm5k+ifasD6d6xwu/b97Fh2+cUbFhCdue+ZKRlMaDnEEa99GOO7HYcmWmt+GztYr428MpEhbc/Xh0xeqgvp9GCeRK3t0fDBbwzCO4pmxl1QM4555yrk+eJKImryNZZW0k9IJXMQ6pZFk2Qf08+AB3O6kCHMzuQnJlM20FtWXrHUlr1a0VSVhK7lu2i80UVDw+sjbRi2zVknh213xUlyFELnmgz8/ibE1JXUfFu3pr1LDecf3eNyhcW7+Kfb4/kW4N/QmZa0Cl47oDvcu6A7wLw7IR7uOCEq5i0cBwLC2bQreNhfPX4KxISax38O6qGXePgSVwcM/teXc6T9CDBLJTx/mpmT+x/VM4555yrg3HAGhrB7QSxwhjrXl9H9i3Z1ZY97LbDSD0glZKtJeT/OZ/0rum0OrIVnc7vRKfzOwGw6vFVdP5mZzZO2Mj2edvJ6JFB52/ULaH73vjYtKRgFu1Gof3WZX0T1Ru3butqNmz9nD++eC0Am3es4+6Xr+PWSx6kbVaHvcqWlpbwj7dHMqj32Qw4bMg+da1cH9zf2Lldd16c+CA/v+h+Hn/3t6zdUkDndt33N9TaWgS81dCNusbF74lLADMbEa7dFv/wBM4555yLSN7wvCLgL1HHAVC0toiidUUsuX0Ji29eTPGmYpb+ZinFm4v3KZt6QHC/XErbFNoc34Zdy/aawJpdnwU/px+UzuaJmzlkxCEUFhRS+HntJylMLrXi82ZYrzo8pXp11ILH972xrQ66dTyMUcNf4q7L/8Vdl/+L9q068ctvjt4ngTMznp1wDwe1P4Szj/lOhXWNnfYEFwy6itJYKcFcdSAlUVQSyeSQfx4xeqhF0bBrPDyJc84551xzNRrYVG2pepbRI4O+D/TlyHuP5Mh7jyT1gFQOv/NwUtvvPcFJrDBG6a7SPdvb528nvdvey9OufXktnS/pjJUYxMKdSRArilFb35oYm5JsHFynJ1WP2m9d3idj94YptT3viXd/x73//SlfbFnJr58ZxqRF/6u07OYd63nof8FKTMs+n8fUT9/hk9Wz+OOL1/LHF69l/oovm5+z/CMO7dSH9q0OJCu9Ndld+vH7F36IUI3vtUugVcAzDd2oa3xk5om8q1h6197Wdfj9UYfhnHN75I+6oM7nSpphZpGtg+Wi0X9M/7sIJwhrKCsfXsmORTso2V5CStsUOl/cmQ5nfNn7s/jmxRw+8nBS2qRQvKmYVU+sIvsX2RStLWLFA8GkJ1ZqtDu53V7DJLfO2MrulbvpfHGwb82/1wTDKbtn0OO6HrWKUWaxZ/5UuiI1Rvb+P+PE29K25+IZx99yZNRxNEK3jBg99N6og3DR8yTOVWrQoEE2ffr0qMNwzrmE8CSuZeo/pv+BwGdEuAZaY3T+1NjHV70XGxx1HFWZdNJdU3Zndjwp6jgakU3AoSNGD90WdSAuej6c0jnnnHPNVt7wvPXAP6KOo7G5bEKsQ/WlonXUgsfb470N8R7yBM6V8STOOeecc83dPcC+s4i0UGfMjU1LL6HRD1Vsty3/yIzdG6ZGHUcjsQv4W9RBuMbDkzjnnHPONWt5w/MKgKejjqOxuOrdWEbUMdTU0QseP8B74wB4YsTooWujDsI1Hp7EOeecc64luJsv53NssQZ9EpvdqpD+UcdRU223fXZExu71tZ6pspkpJehNdm4PT+Kcc8451+zlDc/7BHg56jiidu0bsSaXyB694PGOLbw37vkRo4cujzoI17ikRB2Aa7zyVm0hO2dc1GE451qw/VlSwLkK3AlcTAu9/un3mS1ov5Pjo46jttpuW9E7c/f6j3dldmrUs2nWkx1ATtRBuMbHe+Kcc8451yLkDc+bB9wfdRxR+cm40iY7s+FR8x87sIX2xt01YvTQlVEH4RofT+Kcc84515KMBFZEHURD6/m5Lem0hROjjqOu2m5f2Ttz17rJUcfRwOYD90UdhGucPIlzzjnnXIuRNzxvB3BD1HE0tBteL10rUNRx7I+jFzzeqYX1xl0/YvRQXxrDVciTOOecc861KHnD814HXok6joZy0EZb2X09J0Udx/5qs31lrxbUG/fUiNFDc6MOwjVensQ555xzriW6EdgedRAN4aevl+YLkqOOIxHC3rgmN8NmLW0Cbo06CNe4eRLnnHPOuRYnXAD89qjjqG8dt9rnvVY33Xvhygt649Y2996423xhb1cdT+Kcc84511I9AMyKOoj6dP242GJBetRxJNLR8x/r0ox746YBj0QdhGv8PIlzzjnnXIuUNzyvFPgx0CwTgtY7bVP/fBtUm3NuW7OG05Z8yjeWL9vn2OMbN9Bv8SI2lZRUeO45S5dw0fLlXJK/nO/k5+/Zf++6tVy8fDk5a1bv2ffali08vWljbULbo82OVYdn7fyiOfbGxQgmM2mW70eXWE06iZOULWleAurJl3RgImKqpp2bJGXVdzuNPQbnnHOuscgbnjcNeDDqOOrDtW/G5gpa1eacS9q149HuPfbZv6a4mI937KRrStXrpD/ZowevZPfkhexsALaVljJr1y7+27MnpQafFO5mdyzGK1u38N32B9QmtL0cteDx5tgbd/eI0UNnRB2EaxqadBIXJQVq+/rdBCQkgZJU9V/RBojBOeecayZ+CeRFHUQiZRbatpMW27G1PW9QVhbtkve9vLl77Vpu7tSp1msUJAmKzTAzCi1GCuLxjRu5ov0BpKruKx40w964j2gB92i6xGkOSVyKpDGS5kp6UVKWpLMlzZKUJ+lxSekAle0vIylT0puSflRRQ2HP30JJDwEzgR6SbpU0LWz/zrBcK0njJM2RNE/SMEk3AgcDH0j6ICz3sKTpkuaXnRvu39MzKGmQpPHh9khJj0p6G3gqjCdX0szwcUpY7kxJ48PXY5GkZ8Okc58YKniO14YxTS/duaXuvxXnnHOuicgbnrcL+DawLepYEmX4u7GZgvaJqOv97dvonJJCn4yMKstJ4ocFK/l2/nL+s3kzAK2SkvlK6zZ887N8uqWm0iY5mXm7d3F2mzb7HdfRCx47qJn0xm0ALhsxemhp1IG4pqM5JHFHAo+a2THAVuAXwJPAMDPrD6QA10vKqGh/XD2tgdeBf5nZP6pp7ykzOy7c7g2cCAwABko6HfgqsNrMjjWzo4E3zexvwGrgLDM7K6zrNjMbBBwDnCHpmBo834HARWb2PWAtcK6ZHQ8MA/4WV+44gl63fsBhwKmVxLAXM3vUzAaZ2aDkrHY1CMc555xr+vKG530C/DDqOBIhtcR2n5lnfRNR165YjEc2bOCnB1Z/18mzhxzCS9k9eaR7D57bvInpO3cC8IOOHXkluye/7NyFv61fxw0HduLFzZv5+epVjN6wvs6xtd6x+rCsnZ9/XOcKGgcDrhwxemhB1IG4pqU5JHErzWxiuP0McDaw3Mw+CfeNAU4nSLgq2l/mVeAJM3uqmvY+M7Oy7vuvhI9ZBD1zfQiSujzgHEl3SxpiZpV1aV0qaWZ4/lEECVd1XjOzXeF2KvAPSXnAC+XOn2pmBRZ8QzUbyK5B3c4551yLlTc87z/A36OOY38NmxCbmmR0TkRdK4uLWFVczCX5yzln6RK+KCnhW5/ls66CyU06p6QC0DElhbNbt2bu7l17HV+wezcA2WlpvLp1C/cd3I1PCwvJLyqqc3xHL3j8YMyacg/Wn0eMHvq/qINwTU9zSOKshuWqG3g9EfiaVO0A7R3l6vyjmQ0IH73M7LEwURxIkMz9UdId+wQj9QRuAc4OexHHAWXjFEr48ndTfuxCfPs/B74AjgUGAWlxxwrjtksJeh6dc845V7WbgSbbu5MUs5Lzp9thiarviPQMPurVm3cP78W7h/eiS0oKLx2aTadyE5zsjMXYESvdsz1px056p++9ssED69fx0wMPpMSMWHj1loTYHav7iMjWO1b3bLVzTVO9N+5t4P9FHYRrmppDEneIpMHh9mXAu0C2pF7hvu8DE4BFlewvcwfBmOSHatH2W8A1kloDSOomqbOkg4GdZvYMcA9wfFh+G1A2CLwtQUK2RVIX4Gtx9eYTJIEA36qi/XbAmrC37ftAcg1ijo/BOeecc3HyhucVAZcAK6OOpS4u/timpMToXtfzb1m9iss++4z8oiLOWrqEl8J72yqytqSYHxcEL9OGkhKuWLGCS/KXM+yzfE5v3YohrVrvKfvutm0cnZFJ55RU2iYnc2xmJhctXw5Q7b121TlqfpPsjVsKfNfvg3N11Rx6ZxYCwyU9AnwK/AyYDLwQzuA4DRhtZoWSri6/v1xdNwGPS/qTmf1fdQ2b2duS+gIfhx1424ErgF7AnyXFgGK+vPfuUeANSWvM7CxJs4D5wDKCnsAydwKPSfp/wJQqQngIeEnSd4AP2LuXrjJ7xVCD8s4551yLkjc874v+Y/pfRDBjYJOZ0VlmsW9NjHXdnzruObhblcffPbzXnu3OKak8Ei5H0CMtjVeye1Z63jlt2nBO3GQm/9c5IaM9AWi9c03PVjvWTNzR+uBTE1Zp/doOXDxi9NBNiaxU0k0E80TsrMO5+cAgM6v7TYpBPU8CY83sxf2px1VPZjUdjehamvSuva3r8PujDsM514Llj7ogYXVJmhFOJuX2g6SLgU/MbEEDtjkAONjM/hf+PBLYbmb31Ge7/cf0vxR4vj7bSKSvzIhN/uHbsZOjjiMKO7IOyp9ywq97INVkVFKUDPjOiNFDX6rLyeFtP7IKZuXcn0TMk7impzkMp3TOOedcw7mYmk3ElUgDgPMTVZlqeKEfTnRyZ7UFG4krPoi12GmlW+38PLvVjtVN4V7G62ubwFWwxNVj5ZeoqmQpq69I+jhchuqFstt/qnCrpKnho1dYx6GS3guX0npP0iFV7S8X928lPanar6vsasBf1ApI6ihpdgWPjlHH5pxzziWapNvDdUXfkfScpFskHa5g7dQZ4ZqkfcL1SL9BcMvAbEmHV1BXZ0kzwu1jJVnchd9SBeu5dpL0koJ1VqdJOjU8fqKkSQrWdJ0k6UhJacBdwLCwzWFhU/0UrIm6LLyALWv/ivAidLakR8oSNknbJd0laQowmBrKG543kuD+9kbt1PmxGRnFJGRZgabq6AWP9Wjk98b9bMTooY/U8dz4Ja5uLr9EVfllpBSsN/xr4JxwKarpBMtwVWWrmZ1IMEPr/eG+v4ftHgM8y5fLWVW2HwBJfwI6A1dX1Gvo9p8ncRUwsw1xM07GPzZEHZtzzjmXSJIGEUyidRzwTYLZjiG4h/qnZjaQYDblh8xsEvAacGv4/+LS8vWZ2VogQ1JbYAjBxeMQSYcCa8P7df4K3GdmJ4Rt/zM8fRFwenihegfwBzMrCrefD9ssG97YBziPYK3W30hKDe9TH0awNuoAgtmZLw/LtwLmmdlJZvZRbV6jvOF5twL31uachnbN27HUqGOIWqudXxzaeseqxjpT5f+NGD30b9UXq1T8Elc1WaLq5HD/REmzgeHAodW08Vzcv2VfdAwG/hVuPw2cVs1+gNuB9mb2Y/P7tupNc5jYxDnnnHN1dxrwatkapJJeJ1je5hSCycDKyqVXfHqFJgGnEqzH+gfgqwTL8uSGx88h6EkrK99WUhuCWZfHSOpNcO9QVYnJODMrBAolrQW6EKwVOxCYFtadCawNy5cCdboPCSBveN4t/cf0h2AJgkZlwNLY3Da7OSbqOBqDo+Y/3n3KibeXEExi11jcPmL00D/vZx07YK8lqk4ws03hPWgVTe8p4B0zu6wWbVgl25WVqWz/NGCgpA5mtrEW7bta8J4455xzrmWraH3UJGBzudEotRmql0vQC3co8CrBeqanAR/G1T84ru5uZrYN+C3wgZkdDXydii9Oy1S0HqqAMXH1HmlmI8Myu20/h9rlDc+7BfjL/tRRH677X6zuq2U3M612fXFo6+2Nqjfu9yNGD/1dAuuraomq+GWkJgOnxt3bliXpiGrqHhb3b9n9hZOA74bblxPM2FrVfoA3gVHAuPDLGVcPGtO3FK6R6d+tHdMTODOcc865Rukj4BFJfyS4LrgA+AewXNJ3zOyFcEa8Y8xsDjVbb/RD4HfAh2YWk7SRYGKSX4XH3wZuAP4MweyTZjaboCduVVjmqrj6arrG6XvAq5LuM7O1kjoAbczssxqcWyN5w/Nu7j+mv4CfJ6rO/XFEgS3qsB2fdTXO0QseO2TyiXc0ht64e0aMHvrrRFZoZnNU+RJV5Zeyugp4TlJZL/qvgU+qqD49vGc0iWDtZYAbCZbfuhVYB1xdzf6yOF8IE7jXJJ1f1tPvEseXGHCVGjRokE2fPj3qMJxzLiHkSwxUSsGU/ZcBnxFckI0H3gUeBroSDGv8t5ndFU5C8g+CnrBvV3RfXFjnCuB3ZvaognVPvxtOgkA46cKDQF+CxPFDM7tO0mBgTBjD+8D3zSw7TMbeCuP4Y3jeniUGJM0DLjSz/HDik18RXIgWAyPMbLKk7WZW3ex8NdZ/TP/7CNaXjdRfR5dM7rqJFrmsQFWmDszJ3d6mx5AIQ/jbiNFDfxZh+66Z8yTOVcqTOOdcc+JJXOUktTaz7ZKyCHrRrjWzmVHH1dhFncgdstaW/fmx0mz57TH72JnZaeXkE3/TNaLeuIdHjB76kwjadS2If+idc84592g4g91M4CVP4Gomb3jez4lw1sobXi9d4wlcxbJ2revRZvvKhl43rhS4uTEncJJeqWAJrfOijsvVXtRjhZ1zzjkXMTP7Xl3Ok/QgwSyU8f5qZk/sf1RNQzhr5XyCoae1mcFzv3TZZAWHruWkhmqvKTpqwePZk0/8TTFSQyy/sAX47ojRQ99sgLbqzMwuiToGlxj+7Y1zzjnn6sTMRlSwpmqLSeDK5A3Pe4JgNs6ChmrzJ2NLl8m/jK9S0Bu3oiFmqvwEOKmxJ3CuefF74lyl0rv2tq7D7486DOdcC5JfjzPi+j1xrr71H9O/M/Af4Iz6bKf9dlv3yAOlbVT1EgwO2Jl5YMHkE0d2qcfeuLeBYSNGD91cT/U7VyHviXPOOeecS4C84XlrCRYyf6A+27l+XGyBJ3A1k7Vrffc22+qtN+4+4HxP4FwUPIlzzjnnnEuQvOF5JXnD824EhgO7E11/q122ZcAyOz7R9TZnRy94vCdmxQmssgi4ZsToob8YMXrofi0g71xdeRLnnHPOOZdgecPzngJOA1Ykst4fvB2brZotfO5CmbvXd2+77bNE9catBs4aMXpoi7v30zUunsQ555xzztWDvOF5M4BBBPdN7bf0IttxygLrn4i6WpqjFjx+GGZF+1nNP4F+I0YPnZSImJzbHz6rkXPOOedcPckbnrcOOK//mP5XAH8BOtW1ru+/H5ueVM+TpjRXmbs3dGu7LT93a9ueQ+pw+jLgRyNGD30/0XE5V1feE+ecc845V8/yhuc9A/QBHgNqPTV4SqkVnT3bjkx4YC3IUQseP7yWvXExgslL+nsC5xob74lzzjnnnGsAecPzNgI/7D+m/1PAaKBvTc/9Tm5sSrJRl14kF8rcvfHgtluXf7i13WGn16D4AuAHI0YPbYh15pyrNe+Jc84555xrQHnD8z4EBgB3AIXVlU+KWenXp9gh9R1XS3DUgid6V9MbVwz8DjjOEzjXmLXoJE7SVZL+Xsmx7VWcly3pe/UQT6VtVnNetqR54fYgSX+rpmzCY3fOOedczeUNzyvKG573W6A/UOVQvQun2pSUGIc2TGTNW2bhxq7tti6rLDmbAgwaMXro7SNGD93fSVCcq1ctOonbD9lAo0yEzGy6md1YRZFsGmnszjnnXEuTNzzv07zheWcTrCtXsE8BM/tObqzOk6G4ffVb8GRvzOJ7QOcDl4wYPfTkEaOHzo0qLudqo1kncZL+K2mGpPmSrg33XS3pE0kTgFPjyvaU9LGkaZJ+W03Vo4AhkmZL+rmkDElPSMqTNEvSWVXEdJWkVyW9KWmxpN9UUKa1pPckzQzrvCjc/1tJP4sr93tJN5Y790xJY8PtM8IYZ4dxtSkfewVtXytpuqTppTu3VPMyOOeccy4RwnXlegE3EJfMDZ1j09JL6B1ZYM1Q2Bs3BVhOkDwfM2L00P9GG5VztdOskzjgGjMbSLBGy42SugF3EiRv5wL94sr+FXjYzE4APq+m3hwg18wGmNl9wAgAM+sPXAaMkZRRxfknApcTjIf/jqRB5Y7vBi4xs+OBs4B7JYlgRqvhAJKSgO8Cz1bRzi3ACDMbAAwBdlUQ+17M7FEzG2Rmg5Kz2lX1GjjnnHMugfKG5xXmDc97kLhk7sr3Yq0iDqs5Wt530dNPAn1GjB761IjRQ2NRB+RcbTX3JO5GSXOAyUAP4PvAeDNbZ8FNrc/HlT0VeC7cfrqW7ZxWdo6ZLQI+A46oovw7ZrbBzHYBL4fnxxPwB0lzgXeBbkAXM8sHNkg6DvgKMMvMNlTRzkTgL2FvXXszK6nl83LOOedcAytL5gYsjR2eVcRfgUVRx9RMLACuBI4YOOvDJ/y+N9eUNdskTtKZwDnAYDM7FphF8EewqrVZar1uS1lztSxfvp3yP19OsBjowLAX7QugrGfvn8BVwNXA41U2YjYK+CGQCUyW1KeWcTrnnHMuIk/fNb+o76KF/yAYOfQNIDfikJqqScC3gKP7Llr4dN9FC/1LbdfkNed14toBm8xsZ5i8nEyQzJwpqSOwFfgOMCcsP5FgeOIzBElUVbYBbeJ+/jA8531JRwCHAIurOP9cSR0IhjdeDFxTQexrzaw4vL8ufkaqV4C7gFSqmaBE0uFmlgfkSRpMsMjoynKxO+ecc64R67tooQGvA68v7NP3ROA64BKgfZRxNXL5BKOknu67aOGnEcfiXMI15yTuTeC6cEjiYoIhlWuAkcDH4fZMIDks/zPgX+HEIS9VU/dcoCQcqvkk8BAwWlIeUAJcZXvPelTeRwR/WHoB/zKz6eWOPwu8Lmk6MJu4YRRmViTpA2CzmZVWE+dNYRJYSjCE4A0gFh97RffFOeecc65x6rto4VRg6sI+fa8DvkrwBfQ3AL93LviC/gXgKSA3TH6da5Zk5u/vhiTpKmCQmd1Qx/OTCJLP75hZvX6zlN61t3Udfn99NuGcc3vJH3VBvdUtaYaZlZ9Iyrkmb2GfvlnAhQQJ3flAerQRNahS4G2CxO3VvosW7oo4HucaRHPuiWt2JPUDxgKv1HcC55xzzrmmoe+ihTuB/wD/Wdinb1uCoZbfJZgboDle620nGFX1BvCvvosWfhFxPM41uOb4wU4YSf3Zd6bKQjM7qQbnngfcXW73cjO7hGAIZq2Z2QLgsLqc65xzzrnmr++ihVuBMcCYhX36diCYfftkYDDBEkdNcdjlJoJbUT4MHzN9chLX0nkSV4VwUpABdTz3LeCthAbknHPOOVdDfRct3Eg4IQrAwj59k4GjCRK6wQTJXVVLIkXlc4KZOMuStjy/v825vXkS55xzzjnXAvRdtLCUYFbuOcBogIV9+nYkSOZOBHoSrE3bDTiY+p3NeiPBDJLxj8+AhT6bpHPV8yTOOeecc66F6rto4QZgXPjYy8I+fduwd1IXv92KYL3h5PDfJIJ1cwvjHrvDfzcSJGj5ZY++ixZuq79n5Vzz57NTukoNGjTIpk8vv/qBc841TT47pXPOueYiKeoAnHPOOeecc87VnCdxzjnnnHPOOdeEeBLnnHPOOeecc02IJ3HOOeecc84514R4Euecc84555xzTYgncc4555xzrl5JGi+pUc8OK+kuSedEHYdzNeHrxDnnnHPOuRbPzO6IOgbnasp74pxzzjnnXEJIypa0SNIYSXMlvSgpq1yZhyVNlzRf0p3hvrMlvRJX5lxJL1fRznZJd0uaIeldSSeGvX3LJH0jLHOVpL/HnTNW0pmSkiU9KWmepDxJPw+PPynp2+H2CZImSZojaaqkNol9pZzbP57EOeecc865RDoSeNTMjgG2Aj8pd/w2MxsEHAOcIekY4H2gr6ROYZmrgSeqaKMVMN7MBgLbgN8B5wKXAHdVE98AoJuZHW1m/cu3IykNeB74mZkdC5wD7KqmTucalA+ndJWaMWPGdkmLo44jdCCwPuogQh5LxTyWinksFYsilkMbuD3nWqqVZjYx3H4GuLHc8UslXUtwHdoV6GdmcyU9DVwh6QlgMHBlFW0UAW+G23lAoZkVS8oDsquJbxlwmKQHgHHA2+WOHwmsMbNpAGa2tZr6nGtwnsS5qiwOvymLnKTpHsu+PJaKeSwV81iccw3EKvtZUk/gFuAEM9sk6UkgIzz8BPA6sBt4wcxKqmij2MzK6o0BhQBmFpNUdn1bwt6jzjLCMpskHQucB4wALgWuiSunCp6Dc42KD6d0zjnnnHOJdIikweH2ZcBHccfaAjuALZK6AF8rO2Bmq4HVwK+BJxMQRz4wQFKSpB7AiQCSDgSSzOwl4Hbg+HLnLQIOlnRCWL5NXGLoXKPgb0jnnHPOOZdIC4Hhkh4BPgUeBr4OYGZzJM0C5hMMa5xY7txngU5mtiABcUwElhMMt5wHzAz3dwOekFTWmfGr+JPMrEjSMOABSZkE98OdA2xPQEzOJYQnca4qj0YdQByPpWIeS8U8lop5LM65hhAzs+vK7TuzbMPMrqri3NOAf1TXgJm1jtseWdGxcLjl5ZVUUb73ba+4wvvhTq4uDueioi+HEzvnnHPOOVd3krKBsWZ2dB3OnUEw1PJcMytMdGzONSeexDnnnHPOuUZJ0hQgvdzu75tZXhTxONdYeBLnnHPOOeecc02Iz07p9iHpq5IWS1oiKacB2ush6QNJCyXNl/SzcP9ISaskzQ4f58ed86swvsWSzktwPPmS8sI2p4f7Okh6R9Kn4b8H1Hcsko6Me+6zJW2VdFNDvS6SHpe0VtK8uH21fh0kDQxfzyWS/iZJCYrlz5IWSZor6RVJ7cP92ZJ2xb0+oxsgllr/ThIRSxXxPB8XS76k2eH+enttqvgcR/Kecc4555o1M/OHP/Y8gGRgKXAYkAbMIViEsz7b7AocH263AT4B+gEjgVsqKN8vjCsd6BnGm5zAePKBA8vt+xOQE27nAHc3RCzlfi+fEyxW3CCvC3A6wY3f8/bndQCmEizaKuAN4GsJiuUrQEq4fXdcLNnx5crVU1+x1Pp3kohYKoun3PF7gTvq+7Wh8s9xJO8Zf/jDH/7whz+a88N74lx5JwJLzGyZmRUB/wYuqs8GzWyNmc0Mt7cRTE3crYpTLgL+bWaFZrYcWBLGXZ8uAsaE22OAixs4lrOBpWb2WTUxJiwWM/sQ2FhBGzV+HSR1Bdqa2cdmZsBTcefsVyxm9rZ9uRDsZKB7VXXUZyxVqNfXpbp4wh6sS4HnqqojEfFU8TmO5D3jnHPONWeexLnyugEr434uoOqEKqEUzGp1HDAl3HVDOFzu8bhhWPUdowFvS5oh6dpwXxczWwPBxSrQuYFiKfNd9r4Qj+J1gdq/Dt3C7fqMCeAagh6bMj0lzZI0QdKQuBjrM5ba/E4a6nUZAnxhZp/G7av316bc57ixvmecc865JsuTOFdeRfeeNMjsN5JaAy8BN5nZVoLFQQ8HBgBrCIaFNUSMp5rZ8cDXgBGSTq+ibL2/XpLSgG8AL4S7onpdqlJZ2w3x+twGlBAsEAvBa3KImR0H/AL4l6S29RxLbX8nDfW7uoy9k/96f20q+BxXWrSSNqN8HzvnnHNNgidxrrwCoEfcz92B1fXdqKRUggu/Z83sZQAz+8LMSs0sRrDwZ9nQwHqN0cxWh/+uBV4J2/0iHOZVNvRsbUPEEvoaMNPMvgjjiuR1CdX2dShg72GOCY1J0nDgQuDycOgd4fC8DeH2DIJ7rY6oz1jq8Dup19cFQFIK8E3g+bg46/W1qehzTCN7zzjnnHPNgSdxrrxpQG9JPcMeoO8Cr9Vng+F9O48BC83sL3H7u8YVuwQom33vNeC7ktIl9QR6E0yEkIhYWklqU7ZNMHnGvLDN4WGx4cCr9R1LnL16U6J4XeLU6nUIh89tk3Ry+Hu+Mu6c/SLpq8AvgW+Y2c64/Z0kJYfbh4WxLKvnWGr1O6nPWOKcAywysz1DE+vztansc0wjes8455xzzUbUM6v4o/E9gPMJZpZbCtzWAO2dRjBcai4wO3ycDzwN5IX7XwO6xp1zWxjfYhI4cx3BrJxzwsf8sucPdATeAz4N/+1Q37GEdWcBG4B2cfsa5HUhSBzXAMUEvSM/qMvrAAwiSGqWAn8nXJ8yAbEsIbinquw9Mzos+63wdzcHmAl8vQFiqfXvJBGxVBZPuP9J4LpyZevttaHyz3Ek7xl/+MMf/vCHP5rzwxf7ds4555xzzrkmxIdTOuecc84551wT4kmcc84555xzzjUhnsQ555xzzjnnXBPiSZxzzjnnnHPONSGexDnnnHPOOedcE+JJnHOuyZB0k6SsqONwzjnnnIuSLzHgnGsyJOUDg8xsfdSxOOecc85FxXvinHMJJelKSXMlzZH0tKRDJb0X7ntP0iFhuSclfTvuvO3hv2dKGi/pRUmLJD2rwI3AwcAHkj6I5tk555xzzkUvJeoAnHPNh6SjgNuAU81svaQOwBjgKTMbI+ka4G/AxdVUdRxwFLAamBjW9zdJvwDO8p4455xzzrVk3hPnnEukocCLZUmWmW0EBgP/Co8/DZxWg3qmmlmBmcWA2UB24kN1zjnnnGuaPIlzziWSgOputC07XkL4N0iSgLS4MoVx26X4qAHnnHPOuT08iXPOJdJ7wKWSOgKEwyknAd8Nj18OfBRu5wMDw+2LgNQa1L8NaJOoYJ1zzjnnmiL/dts5lzBmNl/S74EJkkqBWcCNwOOSbgXWAVeHxf8BvCppKkHyt6MGTTwKvCFpjZmdlfhn4JxzzjnX+PkSA84555xzzjnXhPhwSuecc84555xrQjyJc84555xzzrkmxJM455xzzjnnnGtCPIlzzjnnnHPOuSbEkzjnnHPOOeeca0I8iXPOOeecc865JsSTOOecc84555xrQv4/cakILuZbCm8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Plot class distribution\n",
    "\n",
    "plt.figure(figsize=(15,15))\n",
    "fig, axs = plt.subplots(1,2)\n",
    "plt.subplots_adjust(right = 1.7)\n",
    "grouped = snips_df.groupby(\"intent\").count().reset_index()\n",
    "axs[0].barh(grouped[\"intent\"], grouped[\"text\"])\n",
    "axs[0].set_xlabel(\"count\")\n",
    "axs[1].pie(grouped[\"text\"], labels=grouped[\"intent\"], startangle=90, autopct='%1.1f%%')\n",
    "axs[1].axis('equal')\n",
    "plt.savefig(\"class_dist.jpg\", bbox_inches='tight', dpi = 300)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Prepare data for training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(11587,) (2897,) (11587,) (2897,)\n"
     ]
    }
   ],
   "source": [
    "#Split the dataset into train and test\n",
    "\n",
    "TEST_SPLIT = 0.2\n",
    "RANDOM_STATE = 10\n",
    "np.random.seed(RANDOM_STATE)\n",
    "tf.random.set_seed(RANDOM_STATE)\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(snips_df[\"text\"], snips_df[\"intent\"], \n",
    "                                                    test_size = TEST_SPLIT, random_state = RANDOM_STATE)\n",
    "\n",
    "print(X_train.shape, X_test.shape, y_train.shape, y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Extraction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Initialize a Tokenizer\n",
    "\n",
    "tokenizer = Tokenizer()\n",
    "tokenizer.fit_on_texts(list(X_train))\n",
    "\n",
    "#Convert text to sequences\n",
    "X_seq = tokenizer.texts_to_sequences(list(X_train))\n",
    "X_test_seq = tokenizer.texts_to_sequences(list(X_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "35\n",
      "32\n",
      "35\n"
     ]
    }
   ],
   "source": [
    "X_seq_len = [len(x) for x in X_seq]\n",
    "X_max_seq_len = max(X_seq_len)\n",
    "print(X_max_seq_len)\n",
    "\n",
    "X_test_seq_len = [len(x) for x in X_test_seq]\n",
    "X_max_test_seq_len = max(X_test_seq_len)\n",
    "print(X_max_test_seq_len)\n",
    "\n",
    "MAX_SEQ_LEN = max(X_max_seq_len, X_max_test_seq_len)\n",
    "print(MAX_SEQ_LEN)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(11587, 35) (2897, 35)\n"
     ]
    }
   ],
   "source": [
    "# MAX_SEQ_LEN = 35\n",
    "\n",
    "#pad the sequences\n",
    "X = pad_sequences(X_seq, maxlen = MAX_SEQ_LEN, padding = 'post')\n",
    "X_test = pad_sequences(X_test_seq, maxlen = MAX_SEQ_LEN, padding = 'post')\n",
    "\n",
    "print(X.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(11587, 7) (2897, 7)\n"
     ]
    }
   ],
   "source": [
    "#Convert labels to one-hot vectors\n",
    "\n",
    "y = y_train.to_numpy()\n",
    "encoder = LabelEncoder()\n",
    "encoder.fit(y)\n",
    "\n",
    "encoded_y = encoder.transform(y)\n",
    "y_train_encoded = utils.to_categorical(encoded_y)\n",
    "\n",
    "y_test = y_test.to_numpy()\n",
    "encoded_y_test = encoder.transform(y_test)\n",
    "y_test_encoded = utils.to_categorical(encoded_y_test)\n",
    "\n",
    "print(y_train_encoded.shape, y_test_encoded.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10399\n"
     ]
    }
   ],
   "source": [
    "VAL_SPLIT = 0.1\n",
    "BATCH_SIZE = 32\n",
    "EPOCHS = 7\n",
    "EMBEDDING_DIM = 16\n",
    "NUM_UNITS = 16\n",
    "NUM_CLASSES = len(snips_df['intent'].unique())\n",
    "VOCAB_SIZE = len(tokenizer.word_index) + 1\n",
    "print(VOCAB_SIZE)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LSTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding (Embedding)        (None, 35, 16)            166384    \n",
      "_________________________________________________________________\n",
      "lstm (LSTM)                  (None, 16)                2112      \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 7)                 119       \n",
      "=================================================================\n",
      "Total params: 168,615\n",
      "Trainable params: 168,615\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "lstm_model = Sequential()\n",
    "lstm_model.add(Embedding(input_dim = VOCAB_SIZE, output_dim = EMBEDDING_DIM, input_length = MAX_SEQ_LEN, mask_zero = True))\n",
    "lstm_model.add(LSTM(NUM_UNITS, activation='relu'))\n",
    "lstm_model.add(Dense(NUM_CLASSES, activation='softmax'))\n",
    "\n",
    "lstm_model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=[Precision(), Recall(), 'accuracy'])\n",
    "\n",
    "lstm_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/7\n",
      "326/326 [==============================] - 10s 25ms/step - loss: 1.7596 - precision: 0.7442 - recall: 0.0585 - accuracy: 0.3502 - val_loss: 0.7296 - val_precision: 0.9913 - val_recall: 0.4892 - val_accuracy: 0.9223\n",
      "Epoch 2/7\n",
      "326/326 [==============================] - 6s 17ms/step - loss: 0.5155 - precision: 0.9824 - recall: 0.6840 - accuracy: 0.9437 - val_loss: 0.2692 - val_precision: 0.9801 - val_recall: 0.8913 - val_accuracy: 0.9603\n",
      "Epoch 3/7\n",
      "326/326 [==============================] - 5s 16ms/step - loss: 0.2132 - precision: 0.9851 - recall: 0.9100 - accuracy: 0.9725 - val_loss: 0.1658 - val_precision: 0.9838 - val_recall: 0.9448 - val_accuracy: 0.9733\n",
      "Epoch 4/7\n",
      "326/326 [==============================] - 5s 17ms/step - loss: 0.1079 - precision: 0.9928 - recall: 0.9644 - accuracy: 0.9868 - val_loss: 0.1846 - val_precision: 0.9789 - val_recall: 0.9586 - val_accuracy: 0.9741\n",
      "Epoch 5/7\n",
      "326/326 [==============================] - 6s 18ms/step - loss: 0.0639 - precision: 0.9930 - recall: 0.9818 - accuracy: 0.9891 - val_loss: 0.1454 - val_precision: 0.9808 - val_recall: 0.9672 - val_accuracy: 0.9767\n",
      "Epoch 6/7\n",
      "326/326 [==============================] - 6s 17ms/step - loss: 0.0478 - precision: 0.9942 - recall: 0.9870 - accuracy: 0.9920 - val_loss: 0.1247 - val_precision: 0.9783 - val_recall: 0.9724 - val_accuracy: 0.9758\n",
      "Epoch 7/7\n",
      "326/326 [==============================] - 5s 16ms/step - loss: 0.0299 - precision: 0.9965 - recall: 0.9915 - accuracy: 0.9942 - val_loss: 0.1246 - val_precision: 0.9818 - val_recall: 0.9758 - val_accuracy: 0.9802\n"
     ]
    }
   ],
   "source": [
    "lstm_history = lstm_model.fit(X, y_train_encoded, batch_size = BATCH_SIZE, epochs = EPOCHS, verbose = 1, validation_split = VAL_SPLIT)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plot learning curves"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title('Learning Curves')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Cross Entropy')\n",
    "plt.plot(lstm_history.history['loss'], label='train')\n",
    "plt.plot(lstm_history.history['val_loss'], label='val')\n",
    "plt.legend()\n",
    "plt.savefig(\"learning_curve.jpg\", bbox_inches='tight', dpi = 300)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predict on test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "91/91 [==============================] - 0s 3ms/step - loss: 0.1192 - precision: 0.9884 - recall: 0.9724 - accuracy: 0.9821\n"
     ]
    }
   ],
   "source": [
    "#Evaluate the model performance on test data\n",
    "lstm_score = lstm_model.evaluate(X_test, y_test_encoded, batch_size = BATCH_SIZE, verbose = 1)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Classification metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                        precision    recall  f1-score   support\n",
      "\n",
      "       add_to_playlist       1.00      0.98      0.99       428\n",
      "       book_restaurant       0.99      0.99      0.99       394\n",
      "           get_weather       0.99      1.00      0.99       434\n",
      "            play_music       0.97      1.00      0.98       410\n",
      "             rate_book       0.99      0.99      0.99       403\n",
      "  search_creative_work       0.96      0.95      0.95       411\n",
      "search_screening_event       0.98      0.97      0.98       417\n",
      "\n",
      "              accuracy                           0.98      2897\n",
      "             macro avg       0.98      0.98      0.98      2897\n",
      "          weighted avg       0.98      0.98      0.98      2897\n",
      "\n"
     ]
    }
   ],
   "source": [
    "y_pred_labels_lstm = [encoder.classes_[x] for x in np.argmax(lstm_model.predict(X_test), axis=-1)]\n",
    "print(metrics.classification_report(y_test, y_pred_labels_lstm))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "play_music\n"
     ]
    }
   ],
   "source": [
    "sentence = \"Play music on YouTube Music\"\n",
    "input_seq = tokenizer.texts_to_sequences([sentence])\n",
    "input_features = pad_sequences(input_seq, maxlen = MAX_SEQ_LEN, padding = 'post')\n",
    "probs = lstm_model.predict(input_features)\n",
    "predicted_y = probs.argmax(axis=-1)\n",
    "print(encoder.classes_[predicted_y][0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}