job_skills_prediction.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "b862911c",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "df = pd.read_csv('outputs/np_train_skills_no_commas.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "643a4ec6",
   "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>Unnamed: 0</th>\n",
       "      <th>Target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>19769.000000</td>\n",
       "      <td>13004.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>9884.000000</td>\n",
       "      <td>0.304445</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>5706.963072</td>\n",
       "      <td>0.460190</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>4942.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>9884.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>14826.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>19768.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         Unnamed: 0        Target\n",
       "count  19769.000000  13004.000000\n",
       "mean    9884.000000      0.304445\n",
       "std     5706.963072      0.460190\n",
       "min        0.000000      0.000000\n",
       "25%     4942.000000      0.000000\n",
       "50%     9884.000000      0.000000\n",
       "75%    14826.000000      1.000000\n",
       "max    19768.000000      1.000000"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "447a1ed9",
   "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>Unnamed: 0</th>\n",
       "      <th>Phrase</th>\n",
       "      <th>Target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>significant scope impact</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>monitor compliance</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>self-service reporting solutions</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>experience</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>strong attention detail commitment</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0                              Phrase  Target\n",
       "0           0            significant scope impact     0.0\n",
       "1           1                  monitor compliance     0.0\n",
       "2           2    self-service reporting solutions     0.0\n",
       "3           3                          experience     0.0\n",
       "4           4  strong attention detail commitment     1.0"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "02036562",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Unnamed: 0       0\n",
       "Phrase           1\n",
       "Target        6765\n",
       "dtype: int64"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "ce1e150e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Unnamed: 0    0\n",
       "Phrase        0\n",
       "Target        0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#drop french phrases and extra data\n",
    "df = df.dropna()\n",
    "df.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "84d8f59a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "13004"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "c8d31707",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Phrase Index with most characters is:  6593\n"
     ]
    }
   ],
   "source": [
    "print('Phrase Index with most characters is: ', df['Phrase'].str.len().idxmax())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "4c24718e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'www.canada.ca/en/treasury-board-secretariat/services/professional-development/key-leadership-competency-profile/examples-effective-ineffective-behaviours.html create vision strategy'"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['Phrase'][6593]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "b4d1e8d4",
   "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>Unnamed: 0</th>\n",
       "      <th>Phrase</th>\n",
       "      <th>Target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4397</th>\n",
       "      <td>4397</td>\n",
       "      <td>credit risk management modeling</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18061</th>\n",
       "      <td>18061</td>\n",
       "      <td>adopt</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3872</th>\n",
       "      <td>3872</td>\n",
       "      <td>contact member</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11332</th>\n",
       "      <td>11332</td>\n",
       "      <td>various scope</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18005</th>\n",
       "      <td>18005</td>\n",
       "      <td>handled</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1321</th>\n",
       "      <td>1321</td>\n",
       "      <td>primary responsibilities</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>802</th>\n",
       "      <td>802</td>\n",
       "      <td>related technical field</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4508</th>\n",
       "      <td>4508</td>\n",
       "      <td>considerable latitude exercise</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9352</th>\n",
       "      <td>9352</td>\n",
       "      <td>baselining</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8507</th>\n",
       "      <td>8507</td>\n",
       "      <td>high quality data</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>947</th>\n",
       "      <td>947</td>\n",
       "      <td>search data engineer</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2973</th>\n",
       "      <td>2973</td>\n",
       "      <td>vision growth</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18682</th>\n",
       "      <td>18682</td>\n",
       "      <td>patient scheduling</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3671</th>\n",
       "      <td>3671</td>\n",
       "      <td>significant impact health</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18282</th>\n",
       "      <td>18282</td>\n",
       "      <td>transform supply chain</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1258</th>\n",
       "      <td>1258</td>\n",
       "      <td>gender identity expression</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8857</th>\n",
       "      <td>8857</td>\n",
       "      <td>reliance expensive vendor products</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10556</th>\n",
       "      <td>10556</td>\n",
       "      <td>site characteristics</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6754</th>\n",
       "      <td>6754</td>\n",
       "      <td>predictive maintenance</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2095</th>\n",
       "      <td>2095</td>\n",
       "      <td>iteration planning</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19736</th>\n",
       "      <td>19736</td>\n",
       "      <td>cutting-edge</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1783</th>\n",
       "      <td>1783</td>\n",
       "      <td>author</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18788</th>\n",
       "      <td>18788</td>\n",
       "      <td>translation</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18230</th>\n",
       "      <td>18230</td>\n",
       "      <td>related computer vision</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1840</th>\n",
       "      <td>1840</td>\n",
       "      <td>production systems data scientist</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6049</th>\n",
       "      <td>6049</td>\n",
       "      <td>business-logic guardrails</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4671</th>\n",
       "      <td>4671</td>\n",
       "      <td>share passion –</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5336</th>\n",
       "      <td>5336</td>\n",
       "      <td>inference results</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6842</th>\n",
       "      <td>6842</td>\n",
       "      <td>applications development</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8621</th>\n",
       "      <td>8621</td>\n",
       "      <td>managing</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3057</th>\n",
       "      <td>3057</td>\n",
       "      <td>informatics</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18666</th>\n",
       "      <td>18666</td>\n",
       "      <td>data blending</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11512</th>\n",
       "      <td>11512</td>\n",
       "      <td>knowledge statistical analysis techniques</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3604</th>\n",
       "      <td>3604</td>\n",
       "      <td>high impact product features</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1404</th>\n",
       "      <td>1404</td>\n",
       "      <td>complex joins</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>131</th>\n",
       "      <td>131</td>\n",
       "      <td>medical condition</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6265</th>\n",
       "      <td>6265</td>\n",
       "      <td>ask access</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19295</th>\n",
       "      <td>19295</td>\n",
       "      <td>knowledge python</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>620</th>\n",
       "      <td>620</td>\n",
       "      <td>situations</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10269</th>\n",
       "      <td>10269</td>\n",
       "      <td>jam city equal opportunity employer</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18872</th>\n",
       "      <td>18872</td>\n",
       "      <td>deployment</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5898</th>\n",
       "      <td>5898</td>\n",
       "      <td>validation implementation</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5253</th>\n",
       "      <td>5253</td>\n",
       "      <td>acts</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18479</th>\n",
       "      <td>18479</td>\n",
       "      <td>pods</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18617</th>\n",
       "      <td>18617</td>\n",
       "      <td>etl</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1909</th>\n",
       "      <td>1909</td>\n",
       "      <td>model validation model implementation research...</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9562</th>\n",
       "      <td>9562</td>\n",
       "      <td>seattle</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2982</th>\n",
       "      <td>2982</td>\n",
       "      <td>code quality proficient aws products</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9401</th>\n",
       "      <td>9401</td>\n",
       "      <td>etl infrastructure new products</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11839</th>\n",
       "      <td>11839</td>\n",
       "      <td>selection delightful</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Unnamed: 0                                             Phrase  Target\n",
       "4397         4397                    credit risk management modeling     1.0\n",
       "18061       18061                                              adopt     0.0\n",
       "3872         3872                                     contact member     0.0\n",
       "11332       11332                                      various scope     0.0\n",
       "18005       18005                                            handled     0.0\n",
       "1321         1321                           primary responsibilities     0.0\n",
       "802           802                            related technical field     1.0\n",
       "4508         4508                     considerable latitude exercise     0.0\n",
       "9352         9352                                         baselining     1.0\n",
       "8507         8507                                  high quality data     0.0\n",
       "947           947                               search data engineer     1.0\n",
       "2973         2973                                      vision growth     0.0\n",
       "18682       18682                                patient scheduling      1.0\n",
       "3671         3671                          significant impact health     0.0\n",
       "18282       18282                             transform supply chain     0.0\n",
       "1258         1258                         gender identity expression     0.0\n",
       "8857         8857                 reliance expensive vendor products     0.0\n",
       "10556       10556                               site characteristics     0.0\n",
       "6754         6754                             predictive maintenance     1.0\n",
       "2095         2095                                 iteration planning     1.0\n",
       "19736       19736                                      cutting-edge      0.0\n",
       "1783         1783                                             author     0.0\n",
       "18788       18788                                        translation     0.0\n",
       "18230       18230                            related computer vision     0.0\n",
       "1840         1840                  production systems data scientist     0.0\n",
       "6049         6049                          business-logic guardrails     0.0\n",
       "4671         4671                                    share passion –     0.0\n",
       "5336         5336                                  inference results     0.0\n",
       "6842         6842                           applications development     0.0\n",
       "8621         8621                                           managing     0.0\n",
       "3057         3057                                        informatics     1.0\n",
       "18666       18666                                     data blending      1.0\n",
       "11512       11512          knowledge statistical analysis techniques     1.0\n",
       "3604         3604                       high impact product features     0.0\n",
       "1404         1404                                      complex joins     1.0\n",
       "131           131                                  medical condition     0.0\n",
       "6265         6265                                         ask access     0.0\n",
       "19295       19295                                  knowledge python      1.0\n",
       "620           620                                         situations     0.0\n",
       "10269       10269                jam city equal opportunity employer     0.0\n",
       "18872       18872                                        deployment      1.0\n",
       "5898         5898                          validation implementation     0.0\n",
       "5253         5253                                               acts     0.0\n",
       "18479       18479                                              pods      0.0\n",
       "18617       18617                                               etl      1.0\n",
       "1909         1909  model validation model implementation research...     1.0\n",
       "9562         9562                                            seattle     0.0\n",
       "2982         2982               code quality proficient aws products     1.0\n",
       "9401         9401                    etl infrastructure new products     1.0\n",
       "11839       11839                               selection delightful     0.0"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.sample(50, random_state = 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "43f6a9ca",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:xlabel='Target', ylabel='Counts'>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.figure(figsize=(10,7))\n",
    "\n",
    "plt.xlabel(\"Target\")\n",
    "plt.ylabel(\"Counts\")\n",
    "\n",
    "df[\"Target\"].value_counts().plot(kind='bar')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "e337e29a",
   "metadata": {},
   "outputs": [],
   "source": [
    "from collections import Counter\n",
    "\n",
    "#count unique words\n",
    "def counter_word(text):\n",
    "    count = Counter()\n",
    "    for i in text.values:\n",
    "        for word in i.split():\n",
    "            count[word] += 1\n",
    "    return count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "df6c5032",
   "metadata": {},
   "outputs": [],
   "source": [
    "text = df['Phrase']\n",
    "counter = counter_word(text)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "29308340",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5772"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(counter)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "835d0058",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Counter({'significant': 19,\n",
       "         'scope': 13,\n",
       "         'impact': 53,\n",
       "         'monitor': 9,\n",
       "         'compliance': 16,\n",
       "         'self-service': 5,\n",
       "         'reporting': 29,\n",
       "         'solutions': 151,\n",
       "         'experience': 411,\n",
       "         'strong': 127,\n",
       "         'attention': 16,\n",
       "         'detail': 19,\n",
       "         'commitment': 14,\n",
       "         'recombination': 1,\n",
       "         'techniques': 54,\n",
       "         'test': 19,\n",
       "         'cases': 14,\n",
       "         'aim': 3,\n",
       "         'work': 276,\n",
       "         'job': 76,\n",
       "         'requirements': 80,\n",
       "         'day': 21,\n",
       "         'individuals': 25,\n",
       "         'delivery': 42,\n",
       "         'network': 22,\n",
       "         '—': 7,\n",
       "         'accenture': 6,\n",
       "         'quantitative': 41,\n",
       "         'analysis': 145,\n",
       "         'tests': 10,\n",
       "         'problems': 61,\n",
       "         'everyone': 16,\n",
       "         'expert': 21,\n",
       "         'everything': 12,\n",
       "         'long': 4,\n",
       "         'term': 10,\n",
       "         'equity': 18,\n",
       "         'lightspeeders': 1,\n",
       "         'implantable': 1,\n",
       "         'medical': 16,\n",
       "         'device': 5,\n",
       "         'able': 33,\n",
       "         'generate': 8,\n",
       "         'share': 26,\n",
       "         'perl': 2,\n",
       "         'resources': 15,\n",
       "         '’': 127,\n",
       "         'data': 1119,\n",
       "         'science': 175,\n",
       "         'executive-facing': 1,\n",
       "         'dashboards': 18,\n",
       "         'track': 11,\n",
       "         'progress': 10,\n",
       "         'business': 344,\n",
       "         'analyst': 23,\n",
       "         '–': 40,\n",
       "         'description': 21,\n",
       "         'complex': 71,\n",
       "         'problems/issues': 2,\n",
       "         'scaling': 4,\n",
       "         'team': 306,\n",
       "         'passionate': 22,\n",
       "         'algorithms': 36,\n",
       "         'develop': 17,\n",
       "         'models': 99,\n",
       "         'state': 5,\n",
       "         'art': 4,\n",
       "         'ability': 126,\n",
       "         'manage': 33,\n",
       "         'hadoop': 17,\n",
       "         'spark': 29,\n",
       "         'savings': 9,\n",
       "         'position': 46,\n",
       "         'report': 7,\n",
       "         'cto': 3,\n",
       "         'evaluate': 5,\n",
       "         'technical': 124,\n",
       "         'knowledge': 137,\n",
       "         'pertains': 1,\n",
       "         'https': 2,\n",
       "         'increase': 5,\n",
       "         'profitability': 5,\n",
       "         'identification': 6,\n",
       "         'characterization': 1,\n",
       "         'classic': 2,\n",
       "         'movies': 2,\n",
       "         'others': 14,\n",
       "         'fellow': 3,\n",
       "         'scientists': 32,\n",
       "         'engineers': 28,\n",
       "         'excellent': 42,\n",
       "         'proficiency': 30,\n",
       "         'computer': 73,\n",
       "         'productivity': 5,\n",
       "         'software': 118,\n",
       "         'basis': 15,\n",
       "         'environment': 125,\n",
       "         'benefits': 42,\n",
       "         'online/offline': 1,\n",
       "         'customer': 81,\n",
       "         'journey': 8,\n",
       "         'language': 29,\n",
       "         'skills': 177,\n",
       "         'corporate': 10,\n",
       "         'discount': 2,\n",
       "         'gym': 3,\n",
       "         'membership': 2,\n",
       "         'family': 16,\n",
       "         'downtown': 8,\n",
       "         'halifax': 2,\n",
       "         'reports': 33,\n",
       "         'etc': 6,\n",
       "         'unable': 4,\n",
       "         'apply': 13,\n",
       "         'online': 18,\n",
       "         'require': 5,\n",
       "         'accommodations': 8,\n",
       "         'accessiblecareers': 1,\n",
       "         'power': 29,\n",
       "         'platform': 59,\n",
       "         'governance': 21,\n",
       "         'support': 107,\n",
       "         'value': 56,\n",
       "         'sql': 73,\n",
       "         'engineering': 127,\n",
       "         'development': 221,\n",
       "         'automate': 5,\n",
       "         'ingestions': 1,\n",
       "         'transformations': 5,\n",
       "         'concepts': 20,\n",
       "         'cloud': 83,\n",
       "         'solution': 47,\n",
       "         'architect': 6,\n",
       "         'drive': 48,\n",
       "         'traditional': 7,\n",
       "         'affords': 1,\n",
       "         'senior': 49,\n",
       "         'management': 168,\n",
       "         'targets': 4,\n",
       "         'resource': 10,\n",
       "         'relevant': 32,\n",
       "         'ui': 3,\n",
       "         'technologies': 81,\n",
       "         'develops': 2,\n",
       "         'collaborative': 22,\n",
       "         'agile': 34,\n",
       "         'teams': 92,\n",
       "         'welcome': 11,\n",
       "         'applications': 57,\n",
       "         'benchmarking': 1,\n",
       "         'product': 151,\n",
       "         'features': 14,\n",
       "         'large': 41,\n",
       "         'matrix': 1,\n",
       "         'organizations': 12,\n",
       "         'cassandra': 5,\n",
       "         'related': 40,\n",
       "         'field': 47,\n",
       "         'math': 16,\n",
       "         'hr': 5,\n",
       "         'informs': 1,\n",
       "         'demonstrable': 4,\n",
       "         'research': 103,\n",
       "         'innovative': 34,\n",
       "         'full-time': 18,\n",
       "         'employees': 23,\n",
       "         'protect': 1,\n",
       "         'homes': 1,\n",
       "         'application': 47,\n",
       "         'process': 78,\n",
       "         'filtering': 1,\n",
       "         'loyalty': 5,\n",
       "         'active': 11,\n",
       "         'projects': 79,\n",
       "         'egos': 1,\n",
       "         'break': 3,\n",
       "         'tableau': 18,\n",
       "         'bi': 24,\n",
       "         'inclusive': 22,\n",
       "         'workplace': 20,\n",
       "         'culture': 47,\n",
       "         'high': 54,\n",
       "         'initiatives': 22,\n",
       "         'interest': 27,\n",
       "         'periscope': 1,\n",
       "         'career': 32,\n",
       "         'future': 35,\n",
       "         'revenue': 10,\n",
       "         'mills': 1,\n",
       "         'global': 49,\n",
       "         'industry': 74,\n",
       "         'leader': 22,\n",
       "         'innovator': 1,\n",
       "         'opportunity': 92,\n",
       "         'tremendous': 4,\n",
       "         'responsibility': 8,\n",
       "         'customers': 58,\n",
       "         'general': 10,\n",
       "         'office': 35,\n",
       "         'equipment': 6,\n",
       "         'note': 11,\n",
       "         'accelerate': 7,\n",
       "         'pace': 1,\n",
       "         'machine': 123,\n",
       "         'unstructured': 8,\n",
       "         'advice': 13,\n",
       "         'help': 45,\n",
       "         'clients': 54,\n",
       "         'way': 29,\n",
       "         'years': 22,\n",
       "         'multi': 1,\n",
       "         '-functional': 1,\n",
       "         'identify': 17,\n",
       "         'e': 3,\n",
       "         'master': 14,\n",
       "         'bd/ml': 1,\n",
       "         'workshops': 2,\n",
       "         'meaningful': 14,\n",
       "         'conclusions': 11,\n",
       "         'tasks': 17,\n",
       "         'guidance': 16,\n",
       "         'post-launch': 1,\n",
       "         'improve': 9,\n",
       "         'powerpoint': 5,\n",
       "         'advanced': 37,\n",
       "         'analytics': 162,\n",
       "         'care': 31,\n",
       "         'problem': 26,\n",
       "         'hand': 3,\n",
       "         '5-7': 1,\n",
       "         'watch': 3,\n",
       "         'talk': 4,\n",
       "         'dynamic': 16,\n",
       "         'social': 17,\n",
       "         'events': 19,\n",
       "         'capital': 9,\n",
       "         'markets': 12,\n",
       "         'operations': 37,\n",
       "         'similar': 17,\n",
       "         'tools': 111,\n",
       "         'deep': 50,\n",
       "         'exposure': 11,\n",
       "         'ai': 73,\n",
       "         'statistical': 68,\n",
       "         'matching': 2,\n",
       "         'condition': 2,\n",
       "         'analytical': 74,\n",
       "         'methodologies': 23,\n",
       "         'objectives': 13,\n",
       "         'serious': 3,\n",
       "         'brain': 2,\n",
       "         'create': 28,\n",
       "         'insights': 65,\n",
       "         'potential': 17,\n",
       "         'different': 32,\n",
       "         'sensors': 2,\n",
       "         'pipelines': 36,\n",
       "         'high-volume': 1,\n",
       "         'ideas': 22,\n",
       "         'collaboration': 32,\n",
       "         'members': 28,\n",
       "         'sure': 6,\n",
       "         'mark': 3,\n",
       "         'cmb': 1,\n",
       "         'bilingualism': 3,\n",
       "         'asset': 42,\n",
       "         'altaml': 1,\n",
       "         'applied': 8,\n",
       "         'lab': 12,\n",
       "         'inform': 12,\n",
       "         'media': 24,\n",
       "         'strategy': 45,\n",
       "         'actionable': 12,\n",
       "         'marital': 2,\n",
       "         'domestic/civil': 1,\n",
       "         'partnership': 5,\n",
       "         'status': 34,\n",
       "         'dynamodb': 1,\n",
       "         'foundations': 3,\n",
       "         'diversity': 39,\n",
       "         'inclusion': 19,\n",
       "         'sex': 4,\n",
       "         'appropriate': 22,\n",
       "         'equal': 38,\n",
       "         'employer': 35,\n",
       "         'information': 70,\n",
       "         'visit': 15,\n",
       "         'legal': 2,\n",
       "         'professional': 28,\n",
       "         'analyse': 4,\n",
       "         'model': 53,\n",
       "         'performance': 62,\n",
       "         'accuracy': 18,\n",
       "         'production': 34,\n",
       "         'robotics': 3,\n",
       "         'cpp': 6,\n",
       "         'fund': 4,\n",
       "         'comprehension': 1,\n",
       "         'architecture': 49,\n",
       "         'sea': 1,\n",
       "         'build': 30,\n",
       "         'exploratory': 3,\n",
       "         'validation': 16,\n",
       "         'analysts': 19,\n",
       "         'religious': 1,\n",
       "         'beliefs': 2,\n",
       "         'banking': 18,\n",
       "         'ground': 5,\n",
       "         'boxes': 1,\n",
       "         'candidates': 19,\n",
       "         'canada': 45,\n",
       "         'introduction': 2,\n",
       "         'ibm': 10,\n",
       "         'leadership': 50,\n",
       "         'principles': 16,\n",
       "         'remote': 16,\n",
       "         'locations': 7,\n",
       "         'ecc': 1,\n",
       "         's4': 1,\n",
       "         'life-changing': 2,\n",
       "         'structural': 2,\n",
       "         'equation': 1,\n",
       "         'non-technical': 10,\n",
       "         'audiences': 11,\n",
       "         'understand': 28,\n",
       "         'interacting': 1,\n",
       "         'results': 38,\n",
       "         'ideal': 16,\n",
       "         'candidate': 34,\n",
       "         'systems': 78,\n",
       "         'personal': 14,\n",
       "         'branding': 2,\n",
       "         '“': 14,\n",
       "         'clustering': 1,\n",
       "         'advance': 10,\n",
       "         'oral': 4,\n",
       "         'communication': 46,\n",
       "         'great': 30,\n",
       "         'experiences': 35,\n",
       "         'functions': 12,\n",
       "         'johnson': 2,\n",
       "         'chemistry': 11,\n",
       "         'products': 76,\n",
       "         'execute': 12,\n",
       "         'provide': 19,\n",
       "         'age': 8,\n",
       "         'people': 70,\n",
       "         'races': 2,\n",
       "         'familiarity': 17,\n",
       "         'duties': 10,\n",
       "         'database': 31,\n",
       "         'optimization': 24,\n",
       "         'stream': 4,\n",
       "         'join': 39,\n",
       "         'automotive': 5,\n",
       "         'department': 13,\n",
       "         'high-dimensional': 2,\n",
       "         'salary': 18,\n",
       "         'reflective': 1,\n",
       "         'place': 15,\n",
       "         'safe': 4,\n",
       "         'event': 6,\n",
       "         'falls': 2,\n",
       "         'rules': 10,\n",
       "         'objects': 5,\n",
       "         'risks': 7,\n",
       "         'petabytes': 1,\n",
       "         'modern': 26,\n",
       "         'r': 30,\n",
       "         'packages': 12,\n",
       "         'e.g': 3,\n",
       "         'intelligence': 43,\n",
       "         'new': 106,\n",
       "         'credit': 16,\n",
       "         'person': 9,\n",
       "         'thrives': 3,\n",
       "         'identifies': 4,\n",
       "         'developers': 22,\n",
       "         'antuit': 1,\n",
       "         'examples': 1,\n",
       "         'dbt': 2,\n",
       "         'high-quality': 10,\n",
       "         'users': 18,\n",
       "         'qualifications': 37,\n",
       "         'communications': 10,\n",
       "         'enterprise': 33,\n",
       "         'diverse': 45,\n",
       "         'cross': 4,\n",
       "         'functional': 13,\n",
       "         'multidisciplinary': 5,\n",
       "         'health': 58,\n",
       "         'integration': 41,\n",
       "         'platforms': 49,\n",
       "         'ia': 2,\n",
       "         'geomatics': 3,\n",
       "         'implement': 34,\n",
       "         'technology': 97,\n",
       "         'national': 9,\n",
       "         'origin': 7,\n",
       "         'ancestry': 4,\n",
       "         'order': 12,\n",
       "         'reflect': 1,\n",
       "         'nation': 1,\n",
       "         'docker': 7,\n",
       "         'consensus': 3,\n",
       "         'solution/projects': 1,\n",
       "         'spark/impala/hive': 1,\n",
       "         'expertise': 50,\n",
       "         'signal': 3,\n",
       "         'processing': 35,\n",
       "         'multiple': 49,\n",
       "         'sources': 37,\n",
       "         'decision-making': 8,\n",
       "         'optimize': 10,\n",
       "         'stantec': 7,\n",
       "         'bank': 15,\n",
       "         'degree': 69,\n",
       "         'equivalency': 2,\n",
       "         'voice': 8,\n",
       "         'recognition': 10,\n",
       "         'scala': 11,\n",
       "         'fun-loving': 2,\n",
       "         'quality': 63,\n",
       "         'factor': 3,\n",
       "         'define': 14,\n",
       "         'articulate': 1,\n",
       "         'stakeholders': 59,\n",
       "         'backend': 3,\n",
       "         'assessments': 5,\n",
       "         'programs': 22,\n",
       "         'analyze': 21,\n",
       "         'needs': 51,\n",
       "         'solve': 15,\n",
       "         'standard': 10,\n",
       "         'operating': 5,\n",
       "         'procedures': 15,\n",
       "         'control': 15,\n",
       "         'financial': 37,\n",
       "         'designers': 6,\n",
       "         'server': 15,\n",
       "         'recruitment': 18,\n",
       "         'subject': 16,\n",
       "         'reference': 8,\n",
       "         'criminal': 4,\n",
       "         'background': 15,\n",
       "         'checks': 4,\n",
       "         'bigquery': 6,\n",
       "         'and/or': 45,\n",
       "         'kinesis': 6,\n",
       "         'matter': 13,\n",
       "         'competitive': 15,\n",
       "         'comprehensive': 16,\n",
       "         'compensation': 15,\n",
       "         'package': 12,\n",
       "         'challenges': 23,\n",
       "         'toronto': 18,\n",
       "         'location': 20,\n",
       "         'td': 6,\n",
       "         'priorities': 11,\n",
       "         'infrastructure': 46,\n",
       "         'available': 22,\n",
       "         'hiring': 4,\n",
       "         'manager': 27,\n",
       "         'perform': 7,\n",
       "         'forces': 1,\n",
       "         'entrepreneurial': 5,\n",
       "         'mind-set': 1,\n",
       "         'enhance': 16,\n",
       "         'self-motivated': 7,\n",
       "         'individual': 20,\n",
       "         'mainstream': 2,\n",
       "         'learning/deep': 1,\n",
       "         'gans': 2,\n",
       "         'accommodation': 23,\n",
       "         'predictive': 25,\n",
       "         'responsibilities': 27,\n",
       "         'caffe': 1,\n",
       "         'windows': 2,\n",
       "         'scientist': 52,\n",
       "         'smart': 14,\n",
       "         'aggregations': 1,\n",
       "         'discussions': 7,\n",
       "         'peers': 8,\n",
       "         'visualization': 33,\n",
       "         'qualified': 7,\n",
       "         'specifications': 3,\n",
       "         'pro': 1,\n",
       "         'present': 19,\n",
       "         'findings': 16,\n",
       "         'want': 7,\n",
       "         'rocket': 1,\n",
       "         'ship': 4,\n",
       "         'protection': 10,\n",
       "         'building': 53,\n",
       "         'libraries': 13,\n",
       "         'azure': 49,\n",
       "         'cosmos': 2,\n",
       "         'db': 3,\n",
       "         'omers': 2,\n",
       "         'equivalent': 18,\n",
       "         'req': 3,\n",
       "         'mysql': 7,\n",
       "         'big': 47,\n",
       "         'performs': 1,\n",
       "         'conversation': 4,\n",
       "         'offerings': 8,\n",
       "         'businesses': 15,\n",
       "         'pyspark': 1,\n",
       "         'employment': 54,\n",
       "         'program': 40,\n",
       "         'stock': 8,\n",
       "         'mining': 26,\n",
       "         'analytic': 17,\n",
       "         'thought': 8,\n",
       "         'impacts': 5,\n",
       "         'google': 16,\n",
       "         'grow': 9,\n",
       "         'area': 12,\n",
       "         'various': 54,\n",
       "         'types': 15,\n",
       "         'old': 2,\n",
       "         '”': 15,\n",
       "         'levels': 14,\n",
       "         'risk': 30,\n",
       "         'ci/cd': 7,\n",
       "         'devops': 16,\n",
       "         'standards': 15,\n",
       "         'strategic': 29,\n",
       "         'thinker': 10,\n",
       "         'world': 55,\n",
       "         'run': 3,\n",
       "         'combination': 8,\n",
       "         'education': 37,\n",
       "         'consideration': 3,\n",
       "         'decision': 30,\n",
       "         'portfolio': 8,\n",
       "         'mathematics': 16,\n",
       "         'verbal': 9,\n",
       "         'presentation': 15,\n",
       "         'external': 24,\n",
       "         'partnerships': 2,\n",
       "         'strict': 2,\n",
       "         'deadlines': 7,\n",
       "         'actuarial': 4,\n",
       "         '/': 32,\n",
       "         'company': 87,\n",
       "         'powercenter': 1,\n",
       "         'mathematical': 6,\n",
       "         'market': 39,\n",
       "         'learning': 98,\n",
       "         'talented': 4,\n",
       "         'employee': 34,\n",
       "         'polygraph': 1,\n",
       "         'act': 14,\n",
       "         'chance': 8,\n",
       "         'gain': 9,\n",
       "         'action': 17,\n",
       "         'rail': 1,\n",
       "         'transportation': 9,\n",
       "         'conduct': 14,\n",
       "         'system': 34,\n",
       "         '4-6': 1,\n",
       "         'evidence-based': 1,\n",
       "         'decisions': 30,\n",
       "         'investment': 16,\n",
       "         'studies': 11,\n",
       "         'digital': 58,\n",
       "         'transformation': 17,\n",
       "         'economics': 6,\n",
       "         'energy': 23,\n",
       "         'demonstrated': 8,\n",
       "         'religion': 3,\n",
       "         'necessary': 7,\n",
       "         'extract': 8,\n",
       "         'discover': 6,\n",
       "         'patterns': 10,\n",
       "         'economic': 6,\n",
       "         'aspects': 14,\n",
       "         'society': 6,\n",
       "         'hire': 6,\n",
       "         'reliability': 7,\n",
       "         'microsoft': 22,\n",
       "         'engineer': 44,\n",
       "         'craft': 3,\n",
       "         'training': 29,\n",
       "         'awareness': 4,\n",
       "         'topic': 1,\n",
       "         'cybersecurity': 6,\n",
       "         'plan': 37,\n",
       "         'brilliant': 3,\n",
       "         'front': 3,\n",
       "         'back-office': 1,\n",
       "         'following': 4,\n",
       "         'iterative': 5,\n",
       "         'approach': 19,\n",
       "         'format': 7,\n",
       "         'role': 72,\n",
       "         'toronto/vancouver/calgary/montreal': 1,\n",
       "         'lives': 8,\n",
       "         'distributed': 4,\n",
       "         'semi-supervised': 1,\n",
       "         'sets': 19,\n",
       "         'resume': 8,\n",
       "         'interview': 10,\n",
       "         'time': 81,\n",
       "         'community': 24,\n",
       "         'service': 24,\n",
       "         'project': 52,\n",
       "         'efficiency': 18,\n",
       "         'effectiveness': 5,\n",
       "         'services': 73,\n",
       "         '-if': 1,\n",
       "         'fcc': 1,\n",
       "         'scalable': 15,\n",
       "         'fuel': 4,\n",
       "         'history': 5,\n",
       "         'statistics': 26,\n",
       "         'aws': 38,\n",
       "         'vast': 3,\n",
       "         'amounts': 4,\n",
       "         'varieties': 1,\n",
       "         'initiative': 6,\n",
       "         'footprint': 3,\n",
       "         'you.s': 4,\n",
       "         'python': 61,\n",
       "         'sports': 13,\n",
       "         'last': 1,\n",
       "         'millisecond': 1,\n",
       "         'something': 7,\n",
       "         'tracking': 2,\n",
       "         'ticket': 3,\n",
       "         'sales': 17,\n",
       "         'practices': 26,\n",
       "         'message': 2,\n",
       "         'queuing': 1,\n",
       "         'telus': 3,\n",
       "         'marketing': 39,\n",
       "         'publishers': 3,\n",
       "         'transform': 14,\n",
       "         'prepare': 5,\n",
       "         'inspirational': 1,\n",
       "         'energetic': 3,\n",
       "         'hands-on': 19,\n",
       "         'ad': 11,\n",
       "         'servers': 2,\n",
       "         'massage': 2,\n",
       "         'chairs': 2,\n",
       "         'skill': 5,\n",
       "         'summary': 8,\n",
       "         'applicants': 27,\n",
       "         'life': 42,\n",
       "         'bachelor': 20,\n",
       "         'key': 49,\n",
       "         'trends': 11,\n",
       "         'minds': 4,\n",
       "         'fields': 12,\n",
       "         'brings': 5,\n",
       "         'pregnancy': 3,\n",
       "         'formation': 4,\n",
       "         'ingestion': 8,\n",
       "         'partner': 31,\n",
       "         'open': 22,\n",
       "         'source': 13,\n",
       "         'box': 2,\n",
       "         'fixed': 3,\n",
       "         'income': 3,\n",
       "         'publish': 2,\n",
       "         'processed': 1,\n",
       "         'ads': 6,\n",
       "         'rely': 1,\n",
       "         'solving': 3,\n",
       "         '//www.amazon.jobs/en/disability/ontario': 1,\n",
       "         'approaches': 19,\n",
       "         'advantage': 6,\n",
       "         'asynchronous': 1,\n",
       "         'calls': 2,\n",
       "         'communities': 19,\n",
       "         'experienced': 8,\n",
       "         'innovators': 5,\n",
       "         'exigences': 1,\n",
       "         'primary': 17,\n",
       "         'arrangements': 2,\n",
       "         'applicable': 10,\n",
       "         'promote': 6,\n",
       "         'fair': 3,\n",
       "         'montreal': 12,\n",
       "         'persons': 11,\n",
       "         'disabilities': 17,\n",
       "         'managers': 17,\n",
       "         'executives': 9,\n",
       "         'importance': 3,\n",
       "         'staff': 13,\n",
       "         'part': 45,\n",
       "         'games': 6,\n",
       "         'committed': 6,\n",
       "         'innovation': 40,\n",
       "         'ambitions': 2,\n",
       "         'world-leading': 1,\n",
       "         'mandatory': 7,\n",
       "         'linux': 7,\n",
       "         'site': 8,\n",
       "         'autonomous': 3,\n",
       "         'vehicle': 4,\n",
       "         'pm': 3,\n",
       "         'ecosystem': 15,\n",
       "         'points': 9,\n",
       "         'git': 12,\n",
       "         'ignite': 1,\n",
       "         'emotion': 1,\n",
       "         'unit': 9,\n",
       "         'ontario': 9,\n",
       "         'city': 11,\n",
       "         'collaborate': 30,\n",
       "         'thank': 8,\n",
       "         'referral': 2,\n",
       "         'bi/dw': 5,\n",
       "         'number': 7,\n",
       "         'ztr': 4,\n",
       "         'player': 11,\n",
       "         'ownership': 18,\n",
       "         'telecom': 4,\n",
       "         'user-facing': 1,\n",
       "         'tap': 2,\n",
       "         'richness': 1,\n",
       "         'genetic': 5,\n",
       "         'medicine': 4,\n",
       "         'nationality': 3,\n",
       "         'flexibility': 17,\n",
       "         'compatibility': 1,\n",
       "         'tools/technology': 2,\n",
       "         'handful': 1,\n",
       "         'things': 10,\n",
       "         'retailers': 3,\n",
       "         '//www.geosyntec.com/careers/': 1,\n",
       "         'geosyntec': 2,\n",
       "         'offers': 10,\n",
       "         'conversational': 3,\n",
       "         'suite': 3,\n",
       "         'factory': 7,\n",
       "         'cell': 8,\n",
       "         'month': 8,\n",
       "         'contract': 20,\n",
       "         'modelling': 14,\n",
       "         'focus': 17,\n",
       "         'merits': 1,\n",
       "         'hour': 4,\n",
       "         'question': 4,\n",
       "         'major': 15,\n",
       "         'software-centric': 1,\n",
       "         'robust': 7,\n",
       "         'perspective': 7,\n",
       "         'frameworks': 21,\n",
       "         'investments': 4,\n",
       "         'investigate': 2,\n",
       "         'goal': 14,\n",
       "         'receptor': 1,\n",
       "         'real': 26,\n",
       "         'lasting': 3,\n",
       "         'change': 21,\n",
       "         'athena': 1,\n",
       "         'implementation': 38,\n",
       "         'et': 2,\n",
       "         'c': 7,\n",
       "         'sklearn/pandas': 1,\n",
       "         'modeling': 28,\n",
       "         'subsequent': 1,\n",
       "         'pools': 1,\n",
       "         'widen': 1,\n",
       "         'range': 28,\n",
       "         'humans': 1,\n",
       "         'learning/ai': 1,\n",
       "         'cell-based': 1,\n",
       "         'assays': 6,\n",
       "         'group': 44,\n",
       "         'accounting': 3,\n",
       "         'member': 16,\n",
       "         'proud': 11,\n",
       "         'advancement': 3,\n",
       "         'participates': 2,\n",
       "         'experts': 10,\n",
       "         'content': 14,\n",
       "         'mobile': 11,\n",
       "         'iot': 6,\n",
       "         'heights': 2,\n",
       "         'geotab': 3,\n",
       "         'full': 27,\n",
       "         'qualitative': 5,\n",
       "         'adaptability-': 1,\n",
       "         'embrace': 9,\n",
       "         'speed': 11,\n",
       "         'internal': 40,\n",
       "         'monthly': 2,\n",
       "         'executive': 16,\n",
       "         'overview': 10,\n",
       "         'collection': 15,\n",
       "         'passion': 22,\n",
       "         'nlp': 11,\n",
       "         'agents': 2,\n",
       "         'protective': 1,\n",
       "         'definition': 7,\n",
       "         'outlier': 1,\n",
       "         'unitibm': 2,\n",
       "         'cognitive': 6,\n",
       "         'in-person': 1,\n",
       "         'situations': 6,\n",
       "         'consult': 3,\n",
       "         'queries': 4,\n",
       "         'segmentation': 3,\n",
       "         'push': 7,\n",
       "         'boundaries': 6,\n",
       "         'fun': 10,\n",
       "         'gender': 14,\n",
       "         'exquisite': 1,\n",
       "         'details': 8,\n",
       "         'tight': 1,\n",
       "         'maritime': 2,\n",
       "         'codebase': 2,\n",
       "         'scale': 23,\n",
       "         'desire': 10,\n",
       "         'selection': 20,\n",
       "         'encourage': 6,\n",
       "         'kitchen': 2,\n",
       "         'overtime': 3,\n",
       "         'pay': 8,\n",
       "         'schedule': 15,\n",
       "         'understanding': 64,\n",
       "         'appointments': 1,\n",
       "         'pig': 3,\n",
       "         'postgressql': 1,\n",
       "         'outside': 3,\n",
       "         'single': 4,\n",
       "         'node': 1,\n",
       "         'parallel': 3,\n",
       "         'success': 33,\n",
       "         'looks': 1,\n",
       "         'explanation': 1,\n",
       "         'essential': 10,\n",
       "         'describes': 1,\n",
       "         'p': 1,\n",
       "         'regular': 9,\n",
       "         'cadence': 1,\n",
       "         'updates': 2,\n",
       "         'granularity': 1,\n",
       "         'cltv': 1,\n",
       "         'framework': 18,\n",
       "         'creation': 15,\n",
       "         'predictions': 4,\n",
       "         'matplotlib': 2,\n",
       "         'dmbok': 1,\n",
       "         'enjoyable': 1,\n",
       "         'consumer': 13,\n",
       "         'adoption': 14,\n",
       "         'methods': 31,\n",
       "         'issues': 19,\n",
       "         'momentive': 3,\n",
       "         'extensive': 13,\n",
       "         'r2601779': 1,\n",
       "         'title': 4,\n",
       "         'students': 8,\n",
       "         '[': 8,\n",
       "         'k-means': 2,\n",
       "         ']': 8,\n",
       "         'usage': 7,\n",
       "         'flexible': 22,\n",
       "         'adjust': 2,\n",
       "         'stakeholder': 5,\n",
       "         'relationships': 22,\n",
       "         'draw': 9,\n",
       "         'industrial': 7,\n",
       "         'manufacturing': 7,\n",
       "         'css': 3,\n",
       "         'javascript': 9,\n",
       "         'atlantic': 2,\n",
       "         'beyond.our': 1,\n",
       "         'delivers': 6,\n",
       "         'security': 37,\n",
       "         'level': 40,\n",
       "         'sites': 3,\n",
       "         'end': 10,\n",
       "         'user': 26,\n",
       "         'guide': 7,\n",
       "         'consistency': 5,\n",
       "         'review': 20,\n",
       "         'game': 10,\n",
       "         'design': 119,\n",
       "         'doorstep': 2,\n",
       "         'medianet': 1,\n",
       "         'warm': 2,\n",
       "         'recommendation': 10,\n",
       "         'finance': 16,\n",
       "         '%': 12,\n",
       "         'include': 4,\n",
       "         'code': 36,\n",
       "         'north': 8,\n",
       "         'american': 4,\n",
       "         'hours': 16,\n",
       "         'ccpa': 1,\n",
       "         'sense': 10,\n",
       "         'adopter': 1,\n",
       "         'charge': 1,\n",
       "         'year': 13,\n",
       "         'influencer': 1,\n",
       "         'airflow': 6,\n",
       "         'sexual': 2,\n",
       "         'orientation': 3,\n",
       "         'york': 4,\n",
       "         'traffic': 3,\n",
       "         'backbone': 2,\n",
       "         'level-headed': 1,\n",
       "         'making': 9,\n",
       "         'company-matching': 1,\n",
       "         'contributions': 9,\n",
       "         'variety': 33,\n",
       "         'tech': 14,\n",
       "         'simple': 5,\n",
       "         'overall': 13,\n",
       "         'offering': 6,\n",
       "         'core': 27,\n",
       "         'solid': 15,\n",
       "         'sustainable': 8,\n",
       "         'current': 15,\n",
       "         'free': 7,\n",
       "         'treats': 1,\n",
       "         'category': 4,\n",
       "         'impjob': 1,\n",
       "         'blog': 1,\n",
       "         'http': 2,\n",
       "         'permanent': 17,\n",
       "         'datasets': 24,\n",
       "         'custom': 8,\n",
       "         'peer-reviewed': 2,\n",
       "         'journals': 3,\n",
       "         'efficiencies': 2,\n",
       "         'team-player': 2,\n",
       "         'engage': 3,\n",
       "         'organization': 37,\n",
       "         'practice': 14,\n",
       "         'technological': 5,\n",
       "         'changes': 12,\n",
       "         'execution': 16,\n",
       "         'pushes': 1,\n",
       "         'wednesday': 1,\n",
       "         'high-availability': 1,\n",
       "         'formations': 1,\n",
       "         'sdes': 1,\n",
       "         'type': 15,\n",
       "         'actual': 1,\n",
       "         'matters': 12,\n",
       "         '·': 18,\n",
       "         'scikit-learn': 3,\n",
       "         'positions': 7,\n",
       "         'temporal': 2,\n",
       "         'integral': 6,\n",
       "         'privacy': 11,\n",
       "         'facial': 4,\n",
       "         'ar': 1,\n",
       "         'absorb': 3,\n",
       "         'dependent': 1,\n",
       "         'hypothesis': 3,\n",
       "         'generation': 9,\n",
       "         'identity': 8,\n",
       "         'recent': 3,\n",
       "         'critical': 26,\n",
       "         'piece': 3,\n",
       "         'engine': 8,\n",
       "         'admare': 3,\n",
       "         'casual': 3,\n",
       "         'dress': 4,\n",
       "         'ongoing': 5,\n",
       "         'legitimate': 1,\n",
       "         'need': 8,\n",
       "         'prototype': 7,\n",
       "         'formulations': 2,\n",
       "         'refine': 3,\n",
       "         'hypotheses': 6,\n",
       "         'answer': 2,\n",
       "         'problem-solving': 16,\n",
       "         'issue': 3,\n",
       "         'scope/level': 1,\n",
       "         'mitigates': 1,\n",
       "         'implementing': 2,\n",
       "         'eligibility': 3,\n",
       "         'i4.0': 1,\n",
       "         'uses': 3,\n",
       "         'mouse': 3,\n",
       "         'treaty': 3,\n",
       "         'unix-based': 2,\n",
       "         'specialists': 4,\n",
       "         'resume/cv': 1,\n",
       "         'multitask': 2,\n",
       "         'rapid': 5,\n",
       "         'experimentation': 7,\n",
       "         'successes': 1,\n",
       "         'partners': 37,\n",
       "         'english': 11,\n",
       "         'version': 6,\n",
       "         'starters': 1,\n",
       "         'every': 22,\n",
       "         'use': 49,\n",
       "         'outcomes': 10,\n",
       "         ...})"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "counter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "4db0b844",
   "metadata": {},
   "outputs": [],
   "source": [
    "num_words = len(counter)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "e2496ef2",
   "metadata": {},
   "outputs": [],
   "source": [
    "#max number of words in a sequence\n",
    "#replace with length of longest string\n",
    "max_length = 20"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "0f831536",
   "metadata": {},
   "outputs": [],
   "source": [
    "#from sklearn.model_selection import train_test_split\n",
    "#X_train, X_test, y_train, y_test = train_test_split( df['Phrase'], df['Target'], test_size=0.2, random_state=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "912a9f49",
   "metadata": {},
   "outputs": [],
   "source": [
    "def split_train_test(features, target, size):\n",
    "    #size is the size of the training set\n",
    "    train_size = int(features.shape[0]* size) #check me\n",
    "    X_train = features[:train_size]\n",
    "    y_train = target[:train_size]\n",
    "\n",
    "    X_test = features[train_size:]\n",
    "    y_test = target[train_size:]\n",
    "    return X_train, y_train, X_test, y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "9c15590f",
   "metadata": {},
   "outputs": [],
   "source": [
    "train = df[['Phrase','Target']]\n",
    "features = train['Phrase']\n",
    "target = train['Target']\n",
    "X_train, y_train, X_test, y_test = split_train_test(features, target, 0.8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "687b2405",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Init Plugin\n",
      "Init Graph Optimizer\n",
      "Init Kernel\n"
     ]
    }
   ],
   "source": [
    "from keras.preprocessing.text import Tokenizer\n",
    "tokenizer = Tokenizer(num_words=num_words)\n",
    "tokenizer.fit_on_texts(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "fb07921a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'data': 1,\n",
       " 'experience': 2,\n",
       " 'business': 3,\n",
       " 'team': 4,\n",
       " 'work': 5,\n",
       " 'development': 6,\n",
       " 'skills': 7,\n",
       " 'analytics': 8,\n",
       " 'science': 9,\n",
       " 'solutions': 10,\n",
       " 'management': 11,\n",
       " 'technical': 12,\n",
       " '’': 13,\n",
       " 'knowledge': 14,\n",
       " 'analysis': 15,\n",
       " 'product': 16,\n",
       " 'ability': 17,\n",
       " 'strong': 18,\n",
       " 'environment': 19,\n",
       " 'machine': 20,\n",
       " 'time': 21,\n",
       " 'software': 22,\n",
       " 'new': 23,\n",
       " 'engineering': 24,\n",
       " 'tools': 25,\n",
       " 'design': 26,\n",
       " 'support': 27,\n",
       " 'opportunity': 28,\n",
       " 'ai': 29,\n",
       " 'research': 30,\n",
       " 'models': 31,\n",
       " 'high': 32,\n",
       " 'technology': 33,\n",
       " 'company': 34,\n",
       " 'learning': 35,\n",
       " 'cloud': 36,\n",
       " 'teams': 37,\n",
       " 'technologies': 38,\n",
       " 'projects': 39,\n",
       " 'process': 40,\n",
       " 'analytical': 41,\n",
       " 'job': 42,\n",
       " 'complex': 43,\n",
       " 'customer': 44,\n",
       " 'statistical': 45,\n",
       " 'requirements': 46,\n",
       " 'industry': 47,\n",
       " 'products': 48,\n",
       " 'systems': 49,\n",
       " 'quality': 50,\n",
       " 'role': 51,\n",
       " 'services': 52,\n",
       " 'performance': 53,\n",
       " 'problems': 54,\n",
       " 'information': 55,\n",
       " 'people': 56,\n",
       " 'degree': 57,\n",
       " 'understanding': 58,\n",
       " 'world': 59,\n",
       " 'python': 60,\n",
       " 'sql': 61,\n",
       " 'insights': 62,\n",
       " 'digital': 63,\n",
       " 'customers': 64,\n",
       " 'ml': 65,\n",
       " 'stakeholders': 66,\n",
       " 'scientist': 67,\n",
       " 'computer': 68,\n",
       " 'platform': 69,\n",
       " 'various': 70,\n",
       " 'opportunities': 71,\n",
       " 'impact': 72,\n",
       " 'applications': 73,\n",
       " 'key': 74,\n",
       " 'large': 75,\n",
       " 'full': 76,\n",
       " 'global': 77,\n",
       " 'deep': 78,\n",
       " 'life': 79,\n",
       " 'health': 80,\n",
       " 'project': 81,\n",
       " 'value': 82,\n",
       " 'model': 83,\n",
       " 'leadership': 84,\n",
       " 'multiple': 85,\n",
       " 'level': 86,\n",
       " 'techniques': 87,\n",
       " 'senior': 88,\n",
       " 'canada': 89,\n",
       " 'expertise': 90,\n",
       " 'needs': 91,\n",
       " 'big': 92,\n",
       " 'employment': 93,\n",
       " 'position': 94,\n",
       " 'platforms': 95,\n",
       " 'building': 96,\n",
       " 'clients': 97,\n",
       " 'use': 98,\n",
       " 'growth': 99,\n",
       " 'solution': 100,\n",
       " 'related': 101,\n",
       " 'field': 102,\n",
       " 'culture': 103,\n",
       " 'problem': 104,\n",
       " 'or': 105,\n",
       " 'end': 106,\n",
       " 'lead': 107,\n",
       " 'drive': 108,\n",
       " 'strategy': 109,\n",
       " 'communication': 110,\n",
       " 'and': 111,\n",
       " 'engineer': 112,\n",
       " 'group': 113,\n",
       " 'internal': 114,\n",
       " 'client': 115,\n",
       " 'quantitative': 116,\n",
       " 'application': 117,\n",
       " 'help': 118,\n",
       " 'architecture': 119,\n",
       " 'part': 120,\n",
       " 'asset': 121,\n",
       " 'join': 122,\n",
       " 'diverse': 123,\n",
       " '–': 124,\n",
       " 'excellent': 125,\n",
       " 'infrastructure': 126,\n",
       " 'benefits': 127,\n",
       " 'equal': 128,\n",
       " 'program': 129,\n",
       " 'real': 130,\n",
       " 'vision': 131,\n",
       " 'office': 132,\n",
       " 'diversity': 133,\n",
       " 'delivery': 134,\n",
       " 'results': 135,\n",
       " 'market': 136,\n",
       " 'plan': 137,\n",
       " 'innovation': 138,\n",
       " 'organization': 139,\n",
       " 'innovative': 140,\n",
       " 'qualifications': 141,\n",
       " 'financial': 142,\n",
       " 'successful': 143,\n",
       " 'employer': 144,\n",
       " 'production': 145,\n",
       " 'candidate': 146,\n",
       " 'integration': 147,\n",
       " 'implement': 148,\n",
       " 'sources': 149,\n",
       " 'algorithms': 150,\n",
       " 'advanced': 151,\n",
       " 'employee': 152,\n",
       " 'aws': 153,\n",
       " 'implementation': 154,\n",
       " 'code': 155,\n",
       " 'activities': 156,\n",
       " 'career': 157,\n",
       " 'future': 158,\n",
       " 'different': 159,\n",
       " 'intelligence': 160,\n",
       " 'decision': 161,\n",
       " 'education': 162,\n",
       " 'security': 163,\n",
       " 'driven': 164,\n",
       " 'self': 165,\n",
       " 'able': 166,\n",
       " 'relevant': 167,\n",
       " 'create': 168,\n",
       " 'great': 169,\n",
       " 'experiences': 170,\n",
       " 'enterprise': 171,\n",
       " 'based': 172,\n",
       " 'marketing': 173,\n",
       " 'scale': 174,\n",
       " 'variety': 175,\n",
       " 'service': 176,\n",
       " 'agile': 177,\n",
       " 'system': 178,\n",
       " 'decisions': 179,\n",
       " 'user': 180,\n",
       " 'methods': 181,\n",
       " 'canadian': 182,\n",
       " 'manage': 183,\n",
       " 'bi': 184,\n",
       " 'way': 185,\n",
       " 'operations': 186,\n",
       " 'pipelines': 187,\n",
       " 'r': 188,\n",
       " 'responsibilities': 189,\n",
       " 'on': 190,\n",
       " 'collaborate': 191,\n",
       " 'com': 192,\n",
       " 'partners': 193,\n",
       " 'day': 194,\n",
       " 'power': 195,\n",
       " 'members': 196,\n",
       " 'professional': 197,\n",
       " 'visualization': 198,\n",
       " 'modeling': 199,\n",
       " 'success': 200,\n",
       " 'critical': 201,\n",
       " 'insurance': 202,\n",
       " 'spark': 203,\n",
       " 'proficiency': 204,\n",
       " 'reports': 205,\n",
       " 'build': 206,\n",
       " 'processing': 207,\n",
       " 'manager': 208,\n",
       " 'azure': 209,\n",
       " 'strategic': 210,\n",
       " 'external': 211,\n",
       " 'practices': 212,\n",
       " 'applicants': 213,\n",
       " 'partner': 214,\n",
       " 'range': 215,\n",
       " 'areas': 216,\n",
       " 'good': 217,\n",
       " 'capabilities': 218,\n",
       " 'share': 219,\n",
       " 'language': 220,\n",
       " 'interest': 221,\n",
       " 'functional': 222,\n",
       " 'modern': 223,\n",
       " 'predictive': 224,\n",
       " 'core': 225,\n",
       " 'pipeline': 226,\n",
       " 'access': 227,\n",
       " 'individuals': 228,\n",
       " 'analyst': 229,\n",
       " 'care': 230,\n",
       " 'understand': 231,\n",
       " 'accommodation': 232,\n",
       " 'training': 233,\n",
       " 'open': 234,\n",
       " 'to': 235,\n",
       " 'responsible': 236,\n",
       " 'reporting': 237,\n",
       " 'employees': 238,\n",
       " 'years': 239,\n",
       " 'collaboration': 240,\n",
       " 'risk': 241,\n",
       " 'every': 242,\n",
       " 'wide': 243,\n",
       " 'companies': 244,\n",
       " 'operational': 245,\n",
       " 'across': 246,\n",
       " 'passionate': 247,\n",
       " 'engineers': 248,\n",
       " 'status': 249,\n",
       " 'appropriate': 250,\n",
       " 'remote': 251,\n",
       " 'optimization': 252,\n",
       " 'programs': 253,\n",
       " 'challenges': 254,\n",
       " 'available': 255,\n",
       " 'microsoft': 256,\n",
       " 'hands': 257,\n",
       " 'hours': 258,\n",
       " 'web': 259,\n",
       " 'etl': 260,\n",
       " 'plans': 261,\n",
       " 'values': 262,\n",
       " 'network': 263,\n",
       " 'expert': 264,\n",
       " 'scientists': 265,\n",
       " 'leader': 266,\n",
       " 'media': 267,\n",
       " 'database': 268,\n",
       " 'energy': 269,\n",
       " 'bachelor': 270,\n",
       " 'approaches': 271,\n",
       " 'in': 272,\n",
       " 'flexible': 273,\n",
       " 'datasets': 274,\n",
       " 'learn': 275,\n",
       " 'strategies': 276,\n",
       " 'meet': 277,\n",
       " 'programming': 278,\n",
       " 'communicate': 279,\n",
       " 'unique': 280,\n",
       " 'inclusive': 281,\n",
       " 'workplace': 282,\n",
       " 'location': 283,\n",
       " 'statistics': 284,\n",
       " 'www': 285,\n",
       " 'frameworks': 286,\n",
       " 'mission': 287,\n",
       " 'excellence': 288,\n",
       " 'date': 289,\n",
       " 'significant': 290,\n",
       " 'description': 291,\n",
       " 'concepts': 292,\n",
       " 'collaborative': 293,\n",
       " 'initiatives': 294,\n",
       " 'post': 295,\n",
       " 'ideas': 296,\n",
       " 'candidates': 297,\n",
       " 'non': 298,\n",
       " 'recruitment': 299,\n",
       " 'individual': 300,\n",
       " 'present': 301,\n",
       " 'mining': 302,\n",
       " 'community': 303,\n",
       " 'passion': 304,\n",
       " 'human': 305,\n",
       " 'deployment': 306,\n",
       " 'sciences': 307,\n",
       " 'detail': 308,\n",
       " 'term': 309,\n",
       " 'medical': 310,\n",
       " 'online': 311,\n",
       " 'social': 312,\n",
       " 'similar': 313,\n",
       " 'methodologies': 314,\n",
       " 'banking': 315,\n",
       " 'users': 316,\n",
       " 'analyze': 317,\n",
       " 'toronto': 318,\n",
       " 'equivalent': 319,\n",
       " 'sets': 320,\n",
       " 'change': 321,\n",
       " 'framework': 322,\n",
       " 'review': 323,\n",
       " 'permanent': 324,\n",
       " 'clinical': 325,\n",
       " 'right': 326,\n",
       " 'maintain': 327,\n",
       " 'environments': 328,\n",
       " 'safety': 329,\n",
       " 'improvement': 330,\n",
       " 'additional': 331,\n",
       " 'metrics': 332,\n",
       " 'scientific': 333,\n",
       " 'equity': 334,\n",
       " 'dashboards': 335,\n",
       " 'tasks': 336,\n",
       " 'dynamic': 337,\n",
       " 'events': 338,\n",
       " 'potential': 339,\n",
       " 'ideal': 340,\n",
       " 'developers': 341,\n",
       " 'cross': 342,\n",
       " 'action': 343,\n",
       " 'approach': 344,\n",
       " 'solving': 345,\n",
       " 'disability': 346,\n",
       " 'contract': 347,\n",
       " 'focus': 348,\n",
       " 'c': 349,\n",
       " 'member': 350,\n",
       " 'selection': 351,\n",
       " 'relationships': 352,\n",
       " 'natural': 353,\n",
       " 'discipline': 354,\n",
       " 'leaders': 355,\n",
       " 'fast': 356,\n",
       " 'home': 357,\n",
       " 'planning': 358,\n",
       " 'effective': 359,\n",
       " 'ca': 360,\n",
       " 'languages': 361,\n",
       " 'many': 362,\n",
       " 'test': 363,\n",
       " 'governance': 364,\n",
       " 'multi': 365,\n",
       " 'e': 366,\n",
       " 'inclusion': 367,\n",
       " 'accuracy': 368,\n",
       " 'salary': 369,\n",
       " 'background': 370,\n",
       " 'enhance': 371,\n",
       " 'businesses': 372,\n",
       " 'analytic': 373,\n",
       " 'primary': 374,\n",
       " 'major': 375,\n",
       " 'execution': 376,\n",
       " 'type': 377,\n",
       " '·': 378,\n",
       " 'processes': 379,\n",
       " 'assistance': 380,\n",
       " 'policy': 381,\n",
       " 'monitoring': 382,\n",
       " 'proven': 383,\n",
       " 'influence': 384,\n",
       " 'address': 385,\n",
       " 'continuous': 386,\n",
       " 'millions': 387,\n",
       " 'positive': 388,\n",
       " 'compliance': 389,\n",
       " 'attention': 390,\n",
       " 'commitment': 391,\n",
       " 'issues': 392,\n",
       " 'family': 393,\n",
       " 'master': 394,\n",
       " 'meaningful': 395,\n",
       " 'guidance': 396,\n",
       " 'personal': 397,\n",
       " 'bank': 398,\n",
       " 'procedures': 399,\n",
       " 'subject': 400,\n",
       " 'comprehensive': 401,\n",
       " 'smart': 402,\n",
       " 'findings': 403,\n",
       " 'act': 404,\n",
       " 'investment': 405,\n",
       " 'transformation': 406,\n",
       " 'efficiency': 407,\n",
       " 'communities': 408,\n",
       " 'ecosystem': 409,\n",
       " 'schedule': 410,\n",
       " 'creation': 411,\n",
       " 'current': 412,\n",
       " 'class': 413,\n",
       " 'goals': 414,\n",
       " 'enable': 415,\n",
       " 'warehouse': 416,\n",
       " 'creative': 417,\n",
       " 'stack': 418,\n",
       " 'phd': 419,\n",
       " 'public': 420,\n",
       " 'record': 421,\n",
       " 'university': 422,\n",
       " 'hadoop': 423,\n",
       " 'apply': 424,\n",
       " 'features': 425,\n",
       " 'identify': 426,\n",
       " 'analysts': 427,\n",
       " 'familiarity': 428,\n",
       " 'place': 429,\n",
       " 'making': 430,\n",
       " 'competitive': 431,\n",
       " 'compensation': 432,\n",
       " 'types': 433,\n",
       " '”': 434,\n",
       " 'devops': 435,\n",
       " 'source': 436,\n",
       " 'amazon': 437,\n",
       " 'ownership': 438,\n",
       " 'flexibility': 439,\n",
       " 'goal': 440,\n",
       " 'collection': 441,\n",
       " 'nlp': 442,\n",
       " 'game': 443,\n",
       " 'finance': 444,\n",
       " 'year': 445,\n",
       " 'tech': 446,\n",
       " 'solid': 447,\n",
       " 'leverage': 448,\n",
       " 'purpose': 449,\n",
       " 'roadmap': 450,\n",
       " 'architectures': 451,\n",
       " 'experiments': 452,\n",
       " 'first': 453,\n",
       " 'analyses': 454,\n",
       " 'covid': 455,\n",
       " 'documentation': 456,\n",
       " 'ready': 457,\n",
       " 'excel': 458,\n",
       " 'specific': 459,\n",
       " 'top': 460,\n",
       " 'depth': 461,\n",
       " 'broad': 462,\n",
       " 'scope': 463,\n",
       " 'cases': 464,\n",
       " 'everyone': 465,\n",
       " 'long': 466,\n",
       " 'resources': 467,\n",
       " 'executive': 468,\n",
       " 'track': 469,\n",
       " 'state': 470,\n",
       " 'organizations': 471,\n",
       " 'validation': 472,\n",
       " 'principles': 473,\n",
       " '“': 474,\n",
       " 'execute': 475,\n",
       " 'provide': 476,\n",
       " 'credit': 477,\n",
       " 'package': 478,\n",
       " 'levels': 479,\n",
       " 'standards': 480,\n",
       " 'conduct': 481,\n",
       " 'aspects': 482,\n",
       " 'sales': 483,\n",
       " 'disabilities': 484,\n",
       " 'careers': 485,\n",
       " 'modelling': 486,\n",
       " 'essential': 487,\n",
       " 'practice': 488,\n",
       " 'java': 489,\n",
       " 'recommendations': 490,\n",
       " 'strength': 491,\n",
       " 'search': 492,\n",
       " 'ways': 493,\n",
       " 'measure': 494,\n",
       " 'backgrounds': 495,\n",
       " 'nosql': 496,\n",
       " 'contribute': 497,\n",
       " 'deliver': 498,\n",
       " 'cycle': 499,\n",
       " 'improvements': 500,\n",
       " 'study': 501,\n",
       " 'preferred': 502,\n",
       " 'databases': 503,\n",
       " 'efficient': 504,\n",
       " 'regression': 505,\n",
       " 'friday': 506,\n",
       " 'center': 507,\n",
       " 'retail': 508,\n",
       " 'practical': 509,\n",
       " 'translate': 510,\n",
       " 'within': 511,\n",
       " 'everything': 512,\n",
       " 'develop': 513,\n",
       " 'welcome': 514,\n",
       " 'math': 515,\n",
       " 'active': 516,\n",
       " 'note': 517,\n",
       " 'exposure': 518,\n",
       " 'objectives': 519,\n",
       " 'actionable': 520,\n",
       " 'visit': 521,\n",
       " 'department': 522,\n",
       " 'order': 523,\n",
       " 'define': 524,\n",
       " 'solve': 525,\n",
       " 'matter': 526,\n",
       " 'google': 527,\n",
       " 'area': 528,\n",
       " 'mathematics': 529,\n",
       " 'presentation': 530,\n",
       " 'montreal': 531,\n",
       " 'scalable': 532,\n",
       " 'transform': 533,\n",
       " 'jobs': 534,\n",
       " 'managers': 535,\n",
       " 'staff': 536,\n",
       " 'git': 537,\n",
       " 'player': 538,\n",
       " 'content': 539,\n",
       " 'mobile': 540,\n",
       " 'pytorch': 541,\n",
       " 'the': 542,\n",
       " 'thinking': 543,\n",
       " 'make': 544,\n",
       " 'small': 545,\n",
       " 'algorithm': 546,\n",
       " 'balance': 547,\n",
       " 'director': 548,\n",
       " 'manner': 549,\n",
       " 'important': 550,\n",
       " 'familiar': 551,\n",
       " '19': 552,\n",
       " 'talent': 553,\n",
       " 'secondary': 554,\n",
       " 'questions': 555,\n",
       " 'accessible': 556,\n",
       " 'edge': 557,\n",
       " 'methodology': 558,\n",
       " 'interpersonal': 559,\n",
       " 'wellness': 560,\n",
       " 'groups': 561,\n",
       " 'maintenance': 562,\n",
       " 'lifecycle': 563,\n",
       " 'contact': 564,\n",
       " 'private': 565,\n",
       " 'apis': 566,\n",
       " 'environmental': 567,\n",
       " 'art': 568,\n",
       " 'basis': 569,\n",
       " 'tableau': 570,\n",
       " 'advice': 571,\n",
       " 'lab': 572,\n",
       " 'inform': 573,\n",
       " 'audiences': 574,\n",
       " 'duties': 575,\n",
       " 'person': 576,\n",
       " 'optimize': 577,\n",
       " 'standard': 578,\n",
       " 'server': 579,\n",
       " 'motivated': 580,\n",
       " 'libraries': 581,\n",
       " 'thinker': 582,\n",
       " 'vancouver': 583,\n",
       " 'interview': 584,\n",
       " 'ad': 585,\n",
       " 'fields': 586,\n",
       " 'city': 587,\n",
       " 'things': 588,\n",
       " 'month': 589,\n",
       " 'consumer': 590,\n",
       " 'adoption': 591,\n",
       " 'extensive': 592,\n",
       " 'overall': 593,\n",
       " 'free': 594,\n",
       " 'matters': 595,\n",
       " 'english': 596,\n",
       " 'assessment': 597,\n",
       " 'dental': 598,\n",
       " 'app': 599,\n",
       " 'query': 600,\n",
       " 'pharmaceutical': 601,\n",
       " 'clear': 602,\n",
       " 'series': 603,\n",
       " 'preparation': 604,\n",
       " 'components': 605,\n",
       " 'workflows': 606,\n",
       " 'start': 607,\n",
       " 'direction': 608,\n",
       " 'paced': 609,\n",
       " 'tensorflow': 610,\n",
       " 'shape': 611,\n",
       " 'organizational': 612,\n",
       " 'networks': 613,\n",
       " 'graduate': 614,\n",
       " 'acquisition': 615,\n",
       " 'domain': 616,\n",
       " 'international': 617,\n",
       " 'fundamentals': 618,\n",
       " 'tests': 619,\n",
       " 'progress': 620,\n",
       " 'others': 621,\n",
       " 'corporate': 622,\n",
       " 'general': 623,\n",
       " 'conclusions': 624,\n",
       " 'markets': 625,\n",
       " 'ibm': 626,\n",
       " 'functions': 627,\n",
       " 'packages': 628,\n",
       " 'national': 629,\n",
       " 'recognition': 630,\n",
       " 'scala': 631,\n",
       " 'fun': 632,\n",
       " 'control': 633,\n",
       " 'verbal': 634,\n",
       " 'front': 635,\n",
       " 's': 636,\n",
       " 'sports': 637,\n",
       " 'offers': 638,\n",
       " 'cell': 639,\n",
       " 'proud': 640,\n",
       " 'embrace': 641,\n",
       " 'overview': 642,\n",
       " 'regular': 643,\n",
       " 'recommendation': 644,\n",
       " 'sense': 645,\n",
       " 'changes': 646,\n",
       " 'privacy': 647,\n",
       " 'outcomes': 648,\n",
       " 'accessibility': 649,\n",
       " 'presentations': 650,\n",
       " 'of': 651,\n",
       " 'sector': 652,\n",
       " 'flow': 653,\n",
       " 'email': 654,\n",
       " 'owners': 655,\n",
       " 'context': 656,\n",
       " 'roles': 657,\n",
       " 'stage': 658,\n",
       " 'startup': 659,\n",
       " 'relationship': 660,\n",
       " 'race': 661,\n",
       " 'artificial': 662,\n",
       " 'creativity': 663,\n",
       " 'next': 664,\n",
       " 'healthcare': 665,\n",
       " 'commercial': 666,\n",
       " 'b': 667,\n",
       " 'automation': 668,\n",
       " 'mentor': 669,\n",
       " 'engagement': 670,\n",
       " 'regulations': 671,\n",
       " 'capacity': 672,\n",
       " 'neural': 673,\n",
       " 'please': 674,\n",
       " 'proficient': 675,\n",
       " 'a': 676,\n",
       " 'like': 677,\n",
       " 'colleagues': 678,\n",
       " 'feature': 679,\n",
       " 'designs': 680,\n",
       " 'journey': 681,\n",
       " 'traditional': 682,\n",
       " 'revenue': 683,\n",
       " 'responsibility': 684,\n",
       " 'applied': 685,\n",
       " 'advance': 686,\n",
       " 'dimensional': 687,\n",
       " 'peers': 688,\n",
       " 'protection': 689,\n",
       " 'thought': 690,\n",
       " 'grow': 691,\n",
       " 'portfolio': 692,\n",
       " 'transportation': 693,\n",
       " 'studies': 694,\n",
       " 'extract': 695,\n",
       " 'resume': 696,\n",
       " 'trends': 697,\n",
       " 'experienced': 698,\n",
       " 'applicable': 699,\n",
       " 'executives': 700,\n",
       " 'site': 701,\n",
       " 'thank': 702,\n",
       " 'centric': 703,\n",
       " 'experts': 704,\n",
       " 'speed': 705,\n",
       " 'details': 706,\n",
       " 'students': 707,\n",
       " 'draw': 708,\n",
       " 'north': 709,\n",
       " 'contributions': 710,\n",
       " 'custom': 711,\n",
       " 'need': 712,\n",
       " 'providers': 713,\n",
       " 'purchase': 714,\n",
       " 'm': 715,\n",
       " 'rest': 716,\n",
       " 'challenge': 717,\n",
       " 'discovery': 718,\n",
       " 'exceptional': 719,\n",
       " 'clean': 720,\n",
       " 'junior': 721,\n",
       " 'assets': 722,\n",
       " 'story': 723,\n",
       " 'laboratory': 724,\n",
       " 'deploy': 725,\n",
       " 'heart': 726,\n",
       " 'mindset': 727,\n",
       " 'early': 728,\n",
       " 'sustainability': 729,\n",
       " 'innovate': 730,\n",
       " 'terms': 731,\n",
       " 'due': 732,\n",
       " 'explain': 733,\n",
       " 'professionals': 734,\n",
       " 'bonus': 735,\n",
       " 'firm': 736,\n",
       " 'academic': 737,\n",
       " 'biology': 738,\n",
       " 'offer': 739,\n",
       " 'today': 740,\n",
       " 'volumes': 741,\n",
       " 'mobility': 742,\n",
       " 'options': 743,\n",
       " 'countries': 744,\n",
       " 'l': 745,\n",
       " 'willing': 746,\n",
       " 'possible': 747,\n",
       " 'input': 748,\n",
       " 'common': 749,\n",
       " 'detection': 750,\n",
       " 'attitude': 751,\n",
       " 'annual': 752,\n",
       " 'cutting': 753,\n",
       " 'trust': 754,\n",
       " 'complete': 755,\n",
       " 'hard': 756,\n",
       " 'well': 757,\n",
       " 'cost': 758,\n",
       " 'budget': 759,\n",
       " 'throughout': 760,\n",
       " '—': 761,\n",
       " 'generate': 762,\n",
       " 'savings': 763,\n",
       " 'identification': 764,\n",
       " 'downtown': 765,\n",
       " 'architect': 766,\n",
       " 'resource': 767,\n",
       " 'capital': 768,\n",
       " 'changing': 769,\n",
       " 'chemistry': 770,\n",
       " 'rules': 771,\n",
       " 'risks': 772,\n",
       " 'communications': 773,\n",
       " 'voice': 774,\n",
       " 'priorities': 775,\n",
       " 'perform': 776,\n",
       " 'discussions': 777,\n",
       " 'offerings': 778,\n",
       " 'combination': 779,\n",
       " 'deadlines': 780,\n",
       " 'mathematical': 781,\n",
       " 'chance': 782,\n",
       " 'demonstrated': 783,\n",
       " 'patterns': 784,\n",
       " 'calgary': 785,\n",
       " 'summary': 786,\n",
       " 'persons': 787,\n",
       " 'committed': 788,\n",
       " 'mandatory': 789,\n",
       " 'points': 790,\n",
       " 'unit': 791,\n",
       " 'gender': 792,\n",
       " 'guide': 793,\n",
       " 'sustainable': 794,\n",
       " 'engine': 795,\n",
       " 'insight': 796,\n",
       " 'documents': 797,\n",
       " 'certification': 798,\n",
       " 'play': 799,\n",
       " 'wealth': 800,\n",
       " 'bell': 801,\n",
       " 'monday': 802,\n",
       " 'stem': 803,\n",
       " 'extraction': 804,\n",
       " 'competency': 805,\n",
       " 'lines': 806,\n",
       " 'function': 807,\n",
       " 'mental': 808,\n",
       " 'id': 809,\n",
       " 'minimum': 810,\n",
       " 'division': 811,\n",
       " 'college': 812,\n",
       " 'pension': 813,\n",
       " 'travel': 814,\n",
       " 'testing': 815,\n",
       " 'connect': 816,\n",
       " 'path': 817,\n",
       " 'virtual': 818,\n",
       " 'coding': 819,\n",
       " 'computing': 820,\n",
       " 'validate': 821,\n",
       " 'makers': 822,\n",
       " 'text': 823,\n",
       " 'drug': 824,\n",
       " 'transfer': 825,\n",
       " 'agencies': 826,\n",
       " 'student': 827,\n",
       " 'document': 828,\n",
       " 'government': 829,\n",
       " 'publications': 830,\n",
       " 'possess': 831,\n",
       " 'main': 832,\n",
       " 'secure': 833,\n",
       " 'someone': 834,\n",
       " 'venture': 835,\n",
       " 'assist': 836,\n",
       " 'reasonable': 837,\n",
       " 'ms': 838,\n",
       " 'relational': 839,\n",
       " 'popular': 840,\n",
       " 'participate': 841,\n",
       " 'veeva': 842,\n",
       " 'fit': 843,\n",
       " 'sas': 844,\n",
       " 'forecasting': 845,\n",
       " 'local': 846,\n",
       " 'scrum': 847,\n",
       " 'nice': 848,\n",
       " 'abilities': 849,\n",
       " 'evolution': 850,\n",
       " 'apache': 851,\n",
       " 'commerce': 852,\n",
       " 'behavior': 853,\n",
       " 'conferences': 854,\n",
       " 'novel': 855,\n",
       " 'bring': 856,\n",
       " 'space': 857,\n",
       " 'device': 858,\n",
       " 'report': 859,\n",
       " 'accommodations': 860,\n",
       " 'loyalty': 861,\n",
       " 'accelerate': 862,\n",
       " 'unstructured': 863,\n",
       " 'sure': 864,\n",
       " 'cpp': 865,\n",
       " 'locations': 866,\n",
       " 'docker': 867,\n",
       " 'stantec': 868,\n",
       " 'td': 869,\n",
       " 'set': 870,\n",
       " 'qualified': 871,\n",
       " 'want': 872,\n",
       " 'stock': 873,\n",
       " 'necessary': 874,\n",
       " 'discover': 875,\n",
       " 'economic': 876,\n",
       " 'hire': 877,\n",
       " 'format': 878,\n",
       " 'lives': 879,\n",
       " 'effectiveness': 880,\n",
       " 'you': 881,\n",
       " 'something': 882,\n",
       " 'skill': 883,\n",
       " 'ontario': 884,\n",
       " 'promote': 885,\n",
       " 'leading': 886,\n",
       " 'number': 887,\n",
       " 'iot': 888,\n",
       " 'situations': 889,\n",
       " 'desire': 890,\n",
       " 'pay': 891,\n",
       " 'means': 892,\n",
       " 'usage': 893,\n",
       " 'industrial': 894,\n",
       " 'manufacturing': 895,\n",
       " 'airflow': 896,\n",
       " 'availability': 897,\n",
       " 'strive': 898,\n",
       " 'warehousing': 899,\n",
       " 'globe': 900,\n",
       " '2': 901,\n",
       " 'coverage': 902,\n",
       " 'school': 903,\n",
       " 'outputs': 904,\n",
       " 'intelligent': 905,\n",
       " 'batch': 906,\n",
       " 'lake': 907,\n",
       " 'scripts': 908,\n",
       " 'exploration': 909,\n",
       " 'expectations': 910,\n",
       " 'foster': 911,\n",
       " 'commitments': 912,\n",
       " 'physical': 913,\n",
       " 'msc': 914,\n",
       " 'prioritize': 915,\n",
       " 'jira': 916,\n",
       " 'interactions': 917,\n",
       " 'ambiguity': 918,\n",
       " 'supportive': 919,\n",
       " 'entire': 920,\n",
       " 'structure': 921,\n",
       " 'inc': 922,\n",
       " 'growing': 923,\n",
       " 'sound': 924,\n",
       " 'tool': 925,\n",
       " 'demand': 926,\n",
       " 'centre': 927,\n",
       " 'interact': 928,\n",
       " 'logical': 929,\n",
       " 'co': 930,\n",
       " 'week': 931,\n",
       " 'difference': 932,\n",
       " 'labs': 933,\n",
       " 'linkedin': 934,\n",
       " 'experimental': 935,\n",
       " 'researchers': 936,\n",
       " 'informed': 937,\n",
       " 'thrive': 938,\n",
       " 'brands': 939,\n",
       " 'differences': 940,\n",
       " 'keep': 941,\n",
       " 'assess': 942,\n",
       " 'close': 943,\n",
       " 'computational': 944,\n",
       " 'direct': 945,\n",
       " 'visualizations': 946,\n",
       " 'internet': 947,\n",
       " 'channels': 948,\n",
       " 'minimal': 949,\n",
       " 'point': 950,\n",
       " 'consistent': 951,\n",
       " 'workforce': 952,\n",
       " 'brand': 953,\n",
       " 'campaign': 954,\n",
       " 'serve': 955,\n",
       " 'profile': 956,\n",
       " 'up': 957,\n",
       " 'perspectives': 958,\n",
       " 'des': 959,\n",
       " 'saas': 960,\n",
       " 'vacation': 961,\n",
       " 'towards': 962,\n",
       " 'view': 963,\n",
       " 'disciplines': 964,\n",
       " 'days': 965,\n",
       " 'click': 966,\n",
       " 'alternative': 967,\n",
       " 'developer': 968,\n",
       " 'treatment': 969,\n",
       " 'impactful': 970,\n",
       " 'know': 971,\n",
       " 'image': 972,\n",
       " 'celebrate': 973,\n",
       " 'several': 974,\n",
       " 'designing': 975,\n",
       " 'manipulation': 976,\n",
       " 'teamwork': 977,\n",
       " 'stages': 978,\n",
       " 'give': 979,\n",
       " 'monitor': 980,\n",
       " 'accenture': 981,\n",
       " 'evaluate': 982,\n",
       " 'productivity': 983,\n",
       " 'etc': 984,\n",
       " 'require': 985,\n",
       " 'automate': 986,\n",
       " 'transformations': 987,\n",
       " 'equipment': 988,\n",
       " 'improve': 989,\n",
       " 'volume': 990,\n",
       " 'partnership': 991,\n",
       " 'ground': 992,\n",
       " 'stream': 993,\n",
       " 'automotive': 994,\n",
       " 'multidisciplinary': 995,\n",
       " 'hive': 996,\n",
       " 'assessments': 997,\n",
       " 'reference': 998,\n",
       " 'kinesis': 999,\n",
       " 'entrepreneurial': 1000,\n",
       " ...}"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "word_index = tokenizer.word_index\n",
    "word_index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "1b468da2",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_sequences = tokenizer.texts_to_sequences(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "350f04f1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[290, 463, 72]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_sequences[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "53c1a49c",
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.preprocessing.sequence import pad_sequences\n",
    "\n",
    "train_padded = pad_sequences(\n",
    "    train_sequences, maxlen=max_length, padding=\"post\", truncating=\"post\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "e642d181",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([290, 463,  72,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
       "         0,   0,   0,   0,   0,   0,   0], dtype=int32)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_padded[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "d461b0d7",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_sequences = tokenizer.texts_to_sequences(X_test)\n",
    "test_padded = pad_sequences(\n",
    "    test_sequences, maxlen=max_length, padding=\"post\", truncating=\"post\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "e516eca7",
   "metadata": {},
   "outputs": [],
   "source": [
    "#check inverse\n",
    "reverse_word_index = dict([(value,key) for (key,value) in word_index.items()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "87100382",
   "metadata": {},
   "outputs": [],
   "source": [
    "def decode(text):\n",
    "    return \" \".join([reverse_word_index.get(i, \"?\") for i in text])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "923278d8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'significant scope impact'"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "decode(train_sequences[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "9235786a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "significant scope impact\n"
     ]
    }
   ],
   "source": [
    "print(df['Phrase'][0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "6ea15af7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10403, 20)\n",
      "(2601, 20)\n"
     ]
    }
   ],
   "source": [
    "print(train_padded.shape)\n",
    "print(test_padded.shape)\n",
    "#find out if the weight of the tokens affect the accuracy of the lstm"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "83f8a263",
   "metadata": {},
   "source": [
    "## Model 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "0ca8fbe0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Metal device set to: Apple M1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-08-17 15:43:52.109816: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:305] Could not identify NUMA node of platform GPU ID 0, defaulting to 0. Your kernel may not have been built with NUMA support.\n",
      "2021-08-17 15:43:52.110131: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:271] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 0 MB memory) -> physical PluggableDevice (device: 0, name: METAL, pci bus id: <undefined>)\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Embedding, LSTM, Dense, Dropout\n",
    "from keras.initializers import Constant\n",
    "\n",
    "model = tf.keras.models.Sequential([\n",
    "    # Shape [batch, time, features] => [batch, time, lstm_units]\n",
    "    tf.keras.layers.Embedding(num_words, 32, input_length=max_length),\n",
    "    tf.keras.layers.LSTM(64),\n",
    "    tf.keras.layers.Dropout(0.3),\n",
    "    tf.keras.layers.Dense(units=1, activation='sigmoid')\n",
    "])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "22aa342b",
   "metadata": {},
   "outputs": [],
   "source": [
    "opt = tf.keras.optimizers.Adam(learning_rate=0.1)\n",
    "model.compile(loss=\"binary_crossentropy\", optimizer=opt, metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "a786a038",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding (Embedding)        (None, 20, 32)            184704    \n",
      "_________________________________________________________________\n",
      "lstm (LSTM)                  (None, 64)                24832     \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 1)                 65        \n",
      "=================================================================\n",
      "Total params: 209,601\n",
      "Trainable params: 209,601\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "00e5a149",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-08-17 15:46:13.514207: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:176] None of the MLIR Optimization Passes are enabled (registered 2)\n",
      "2021-08-17 15:46:13.517179: W tensorflow/core/platform/profile_utils/cpu_utils.cc:128] Failed to get CPU frequency: 0 Hz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-08-17 15:46:13.989376: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:112] Plugin optimizer for device_type GPU is enabled.\n",
      "2021-08-17 15:46:14.111848: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:112] Plugin optimizer for device_type GPU is enabled.\n",
      "2021-08-17 15:46:14.217816: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:112] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "261/261 [==============================] - ETA: 0s - loss: 0.6498 - accuracy: 0.6587"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-08-17 15:46:21.494578: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:112] Plugin optimizer for device_type GPU is enabled.\n",
      "2021-08-17 15:46:21.534089: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:112] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "261/261 [==============================] - 9s 30ms/step - loss: 0.6498 - accuracy: 0.6587 - val_loss: 0.5841 - val_accuracy: 0.7506\n",
      "Epoch 2/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6300 - accuracy: 0.6571 - val_loss: 0.5561 - val_accuracy: 0.7751\n",
      "Epoch 3/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6372 - accuracy: 0.6598 - val_loss: 0.5942 - val_accuracy: 0.7189\n",
      "Epoch 4/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6424 - accuracy: 0.6549 - val_loss: 0.5164 - val_accuracy: 0.7751\n",
      "Epoch 5/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6627 - accuracy: 0.6486 - val_loss: 0.5347 - val_accuracy: 0.7751\n",
      "Epoch 6/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6500 - accuracy: 0.6554 - val_loss: 0.5875 - val_accuracy: 0.7059\n",
      "Epoch 7/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6813 - accuracy: 0.6316 - val_loss: 0.6662 - val_accuracy: 0.7751\n",
      "Epoch 8/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6676 - accuracy: 0.6363 - val_loss: 0.5469 - val_accuracy: 0.7751\n",
      "Epoch 9/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6727 - accuracy: 0.6365 - val_loss: 0.8644 - val_accuracy: 0.3436\n",
      "Epoch 10/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.7076 - accuracy: 0.6223 - val_loss: 0.5775 - val_accuracy: 0.7136\n",
      "Epoch 11/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6500 - accuracy: 0.6465 - val_loss: 0.5874 - val_accuracy: 0.7549\n",
      "Epoch 12/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6561 - accuracy: 0.6447 - val_loss: 0.5327 - val_accuracy: 0.7698\n",
      "Epoch 13/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6401 - accuracy: 0.6565 - val_loss: 0.5531 - val_accuracy: 0.7679\n",
      "Epoch 14/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6430 - accuracy: 0.6520 - val_loss: 0.5630 - val_accuracy: 0.7492\n",
      "Epoch 15/15\n",
      "261/261 [==============================] - 7s 28ms/step - loss: 0.6299 - accuracy: 0.6539 - val_loss: 0.5372 - val_accuracy: 0.7689\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(\n",
    "    train_padded, y_train, epochs=15, validation_split=0.2\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "4752cc30",
   "metadata": {},
   "outputs": [],
   "source": [
    "def accuracy_loss_plot(history):\n",
    "    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(14, 4))\n",
    "    ax1.plot(history.history['accuracy'])\n",
    "    ax1.plot(history.history['val_accuracy'])\n",
    "    ax1.set_title('Model accuracy')\n",
    "    ax1.set_ylabel('Accuracy')\n",
    "    ax1.set_xlabel('Epoch')\n",
    "    ax1.legend(['Train', 'Val'], loc='lower right')\n",
    "\n",
    "    ax2.plot(history.history['loss'])\n",
    "    ax2.plot(history.history['val_loss'])\n",
    "    ax2.set_title('Model loss')\n",
    "    ax2.set_ylabel('loss')\n",
    "    ax2.set_xlabel('Epoch')\n",
    "    ax2.legend(['Train', 'Val'], loc='upper right')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "034ef11b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "accuracy_loss_plot(history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "0fd2e4bb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "82/82 [==============================] - 1s 10ms/step - loss: 0.5735 - accuracy: 0.7132\n",
      "Test set\n",
      "  Loss: 0.574\n",
      "  Accuracy: 0.713\n"
     ]
    }
   ],
   "source": [
    "accr = model.evaluate(test_padded,y_test)\n",
    "print('Test set\\n  Loss: {:0.3f}\\n  Accuracy: {:0.3f}'.format(accr[0],accr[1]))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "04ccf0a2",
   "metadata": {},
   "source": [
    "## Using GloVe Embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "5d4a3027",
   "metadata": {},
   "outputs": [],
   "source": [
    "from tqdm import tqdm\n",
    "from keras.layers import Dense,SpatialDropout1D\n",
    "from nltk.tokenize import word_tokenize\n",
    "\n",
    "\"\"\"Creates corpus from feature column, which is a pandas series\"\"\"\n",
    "def create_corpus(df):\n",
    "    corpus=[]\n",
    "    for phrase in tqdm(df['Phrase']):\n",
    "        words=[word.lower() for word in word_tokenize(phrase) if(word.isalpha()==1)]\n",
    "        corpus.append(words)\n",
    "    return corpus"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "650bee35",
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"Creates an embedding dictionary using GloVe\"\"\"\n",
    "embedding_dict={}\n",
    "with open('GloVe/glove.6B.100d.txt','r') as f:\n",
    "    for line in f:\n",
    "        values=line.split()\n",
    "        word=values[0]\n",
    "        vectors=np.asarray(values[1:],'float32')\n",
    "        embedding_dict[word]=vectors\n",
    "f.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "96189709",
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"Create padded sequences of equal lenght as input to LSTM\"\"\"\n",
    "def create_padded_inputs(corpus):\n",
    "    MAX_LEN=20\n",
    "    tokenizer_obj=Tokenizer()\n",
    "    tokenizer_obj.fit_on_texts(corpus)\n",
    "    sequences=tokenizer_obj.texts_to_sequences(corpus)\n",
    "\n",
    "    phrase_pad=pad_sequences(sequences,maxlen=MAX_LEN,truncating='post',padding='post')\n",
    "    return phrase_pad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "190c2672",
   "metadata": {},
   "outputs": [],
   "source": [
    "corpus=create_corpus(df)\n",
    "phrase_pad = create_padded_inputs(corpus)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "62ee0d96",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of unique words: 4974\n"
     ]
    }
   ],
   "source": [
    "word_index=tokenizer_obj.word_index\n",
    "print('Number of unique words:',len(word_index))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "35ae2e01",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████| 4974/4974 [00:00<00:00, 787500.68it/s]\n"
     ]
    }
   ],
   "source": [
    "num_words=len(word_index)+1\n",
    "embedding_matrix=np.zeros((num_words,100))\n",
    "\n",
    "for word,i in tqdm(word_index.items()):\n",
    "    if i > num_words:\n",
    "        continue\n",
    "    \n",
    "    emb_vec=embedding_dict.get(word)\n",
    "    if emb_vec is not None:\n",
    "        embedding_matrix[i]=emb_vec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "364b7bbd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Layer lstm_1 will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n"
     ]
    }
   ],
   "source": [
    "model_embed = tf.keras.models.Sequential([\n",
    "    # Shape [batch, time, features] => [batch, time, lstm_units]\n",
    "    tf.keras.layers.Embedding(num_words,100,embeddings_initializer=Constant(embedding_matrix),\n",
    "                   input_length=MAX_LEN,trainable=False),\n",
    "    tf.keras.layers.SpatialDropout1D(0.2),\n",
    "    tf.keras.layers.LSTM(64, dropout=0.2, recurrent_dropout=0.2),\n",
    "    tf.keras.layers.Dense(units=1, activation='sigmoid')\n",
    "])\n",
    "\n",
    "\n",
    "opt = tf.keras.optimizers.Adam(learning_rate=1e-5)\n",
    "\n",
    "model_embed.compile(loss='binary_crossentropy',optimizer=opt,metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "fd2ab23b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_1 (Embedding)      (None, 20, 100)           497500    \n",
      "_________________________________________________________________\n",
      "spatial_dropout1d (SpatialDr (None, 20, 100)           0         \n",
      "_________________________________________________________________\n",
      "lstm_1 (LSTM)                (None, 64)                42240     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 65        \n",
      "=================================================================\n",
      "Total params: 539,805\n",
      "Trainable params: 42,305\n",
      "Non-trainable params: 497,500\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model_embed.summary()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "c55f71bb",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, y_train, X_test, y_test = split_train_test(phrase_pad, df['Target'], 0.8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "3917db18",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/15\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-08-17 15:48:12.217044: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:112] Plugin optimizer for device_type GPU is enabled.\n",
      "2021-08-17 16:13:30.917629: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:112] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2081/2081 - 1542s - loss: 0.6745 - accuracy: 0.6673 - val_loss: 0.5978 - val_accuracy: 0.7751\n",
      "Epoch 2/15\n",
      "2081/2081 - 1537s - loss: 0.6047 - accuracy: 0.6778 - val_loss: 0.4797 - val_accuracy: 0.7987\n",
      "Epoch 3/15\n",
      "2081/2081 - 1544s - loss: 0.5415 - accuracy: 0.7356 - val_loss: 0.4606 - val_accuracy: 0.7799\n",
      "Epoch 4/15\n",
      "2081/2081 - 1533s - loss: 0.5208 - accuracy: 0.7490 - val_loss: 0.4567 - val_accuracy: 0.7631\n",
      "Epoch 5/15\n",
      "2081/2081 - 1526s - loss: 0.5146 - accuracy: 0.7483 - val_loss: 0.4603 - val_accuracy: 0.7593\n",
      "Epoch 6/15\n",
      "2081/2081 - 1527s - loss: 0.5096 - accuracy: 0.7540 - val_loss: 0.4514 - val_accuracy: 0.7713\n",
      "Epoch 7/15\n",
      "2081/2081 - 1526s - loss: 0.5032 - accuracy: 0.7593 - val_loss: 0.4563 - val_accuracy: 0.7693\n",
      "Epoch 8/15\n",
      "2081/2081 - 1529s - loss: 0.5025 - accuracy: 0.7555 - val_loss: 0.4514 - val_accuracy: 0.7775\n",
      "Epoch 9/15\n",
      "2081/2081 - 1529s - loss: 0.4958 - accuracy: 0.7605 - val_loss: 0.4479 - val_accuracy: 0.7751\n",
      "Epoch 10/15\n",
      "2081/2081 - 1539s - loss: 0.4949 - accuracy: 0.7580 - val_loss: 0.4518 - val_accuracy: 0.7737\n",
      "Epoch 11/15\n",
      "2081/2081 - 1560s - loss: 0.4921 - accuracy: 0.7584 - val_loss: 0.4512 - val_accuracy: 0.7761\n",
      "Epoch 12/15\n",
      "2081/2081 - 1595s - loss: 0.4914 - accuracy: 0.7648 - val_loss: 0.4463 - val_accuracy: 0.7804\n",
      "Epoch 13/15\n",
      "2081/2081 - 1577s - loss: 0.4887 - accuracy: 0.7614 - val_loss: 0.4466 - val_accuracy: 0.7799\n",
      "Epoch 14/15\n",
      "2081/2081 - 1546s - loss: 0.4855 - accuracy: 0.7662 - val_loss: 0.4493 - val_accuracy: 0.7785\n",
      "Epoch 15/15\n",
      "2081/2081 - 1546s - loss: 0.4852 - accuracy: 0.7668 - val_loss: 0.4458 - val_accuracy: 0.7799\n"
     ]
    }
   ],
   "source": [
    "history=model_embed.fit(X_train,y_train,batch_size=4,epochs=15,validation_split=0.2,verbose=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "58187df8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0YAAAEWCAYAAABCNYfGAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8rg+JYAAAACXBIWXMAAAsTAAALEwEAmpwYAABoKUlEQVR4nO3dd3yV5f3/8dcnJ3sPSAJJIOwlChKDgFrRuqoVd8EOR1u77LDDn3bafru+nbbVb79faq3WWtHiKNY9W0UUggLKEAIywk4YYWVfvz/uO+QQEgiQc+6M9/PxuB/n3PtzGLnzOdd1fS5zziEiIiIiItKbxQQdgIiIiIiISNCUGImIiIiISK+nxEhERERERHo9JUYiIiIiItLrKTESEREREZFeT4mRiIiIiIj0ekqMRDrIzIrNzJlZbAeOvd7MXo9GXCIiIp2hs55zx3Idka5EiZH0SGa21szqzKxPq+3v+D+siwMKTURE5ITpOSfS+ZQYSU/2ATCjecXMxgLJwYXTNegbPBGRHkPPOZFOpMRIerIHgE+FrV8H/DX8ADPLMLO/mtl2M1tnZt81sxh/X8jMfmVmlWa2Bri4jXP/bGabzWyjmf3YzEIdCczM/mFmW8xst5n9x8zGhO1LMrNf+/HsNrPXzSzJ33eGmb1hZrvMbIOZXe9vf9XMPhN2jUO6OPjfHn7JzFYBq/xtv/OvUW1mC83szLDjQ2b2bTNbbWZ7/P1FZna3mf261WeZY2a3dORzi4hIp+qyz7lW1+nvPyt2mFm5mX02bF+pmZX5z6KtZvYbf3uimf3NzKr8Z94CM8s71nuLHAslRtKTvQmkm9ko/wf5dOBvrY75A5ABDAY+hPeAucHf91ngEmA8UAJc1erc+4AGYKh/zPnAZ+iYZ4BhQC7wNvBg2L5fAROAyUA2cCvQZGYD/fP+APQFxgGLOng/gMuAicBof32Bf41s4O/AP8ws0d/3dbxvIT8CpAM3AvuB+4EZYQ/VPsCH/fNFRCS6uvJzLtwsoALo79/jp2Z2jr/vd8DvnHPpwBDgEX/7dX7cRUAO8HngwHHcW6TDlBhJT9f8bdp5wHJgY/OOsIfI7c65Pc65tcCvgU/6h1wD3Omc2+Cc2wH8LOzcPLyk4WvOuX3OuW3Ab/3rHZVz7l7/nrXAHcAp/jdzMXhJyFedcxudc43OuTf8464FXnTOPeScq3fOVTnnFh3Dn8XPnHM7nHMH/Bj+5l+jwTn3ayABGOEf+xngu865951nsX/sfGA3cK5/3HTgVefc1mOIQ0REOk+XfM6FXacImAL8P+dcjf/cuoeWlq56YKiZ9XHO7XXOvRm2PQcY6j8LFzrnqo/l3iLHSmMNpKd7APgPMIhW3QuAPkAcsC5s2zqgwH/fH9jQal+zgf65m82seVtMq+Pb5D+ofgJcjdfy0xQWTwKQCKxu49SidrZ31CGxmdk3gU/jfU6H1zLUPIj3SPe6H/gE8IL/+rsTiElERE5Ml3vOtdIf2OGc29PqPiX++08DPwJWmNkHwA+dc//yP1cRMMvMMvFawr7jnKs/xvuLdJhajKRHc86twxuc+hHgsVa7K/G+kRoYtm0ALd+2bcb7oRy+r9kGoBbo45zL9Jd059wYju5aYBpeF7QMoNjfbn5MNXjdCVrb0M52gH0cOuA2v41jXPMbfzzRrXjfFmY55zLxWoKan35HutffgGlmdgowCniineNERCTCuuhzLtwmINvM0tqKwTm3yjk3A69r+X8Ds80sxe8Z8UPn3Gi8ruWXcOh4KpFOp8RIeoNPA+c45/aFb3TONeL1Zf6JmaX5Y3i+Tkv/7EeAr5hZoZllAbeFnbsZeB74tZmlm1mMmQ0xsw91IJ40vIdNFV4y89Ow6zYB9wK/8Qerhsxskpkl4I1D+rCZXWNmsWaWY2bj/FMXAVeYWbKZDfU/89FiaAC2A7Fm9n28FqNm9wD/ZWbDzHOymeX4MVbgjU96AHi0uWueiIgEpqs958Jj2AC8AfzML6hwsh/v3wDM7BNm1td//u3yT2sys6lmNtbvZVGNl+A1HX4Hkc6jxEh6POfcaudcWTu7v4zX2rIGeB2viMC9/r4/Ac8Bi/EKJLT+Ju5TQDywDNgJzAb6dSCkv+J1I9jon/tmq/3fBN7FSz524H2DFuOcW4/3jeA3/O2LgFP8c34L1AFb8bq6PciRPQc8C6z0Y6nh0O4Rv8F7YD6P90D6M5AUtv9+YCxeciQiIgHqgs+51mbg9Y7YBDwO/MA596K/70JgqZntxeuaPd3/wi3fv1813tipf6NnjkSYOeeOfpSISBgzOwvv276BTj9EREREpAdQi5GIHBMziwO+CtyjpEhERER6CiVGItJhZjYKrw94P+DOQIMRERER6UTqSiciIiIiIr2eWoxERERERKTX6zETvPbp08cVFxcHHYaISK+3cOHCSudc36Dj6Ir0rBIRCV57z6kekxgVFxdTVtZepUoREYkWM1sXdAxdlZ5VIiLBa+85pa50IiIiIiLS6ykxEhERERGRXk+JkYiIiIiI9HoRHWNkZhcCvwNCeJNB/rzV/gHA/UCmf8xtzrmn/X23A58GGoGvOOeei2SsIiIiIiI9XX19PRUVFdTU1AQdSsQlJiZSWFhIXFxch46PWGJkZiHgbuA8oAJYYGZznHPLwg77LvCIc+6PZjYaeBoo9t9PB8YA/YEXzWy4c64xUvGKiIiIiPR0FRUVpKWlUVxcjJkFHU7EOOeoqqqioqKCQYMGdeicSHalKwXKnXNrnHN1wCxgWqtjHJDuv88ANvnvpwGznHO1zrkPgHL/eiIiIiIicpxqamrIycnp0UkRgJmRk5NzTC1jkUyMCoANYesV/rZwdwCfMLMKvNaiLx/DuZjZTWZWZmZl27dv76y4RURERER6rJ6eFDU71s8ZdPGFGcB9zrlC4CPAA2bW4ZicczOdcyXOuZK+fbvoXIJNTbB4FuzdFnQkIiLSTVXX1POHl1bx3sbdQYciItJjRTIx2ggUha0X+tvCfRp4BMA5Nw9IBPp08NzuoezP8Pjn4MGrof5A0NGIiEg3ZMD/vLqaB9/S3Lki0r1VVVUxbtw4xo0bR35+PgUFBQfX6+rqjnhuWVkZX/nKVyIWWyQTowXAMDMbZGbxeMUU5rQ6Zj1wLoCZjcJLjLb7x003swQzGwQMA+ZHMNbI2LUeXrwD+oyAzYthzpfBuaCjEhGRbiYtMY6PjO3Hk4s3c6BOdYhEpPvKyclh0aJFLFq0iM9//vPccsstB9fj4+NpaGho99ySkhJ+//vfRyy2iCVGzrkG4GbgOWA5XvW5pWb2IzO71D/sG8BnzWwx8BBwvfMsxWtJWgY8C3yp21Wkcw6e/Jr3+vF/wDnfhXf/AXN/F3RkIiLSDV1dUsje2gaeeW9z0KGIiHSq66+/ns9//vNMnDiRW2+9lfnz5zNp0iTGjx/P5MmTef/99wF49dVXueSSSwC44447uPHGGzn77LMZPHhwpyRMEZ3HyJ+T6OlW274f9n4ZMKWdc38C/CSS8UXU4lmw+iW46JeQNRDO/AZsXeq1IOWOguEXBB2hiIh0IxMHZTMwJ5lHyjZwxamFQYcjIj3AD59cyrJN1Z16zdH90/nBR8cc83kVFRW88cYbhEIhqquree2114iNjeXFF1/k29/+No8++uhh56xYsYJXXnmFPXv2MGLECL7whS90eM6itkQ0Meq19myFZ2+DotPhtM9428xg2t1QVQ6PfgY+8yL0HRFsnCIi0m2YGVdPKORXz69kXdU+BuakBB2SiEinufrqqwmFQgDs3r2b6667jlWrVmFm1NfXt3nOxRdfTEJCAgkJCeTm5rJ161YKC4//iyMlRpHwzLe8QgvT7oKYsN6K8ckw4yGYeTY8NAM++xIkZQUWpoiIdC9XTijk1y+sZPbCCr5xvr5cE5ETczwtO5GSktLyZc/3vvc9pk6dyuOPP87atWs5++yz2zwnISHh4PtQKHTE8UkdEXS57p5n2RxY9k84+/9Bn2GH788ohI/9zSvMMPtGaDyxv0AREek9+mUkcdawvjy6sILGJhXzEZGeaffu3RQUeFOY3nfffVG7rxKjzrR/Bzz1Dcg/GSYfoZTggNPhkt/A6pfhxR9ELz4REen2ri4pZNPuGuaWVwYdiohIRNx6663cfvvtjB8//oRbgY6FuR5SPrqkpMSVlZUFG8QTX/SKLtz0KvQ7+ejHP30rzP8/uOyPMO7aiIcnIhINZrbQOVcSdBxdUWc8q2obGpn405c4Y2gf7rr21E6KTER6i+XLlzNq1Kigw4iatj5ve88ptRh1lvIXYdGDcMbXOpYUAVzwExh0Fjz5VdiwIKLhiYhIz5AQG+KycQU8v3Qru/YfeTJEERHpOCVGnaF2jzdnUZ/hcNatHT8vFAdX3w/p/eHhj0P1poiFKCIiPcfVJYXUNTbxz0V6boiIdBYlRp3hpR/B7gq49C6ISzy2c5OzYcYsqNsHsz7uVbMTERE5gjH9MxjTP51HyjYEHYqISI+hxOhErZsH82fCxM/BgInHd43cUXDFTNj0Nsz5CvSQcV8iIhI515QUsXRTNUs37Q46FBGRHkGJ0Ymor4E5N0PmADjneyd2rZEXw9TvwruPwBu/75z4RESkx5o2rj/xoRj+UVYRdCgiIj2CEqMT8e+fQ1U5fPT3kJB64tc765sw+jJ44Qew8vkTv56IiPRYmcnxnDcmjycWbaS2oTHocEREuj0lRsdr0yKY+3sY/wkYMrVzrmkGl/0P5J8Ej34atq/snOuKiEiPdE1JEbv21/Pism1BhyIi0iFTp07lueeeO2TbnXfeyRe+8IU2jz/77LOJ1pQ8SoyOR2M9/PNmSOkL5/+kc68dnwLTH4JQPMyaAQd2de71RUSkxzhjaB/6ZSSqCIOIdBszZsxg1qxZh2ybNWsWM2bMCCiiFkqMjsfcO2Hru3DxryEps/Ovn1kEH/sb7FwHs2+EJnWREBGRw4VijKsmFPLaqu1s3q2qpiLS9V111VU89dRT1NV587CtXbuWTZs28dBDD1FSUsKYMWP4wQ9+EEhssYHctTvb/j78+xcw5nIYdUnk7jNwElz8K2/y1xe+700GKyIincrMLgR+B4SAe5xzP2/jmGuAOwAHLHbOXetvbwTe9Q9b75y7NCpBt3LVhEL+8HI5j729kS9NHRpECCLSXT1zG2x59+jHHYv8sXDRYT9KD8rOzqa0tJRnnnmGadOmMWvWLK655hq+/e1vk52dTWNjI+eeey5Llizh5JNP7tzYjkItRseiqdHrQhefAhf9MvL3m3A9nPZZmHcXLHoo8vcTEelFzCwE3A1cBIwGZpjZ6FbHDANuB6Y458YAXwvbfcA5N85fAkmKAAbmpDBxUDaPlG3AaboHEekGwrvTNXeje+SRRzj11FMZP348S5cuZdmyZVGPSy1Gx2L+TKiYD5fPhNS+0bnnhT+D7Su8lqM+w6CwJDr3FRHp+UqBcufcGgAzmwVMA8Kfxp8F7nbO7QRwznXJKgfXlBTxjX8sZv4HO5g4OCfocESkuzhCy04kTZs2jVtuuYW3336b/fv3k52dza9+9SsWLFhAVlYW119/PTU1NVGPSy1GHbVzLbz0Ixh2Ppx8TfTuG4qDa/4Kafkw6+NQvTl69xYR6dkKgPCqBRX+tnDDgeFmNtfM3vS73jVLNLMyf/tl7d3EzG7yjyvbvn17pwUf7qKx+aQmxPKI5jQSkW4gNTWVqVOncuONNzJjxgyqq6tJSUkhIyODrVu38swzzwQSlxKjjnDOa7GxEFzyW6+sdjQlZ8OMh6B2Dzz8cW9iWRERiYZYYBhwNjAD+JOZZfr7BjrnSoBrgTvNbEhbF3DOzXTOlTjnSvr2jUxvg+T4WD56Sj+efncze2sbInIPEZHONGPGDBYvXsyMGTM45ZRTGD9+PCNHjuTaa69lypQpgcSkrnQd8c7fYM2rcPFvIKMwmBjyxsAVM73E6MmvwuX/G/0ETUSkZ9kIFIWtF/rbwlUAbznn6oEPzGwlXqK0wDm3EcA5t8bMXgXGA6sjHnU7ri4p4qH5G3hqySY+dtqAoMIQEemQyy677JBxkffdd1+bx7366qvRCQi1GB1d9WZ47jsw8AyYcEOwsYy6BKZ+B5bMgjf+EGwsIiLd3wJgmJkNMrN4YDowp9UxT+C1FmFmffC61q0xsywzSwjbPoVDxyZF3fiiTIbmpqo7nYjIcYpoYmRmF5rZ+2ZWbma3tbH/t2a2yF9WmtmusH2/MLOlZrbczH5vFkDziHPw1DegsRYu/T3EdIE88qxvwehp8OIPYNWLQUcjItJtOecagJuB54DlwCPOuaVm9iMza64y9xxQZWbLgFeAbznnqoBRQJmZLfa3/9w5F2hiZGZcPaGQhet2Ur5tb5ChiIh0SxH7Tb8jZVCdc7c0lzoF/gA85p87Ge/bt5OBk4DTgA9FKtZ2LX0c3n/Ka6XJabPrePSZwWV/hNwx3uSvlauCjkik+2qs974AkV7LOfe0c264c26Ic+4n/rbvO+fm+O+dc+7rzrnRzrmxzrlZ/vY3/PVT/Nc/B/k5ml1+agGhGOMfCzcc/WAR6bV6S2n/Y/2ckRxj1JEyqOFmAM3T3DogEYgHDIgDtkYw1sPtq4KnvwX9T4XTvxjVWx9VfArM+DvMnAoPTYfPvARJmUFHJdL11e2DD16D8he9ZecHgEFsIsQmdPC1g8fGp0BqLqTmQUpfr8JkV1W3D/ZuhT1bvddh50N8ctBRyXHITUtk6ohcHnt7I986fwSxoS7Q00FEupTExESqqqrIyckhiA5Z0eKco6qqisTExA6fE8nEqK0yqBPbOtDMBgKDgJcBnHPzzOwVYDNeYnSXc255G+fdBNwEMGBAJw80fe52qNkN0+6GUBesUZE5AD72ANz/UXj4E/Dx2RDX8b94kV7BOW8esOZEaN0b0FgHcckw6Cw4ZQa4RmiogYba9l8P7Gh/f2Ndx2JJzvGSpOZk6eBrq/dJWZ1TWKWxAfZt9xKdQ5Zth7/Wtep29aX50HfEiccggbimpJAXl2/l3yu3c+6ovKDDEZEuprCwkIqKCiI1fUBXkpiYSGFhxwundZXf+KcDs51zjQBmNhSv/3bzJ3nBzM50zr0WfpJzbiYwE6CkpKTz2gRXPg9LHoYP3QZ5o49+fFAGTva61T32WXjsM3D1/RATCjoqkWDV7IY1//aToZeg2h+I3ncUTPwcDP0wDJjktep0hqYmbxxi/YFDk6a6vW0kIf779fO814Y2Su/HxLWTQIUlTwlpsL+ypYWnrYRnfxVe43sriRkt1+k/vu17ZBV3zp+NBGLqyFz6pMbzSNkGJUYicpi4uDgGDRoUdBhdUiQTo46UQW02HfhS2PrlwJvOub0AZvYMMAl4rY1zO1dNNfzra5A7Gs78RsRvd8JOvgb2VXotXE99HS65U2W8pXdpaoKt73qJ0KoXYcNbXitQQjoMPhs+dCsMPTdypfZjYiAmCeKSju0856C2+vCkKfy1ugI2LvRaftpKcpqFEloSm6xiKCo9PJlKy4OUXLUs9wJxoRiuOLWQe1//gMq9tfRJ7aQvAUREerhIJkYHy6DiJUTT8SbBO4SZjQSygHlhm9cDnzWzn+F1pfsQcGcEY23x4g9gz2a45gGIjY/KLU/YpC96vzi9/hvvF59zvhN0RHKsDuyEef8DuzdAn2HQZ7i3ZA/u2mNTgrJ/B6x+uaVVaN82b3u/U+CMr3mtQoWnde0/OzOv9SYxw/s7P5LGBq8FqDlpqq32xi01Jz+JGfpCRA5x9YRCZv5nDU+8s5HPnDk46HBERLqFiCVGzrkGM2sugxoC7m0ugwqUNVf8wUuYZrlDy0bMBs4B3sX7mvRZ59yTkYr1oA9eg7J7YdLNUDgh4rfrVOd+30uO/vML7xemiTcFHZF0RN1+mP9/8PpvvdbK1DxY/FDL/phYyBrkJ0phCVOfYb2r4EZTI2x8u2Ws0MaFgPPG4ww510uEhpzjtYr0RKFY77P11M8nnW5YXhrjijJ5pGwDnz5jUI8eYC0i0lkiOsbIOfc08HSrbd9vtX5HG+c1Ap+LZGyHqdsPc77s/RI6tRu2uJh53ej274BnboWUHDjpyqCjkvY01sM7D8Cr/w17t8CwC7zkNv8kL0GqWuWVYq9c6S+rYNXz0FTfco3UvFYJk/+aXtg15tw6Xk1NXhey5s9dscBrHTqwEzAoLIGzb/OSof7jNa5OpB3XlBTx7cffZUnFbk4pygw6HBGRLq+rFF8I3qs/9Ur3Xvev7lumNhQLV/0ZHrgCHvuc/236OUFHJeGammDZE/Dyj2HHaiiaCFf/xSuk0SwxHQomeEu4xgbYtQ62v9+SNFSuhPce9QoONItLhpyhh7Yu9RkO2YO8EtJdRf0BqFod9lmaP1c5NBxoOS41D4Zf5I0TGnIOJGcHF7NIN3LJKf340b+W8kjZBiVGIiIdoMQIYNMimHc3TLgBBp0ZdDQnJi4JZjwE910Msz4B1z95+C/YEozVL8OLP4TNi7wKaTNmwfALOz42JBTrTTScMwT4SMt257wCHAdbl/ylYr6XNIUP2o9PPXK1s+YlpU/njM9xzhsbU7nST+jCWsF2rQ+LzbwS9H2GQ/FZh3YbTOmj8TMixyE9MY6PnNSPOYs28d2LR5MUr9ZVEZEjUWIEkDsKzvkenPaZoCPpHEmZ8IlH4c/nwYNXw43PHX1wt0TOxoVeQvTBvyFjAFz2v141wc7qAmYGqX29pXjKofvq9nstU9vf9wo7hFc927Yc1rx6aGtTy0U7OO9Ortcy2dTotWYdkpz5SdCBnS2XjU2CPkO97nDjrg1rzRrSfVtqRbqwq0uKeOydjTy3dAuXjS8IOhwRkS5NiRF485mc+fWgo+hcafnwySfgz+d7Xes+/Tyk9ws6qt6lchW89CNYPsdLMi78OZTc2Hnz53REfDLkj/WW9tTXeFXdmpOmPVsOLxtdtdp7baw9/PyQX70xfKLTlFwv4Rl92aFd+jKKuvf4J5FuZuKgbIqyk3ikbIMSIxGRo1Bi1JPlDIFPzIb7LoG/XQE3PO19uy+RtXsj/Pvn8M6DXtfGs2+HSV/yJuXsiuISvW5smQOOfJxzXuvSYfPubAEsLAEaqn9nIl1ETIxx9YQifvPCSjbs2E9RtlpmRUTao8Sop+s/HqY/6HWp+/t0+OTj6rIUKft3eGW358/0upaV3uRNEpzaN+jIOoeZ100zKRP6Dg86GhHpoCsnFPLbF1cye2EFt5yn/7siIu1Rn5beYPDZcMVM2PAWzL7Bq27Wk1WthvVvQvVmrwpcpNXtg9d+Db8bB2/8AcZcDl9eCBf9vOckRSLSbRVkJnHG0D7MXlhBU5M7+gkiIr2UWox6izGXe9XBnvoGPPlVmHZXz6v05RzM/xM8dzs0+clfKN4b19LcVSxzAGQObHmfmnf8Y14a6+Ht++Hfv/C6lQ2/CM79HuSN6bzPJCLSCa4pKeLLD73DG6urOGNYn6DDERHpkpQY9SanfQb2bvfGv6T0gfN+GHREnaduP/zra7DkYS9BKbnBq8K2a33L8v7TsG/7oeeFEiDzKIlT6wSyqQmWPubNRbTzAxgwCa75Kww4PWofV0TkWJw3Oo+MpDgeKdugxEhEpB1KjHqbs2/zKpDNvRNS+sLkm4OO6MTtWAMPfxK2LoWp3/XG9bTXClS3D3Y1J0zrDk2cNi+B/ZWHHh+beGiLU3qBV2VuyxLIHQ0zHobhF/S81jcR6VES40JMG9efWQs2sHt/PRnJnTBPmYhID6PEqLcxg4/8yutW9/x3vJajU6YHHdXxe/9ZeOwm73N9fDYM+/CRj49PgdyR3tKWun2HJkvhydPmRd6fW+YAuHwmjL2q8+YiEhGJsGtKivjrvHXMWbKJT54+MOhwRES6HCVGvVFMCK74kzfx5j+/5M2xM+y8oKM6Nk1NXpfAf/835J8MH3sAsopP/LrxKd6Ev7mj2t5fu9crwa2ESES6mTH90xnVL51/lG1QYiQi0gZVpeutYhPgYw96hQIe+RRsWBB0RB23fwf8/RovKRr3cW/y2s5IijoiIVVJkYh0S2bGNSWFLKnYzfLN1UGHIyLS5ajFqDdLTPe6n/35fPj71XDDs+13MesqNi+Bhz8B1Zvg4t9AyY0a3yPSBucc9Y2OA/WN1NQ3cqCukQP13lIT9v5Anb+/vpEDdU0cqG/kc2cNJislPuiPIBFw2bgCfvb0Cv5RVsH3Pzo66HBERLoUJUa9XWquN+nrvRfA367wWl8yCoOOqm2LHvIqzyVlw43PQmFJ0BGJRNX2PbW8uHwrC9ftZH9dAwfqGtkfntj4yU3zeuNxzFkTHxvD1SWFSox6qKyUeM4bnccTizZy20UjiY9VxxERkWZKjASyB8EnHoW/fAQeuBxufA6Ss4OOqkVDnTc30YJ7oPhMuOovmjhVeo21lft4ftkWnlu6lbfX78Q56JOaQFZyHEnxIRLjQmQmx9MvLnRwPSkuRFJ8DElx/nq8vy0uRGLY+4Pb/dfEuBChGLXA9nRXlRTy1LubeXnFVi48qV/Q4YiIdBlKjMSTPxZmzPISowevhuvmeIUIgla9yRsDVbEAJn8Fzv0BhPTPVk7M/roG/rloE+9v2cPJhRlMGJjFgOxkrAt0y3TO8d7Gaj8Z2sLKrXsBb+D8184dzgUn5TEiL61LxCrd01nD+pKfnsgjZRVKjEREwug3TGlRPAWuuhce+aSXjMyYBaEA57r44DWYfQPUH4Cr74cxlwUXi/QI66v288Cba3l4wQaqaxqIj43hvjeaAK8VZsLATCYMzGLCwCzG9M8gMS46hTbqG5tY8MEOnlu6hReWbWXT7hpiDEoHZfP9S0Zz/pg8CrOSoxKL9HyhGOPKCQX88dXVbK2uIS89MeiQRES6BCVGcqhRl8Ald8KTX4EnvgiX/1/7k6VGinMw7y544QeQPRiufwr6johuDNJjNDU5Xiuv5K9vrOXl97cRY8aFJ+Vz3aRiJgzMYtW2PSxct5OF63by9rqdPLd0KwDxoRjG+q1Jpw7wkqW+aQmdFtf+ugb+s7KS55du4aUV29h9oJ6E2BjOGt6Xr58/gnNG5pKtcT4SIVdPKOLuV1bz6NsVfPHsoUGHIyLSJSgxksNNuA72bYeX/ws2lsGgs7yxPcVnQFp+ZO9duxfm3AxLH4dRl8K0u73qeSLHqLqmnkcXVvDAvHWsqdxHn9R4vjx1KNdOHEh+Rss35CPz0xmZn87HJ3rzulTurT2YJC1ct5P73ljLzP+sAWBAdjIlA7M41W9VGp6Xdkxjcnbsq+Ol5Vt5bulWXlu1ndqGJjKT4/jwqDzOH5PHmcP6kByvH8sSecV9UigdlM0/yir4woeGqGumiAhKjKQ9Z34DUvNg+ZPw3mOw8D5ve84wL0EqPsNLltLyOu+elatg1sehahV8+Icw5asqxS3HbNXWPdw/by2Pvb2R/XWNjB+QyZ0fG8dFY/NJiD1617g+qQlcMCafC8Z4XwLUNjSydFM1C9d6idJ/VlXy2DsbAUhLiGXcgJbud+OKMklLPLT76YYd+3lh2VaeW7qFBWt30OSgIDOJGaUDOH9MHqXF2cSGVBlMou+akiK++Y/FLFy3k5LiLlRwR0QkIObcsZdz7fDFzS4EfgeEgHuccz9vtf+3wFR/NRnIdc5l+vsGAPcARYADPuKcW9vevUpKSlxZWVlnfwQBaGyALUtg7evesu4NqNvj7esz/NBEKTX3+O6x/El4/AsQG+9VnRv8oc6LX3q8hsYmXlqxjfvfWMsbq6uIj43hoyf357rJAzm5MLNT7+Wco2LnARau20nZuh0sXLeL97dU0+S8PH5EXholxVlkJcfz0vJtLPMn0hyZn8b5o/M4f0w+Y/qn9+hv6M1soXNO9fTb0JWeVfvrGjjtxy9y8cn9+MVVpwQdjohI1LT3nIpYYmRmIWAlcB5QASwAZjjnlrVz/JeB8c65G/31V4GfOOdeMLNUoMk5t7+9+3Wlh02P19gAWxaHJUrzwhKlEWGJ0hlHT5QaG+CVH8Prv4WCCXDNX7vuPErdWE19I3WNTaQnBlhMIwJ27Kvj4QUb+Nub69i46wD9MxL5+OkDmX5aETmpnTce6Gj21NSzeMPug8nSovW72FvXQMnALM4fnc/5Y/IYmNMFqjxGiRKj9nW1Z9X/m72EJ5dsYsF3PkxKgjqRiEjv0N5zKpI/BUuBcufcGj+AWcA0oM3ECJgB/MA/djQQ65x7AcA5tzeCccqxCsV6SUzBBK+7W2MDbF4Ma1/zEqUlD0PZn71j+45sSZIGnnHo/EP7KmH2jfDBv2HCDXDRf0Ns9H6Z7ekO1DXy6vvbePq9Lby8fCv76hrJTUtgWF4qw3LTWl5zU7vdZJ7vbdzN/W+s5Z+LN1HX0MSkwTl875LRfHhUbiDd0tIS4zhjWB/OGNYHgMYmx4H6RlL1i6Z0cdecVsjDZRt46t3NXFNSFHQ4IiKBiuRTuwDYELZeAUxs60AzGwgMAl72Nw0HdpnZY/72F4HbnHONrc67CbgJYMCAAZ0avByDUCwUTvCWM77WKlF6DRbP8iZnBeg7ykuS8k+Cf//SK/Iw7W4Y/4lAP0JHOOdYuqka52B0//QuORHmvtoGXnl/G0+/u5lXVmznQH0jOSnxTBtfwIDsZMq37WXVtr38o2wD++pa/jv1SY1naG4qw/O8RGmonzjlpMR3mS5fdQ1NPPPeZu5/Yy1vr99FUlyIqycUct3kYobnpQUd3iFCMaakSLqFUwdkMbhvCv8o26DESER6va7y5J4OzA5LfGKBM4HxwHrgYeB64M/hJznnZgIzweueEK1g5SgOS5TqD21RWvR3qN8HGQPg089B//FBR9wu5xzvb93Dk4s38eTizazf4fXmTEuIpXRQNqcPzmHSkBxG9QsuUdpTU8/LK7xk6NX3vUpnfVITuGpCIReNzW9zcL9zjk27a1i1dY+XLG3dy6pte3j87Y3sqW04eFxWchzDctMYmpfK8NxUhvmJU9+0hKglTFura3jwrfX8/a31VO6tpTgnme9fMporJxSSkdSzugaKRJuZcU1JET9/ZgVrtu9lcN/UoEMSEQlMJBOjjXiFE5oV+tvaMh34Uth6BbAorBveE8DptEqMpJsIxUFhibeccYuXKG1b5s1RlNC1vulv9kHlPj8Z2sSqbXsJxRiTh+Rw8zlDSYwLMW91FW+tqeKlFdsASEuMZaKfKJ0+OPKJ0u4D9by0fCtPv7uF/6zaTl1DE3npCcwoHcBFJ+VTUpx9xPubGQWZSRRkJnH2iJZxYM45tlbXsmrbHj9Z2suqrXv41+JNVNe0JEzpibEHk6ShuankpifS0NhEfWMTdY2O+gbv/cH1xqaD2w6uN+9vaLXe6vyKnQdodI6pI3L51KSBnDWsLzFdsLVOpLu6YnwBv3zufR4u28DtF40KOhwRkcBEMjFaAAwzs0F4CdF04NrWB5nZSCALmNfq3Ewz6+uc2w6cA3Sd0apyYkJx0K/rVUDatOsA/1ritQy9u3E3AKXF2fzXZSdx0Un59AkbzH/pKf0BrzXjzTVV/rKDF5d7iVJ6Yiylg3I4fXB2pyVKu/fX8/yyLTzz3hZeW7Wd+kZHv4xEPjFxIB8Zm8+pA7JOOGEwM/IzEsnPSOTMYS3jwZxzbN9bS3lzsrRtDyu37uW5pVuYtaD+qNeND8UQFzLiYmOIC8W0rIe89bjYGOL99eR4/5hYb/3Ck/ox/bQiivv0nuIFEj1Hq57qH3MNcAdehdTFzrlr/e3XAd/1D/uxc+7+qATdyXLTEzl/dB6PLNjALR8eTmLc0cvai4j0RBFLjJxzDWZ2M/Ac3gPnXufcUjP7EVDmnJvjHzodmOXCyuM55xrN7JvAS+b111kI/ClSsUrvVbm3lqff3cyTizexYO1OAE4uzOC7F4/i4pP70S8j6Yjn56UnMm1cAdPGFQCwZXcNb31QxbzVXrL04vKtAGQkxR3senf64GxG5ad3KInZsa+OF5Zt4el3tzC3vJKGJkdBZhLXTy7morH9GFeYGZXWEzMjNy2R3LREJg/tc8i+yr217NxXR7yf9BxMfPzEJjbGusw4JZFwfvXUuwmrnmpmc8Krp5rZMOB2YIpzbqeZ5frbs/EKBpXgJUwL/XN3RvtzdIZPTSrmmfe28OTiTVytsUYi0ktFdB6jaOpqJVCl69p9oJ7n3tvCk0s2Mbe8kiYHw/NSufSU/lxycv9ObZnYvPsAb63Z4SVKH1Sxrsobo5SRFHdI17uR+WkHE5zKvbU8v3Qrz7y3mTdWV9HY5BiQncxFY/O5eGw/xhZkKNGQLq27lOs2s0nAHc65C/z12wGccz8LO+YXwErn3D2tzp0BnO2c+5y//n/Aq865h450z676rHLOccGd/yE+NoYnbz5DP2NEpEcLoly3SJexv66BF5Zt5cnFm/nPyu3UNTYxMCeZL549lI+e0p8R+ZEZ69QvI4nLxhdw2XivRWnTrgO89UEVb67ewbw1VTy/zGtRykyOo7Q4m721Dby5poomB8U5yXzurMF8ZGy/Hj8hqEhAOlI9dTiAmc3F6/1wh3Pu2XbOLYhcqJFlZnxqUjHffeI93l6/iwkDs4IOSUQk6pQYSY9V29DIq+9v58nFm3hp+TYO1DeSn57IpyYN5NJx/QNpeemfmcTl4wu5fLw3ie3GXQd4K2yMUnxsDF+aOpSPjO3HyPw0JUMiwYsFhgFn4xUR+o+ZjT2WC3TK1BLN0yCk5UVsEuzLxxfw38+s4K/z1ioxEpFeSYmR9CiNTY43Vlfyz0WbeG7pFvbUNJCdEs+VEwq49JQCSgaeeIGCzlSQmcQVpxZyxamR+UVHRI6oI9VTK4C3nHP1wAdmthIvUdqIlyyFn/tqWzfplKklanbDPefAh+/wqntGQEpCLFeVFPK3N9fxnYtHkZuWGJH7iIh0VUqMpEdYs30vj75dwWNvb2Tz7hrSEmK54KR8Lj2lP5OH5Bw2j4+ICB2rnvoEMAP4i5n1wetatwZYDfzUzJqbVs7HK9IQGSk5kDMUNiyI2C0APnn6QP4ydy2z5m/gK+cOi+i9RES6GiVG0m3tqannqSWbmb2wgrJ1O4kx+NDwvnz34tGcOypXJWdF5Ig6WD31OeB8M1sGNALfcs5VAZjZf+ElVwA/cs7tiGjARRNh5XPgHESom+3gvqmcNbwvD761ji+cPYQ4fakkIr2IEiPpVpqaHPPWVDF7YQXPvLeZmvomhuamcttFI7lifAG56er6ISId55x7Gni61bbvh713wNf9pfW59wL3RjrGg4pKYdGDsGMN5AyJ2G2unzyQG+8r4/mlW7n45H4Ru4+ISFejxEi6hXVV+5i90Osqt3HXAdISY7ny1EKuLinilEKVrxaRXqCw1HvdMD+iidGHhucyIDuZ++etVWIkIr2KEiPpsvbWNvC031Vu/todmMGZw/py20UjOW90nrrKiUjv0nckJKTDhrdg3IyI3SYUY3zy9IH85OnlLN9czah+6RG7l4hIV6LESLqUpibHmx/4XeXe3cKB+kYG90nhWxeM4IpTC+iXkRR0iCIiwYiJgcLToCKyBRgAri4p5NcvvM9f563jZ1ccU3VyEZFuS4mRdAkbduxn9sIKHn27goqdB0hLiOWy8f25akIRpw7IVFc5ERHwCjC8+jOoqYbEyLXkZCbHc9m4Ap54ZyO3XTiSjOS4iN1LRKSrUGIkbdq5r44fPrmUbXtqSY6PJSUh5L3Gh0hOaPUatj81IZbk+BAp/mtCbEy7Sc2+2gaeeW8Lsxdu4M01Xle5KUP68K0LRnD+6HyS4tVVTkTkEEWlgIONZTDknIje6pOTBjJrwQb+sXADnzlzcETvJSLSFSgxksOsr9rP9X+ZT8XOA4wtzGDn/gPsr2tgX20j++sa2F/X2OFrhWLMS5TiY0lO8F/jQ8THxrBw3U721zUyMCeZb5w3nCsmFFKQqa5yIiLtKpgAmFeAIcKJ0Zj+GZxWnMUDb67jximDutTk2CIikaDESA7xzvqdfOb+MhqaHH/7zERKB2UfdkxTk+NAfSP7/GRpX62XLO2ra2B/bfNrA/vqGg9JqPbVNR7cvnN/HZec3I+rS4ooGZilrnIiIh2RmA55Y7wCDFHwqUnFfPmhd/j3yu1MHZkblXuKiARFiZEc9Ox7W/jaw++Qm5bIX244jSF9U9s8LibGSEmIJSUhFtKiHKSISG9XVArvzoamJq8gQwRdMCaf3LQE7p+3VomRiPR4mtJaAPjz6x/whQcXMjI/nce+OLndpEhERAJWNBFqq2H7iojfKj42hmsnDuDV97eztnJfxO8nIhKkoyZGZvZRM1MC1UM1NjnumLOU//rXMs4fncdDnz2dPqkJQYclIiLtKWqe6DU63emuLR1AbIzxwJvronI/EZGgdCTh+Riwysx+YWYjIx2QRM+Buka+8LeF3PfGWm6cMoj/+fgEVYITEenqsgZBSl+vAEMU5KYnctHYfjxStoH9dQ1RuaeISBCOmhg55z4BjAdWA/eZ2Twzu8nMNLqkG9u+p5bpM+fxwvKt/OCjo/n+R0cTUsUhEZGuzwwKS6PWYgRw/eSB7Klp4Il3NkXtniIi0dahLnLOuWpgNjAL6AdcDrxtZl+OYGwSIeXb9nLFH+fy/tY9/N8nJnDDlEFBhyQiIseiqBR2rIZ9lVG53akDshjTP52/zluLcy4q9xQRibaOjDG61MweB14F4oBS59xFwCnANyIbnnS2t9ZUceUf3+BAXSOzbprE+WPygw5JRESOVdFE77ViQVRuZ2ZcN6mYFVv2MP+DHVG5p4hItHWkxehK4LfOubHOuV8657YBOOf2A5+OaHTSqf65aCOf/PN8clLjefyLUxhXlBl0SCIicjz6j4OYuKh2p7t0XH8yk+P46zwVYRCRnqkjidEdwMERnmaWZGbFAM65l450opldaGbvm1m5md3Wxv7fmtkif1lpZrta7U83swozu6sDcUo7nHPc/Uo5X521iHEDMnnsC5Mpyk4OOiwRETlecUnQ75SoFWAASIwL8bGSIp5duoUtu2uidl8RkWjpSGL0D6ApbL3R33ZEZhYC7gYuAkYDM8xsdPgxzrlbnHPjnHPjgD8Aj7W6zH8B/+lAjNKOhsYmvv34u/zyufeZNq4/D3y6lMzk+KDDEhGRE1VUChsXQmN91G75idMH0uQcD76lViMR6Xk6khjFOufqmlf89x35zboUKHfOrfHPmQVMO8LxM4CHmlfMbAKQBzzfgXtJG/bWNvDp+8t4aP4GvjR1CL+9ZhwJsSrHLSLSIxSVQkMNbFkSvVtmJ3PuyFwemr+e2obGqN1XRCQaOpIYbTezS5tXzGwa0JEyOAXAhrD1Cn/bYcxsIDAIeNlfjwF+DXzzSDfwy4aXmVnZ9u3bOxBS77Fldw3X/O88Xi+v5OdXjOVbF4wkRuW4RUR6jsLmiV6jU4Ch2acmFVO5t45n3t0S1fuKiERaRxKjzwPfNrP1ZrYB+H/A5zo5junAbOdc89dPXwSeds5VHOkk59xM51yJc66kb9++nRxS97V8czWX/89c1lXt497rT2N66YCgQxIRkc6WUQAZRVEtwABwxtA+DO6Twv3z1kb1viIikRZ7tAOcc6uB080s1V/f28FrbwSKwtYL/W1tmQ58KWx9EnCmmX0RSAXizWyvc+6wAg5yqP+s3M4XH3yb1IRY/vH5yYzunx50SCIiEilFpbA+uolRTIzxyUkD+eGTy1hSsYuTCzOjen8RkUjp0ASvZnYxXivO183s+2b2/Q6ctgAYZmaDzCweL/mZ08a1RwJZwLzmbc65jzvnBjjnivG60/1VSdHRPbJgAzfet4DCrCQe/5KSIhGRHq+wFKorYPcRO1h0uisnFJIcH1LpbhHpUToywev/Ah8DvgwYcDUw8GjnOecagJuB54DlwCPOuaVm9qPwMUt4CdMsp6m0j5tzjl8//z63PrqESUNy+MfnJ9EvIynosEREJNKKmscZRa9sN0B6YhxXnFrAnMWb2LGv7ugniIh0Ax1pMZrsnPsUsNM590O8bm7DO3Jx59zTzrnhzrkhzrmf+Nu+75ybE3bMHUdqDXLO3eecu7kj9+uN6hqa+Poji/nDy+V8rKSIe68/jbTEuKDDEhGJKjP7qj/3nZnZn83sbTM7P+i4Ii5/LMQmQUV0CzCAV4ShrqGJhxdsOPrBIiLdQEcSo+ZZ3PabWX+gHugXuZCko5qaHJ++fwGPv7ORb54/nJ9fOZa4UId6R4qI9DQ3OueqgfPxumd/Evh5sCFFQSgOCiZEvQADwPC8NCYPyeFvb66jsUmdPkSk++vIb9FPmlkm8EvgbWAt8PcIxiQdtGxzNa+tquT2i0Zy8znDMFM5bhHptZp/AH4EeMA5tzRsW89WVAqbF0P9gajf+lOTitm46wAvLd8a9XuLiHS2IyZG/nxCLznndjnnHsUbWzTSOdeR4gsSYXPLvemkLh/f5vRQIiK9yUIzex4vMXrOzNKApoBjio6iUmhqgE3vRP3WHx6VS/+MRBVhEJEe4YiJkXOuCbg7bL3WObc74lFJh7xeXsnwvFRy0xODDkVEJGifBm4DTnPO7QfigBuCDSlKDk70Gv3udLGhGD5++kBeL6+kfNueqN9fRKQzdaQr3UtmdqWpn1aXUlPfyIK1O5g8pE/QoYiIdAWTgPedc7vM7BPAd4He8UVeSg7kDIUN0S/AADD9tCLiQzFqNRKRbq8jidHngH8AtWZWbWZ7zKw6wnHJUby9fic19U2cMVSJkYgI8Ee8IkGnAN8AVgN/DTakKCqa6LUYBTDzRU5qApec0o9HF1awp6Y+6vcXEeksR02MnHNpzrkY51y8cy7dX9fMoQGbW15JKMaYODg76FBERLqCBn8+vGnAXc65u4G0gGOKnqJS2F8JO9YEcvvrJhWzr66Rx97eGMj9RUQ6Q+zRDjCzs9ra7pz7T+eHIx31enkV44oyNWeRiIhnj5ndjlem+0y/eFDv+QFZGDbRa86QqN/+lKJMTinK5P55a/nUpIGqkioi3VJHutJ9K2z5HvAkcEcEY5Kj2H2gnncrdjFF3ehERJp9DKjFm89oC1CIN81E79B3JCSkB1KAodl1kwayZvs+5pZXBRaDiMiJ6EhXuo+GLecBJwE7Ix+atOfNNVU0OTS+SETE5ydDDwIZZnYJUOOc6z1jjGJioPA0qAimAAPAR8b2IyclnvvnrQ0sBhGRE9GRFqPWKoBRnR2IdNzc8kqS40OMK8oMOhQRkS7BzK4B5gNXA9cAb5nZVcFGFWVFE2HrUqgJpj5SYlyI6aVFvLR8Kxt27A8kBhGRE3HUxMjM/mBmv/eXu4DXgLcjH5q05/XySiYOyiY+9njyWhGRHuk7eHMYXeec+xRQitf9+4jM7EIze9/Mys3stjb2X29m281skb98JmxfY9j2OZ36aY5HUSngYGNZYCF8fKI3vujBt9YHFoOIyPE6avEFIPwnbAPwkHNuboTikaPYtOsAa7bv49rSAUGHIiLSlcQ457aFrVdxlC//zCyEN4n5eXi9IRaY2Rzn3LJWhz7snLu5jUsccM6NO4GYO1fBBMC8AgxDzgkkhP6ZSZw/Oo+HF6znax8eRmJcKJA4RESOR0cSo9l4fbUbwXuQmFmyP7O4RNnc8koAFV4QETnUs2b2HPCQv/4x4OmjnFMKlDvn1gCY2Sy8ct+tE6PuITEd8sYEWoAB4FOTinnmvS08uXgTV5cUBRqLiMix6EhfrJeApLD1JODFyIQjRzO3vJI+qfGMyOs903OIiByNc+5bwEzgZH+Z6Zz7f0c5rQDYELZe4W9r7UozW2Jms80s/Df9RDMrM7M3zeyy9m5iZjf5x5Vt3769Q5/nuBWVQkUZNDVF9j5HcPrgbIbnpXL/vLW4ACacFRE5Xh1JjBKdc3ubV/z3yZELSdrjnGPu6iomD+lDTIzmiBARCeece9Q593V/ebyTLvskUOycOxl4Abg/bN9A51wJcC1wp5m1OYGQc26mc67EOVfSt2/fTgqrHUUTobYatq+I7H2OwMz41KRi3ttYzdvrdwUWh4jIsepIYrTPzE5tXjGzCcCByIUk7Vm1bS/b99SqTLeIiM/M9phZdRvLHjM7Wnm2jUB4C1Chv+0g51yVc67WX70HmBC2b6P/ugZ4FRh/gh/nxBU1T/QabHe6y8cXkJYQy19VultEupGOJEZfA/5hZq+Z2evAw0Bbg1Alwl5f5Y8vGqbESEQEwDmX5pxLb2NJc86lH+X0BcAwMxtkZvHAdOCQ6nJm1i9s9VJgub89y8wS/Pd9gCl0hbFJWYMgpa9XgCFAKQmxXFVSyNPvbmbbnppAYxER6aiOTPC6ABgJfAH4PDDKObcw0oHJ4eaWVzKoTwoFmUlHP1hERI7IOdeA90Xfc3gJzyPOuaVm9iMzu9Q/7CtmttTMFgNfAa73t48CyvztrwA/b6OaXfSZQWFp4C1GAJ88fSD1jY5Z8zcc/WARkS7gqFXpzOxLwIPOuff89Swzm+Gc+5+IRycH1Tc28eaaKi4/ta1xwSIicjycc0/Tqnqdc+77Ye9vB25v47w3gLERD/B4FJXC+0/BvkpICa6HweC+qZw1vC8PvrWOL5w9hLiQ5t4Tka6tIz+lPuuc29W84pzbCXw2YhFJmxZv2MW+ukamDFE3OhEROYKiid5rxYJg4wCumzSQrdW1zFm0KehQRESOqiOJUcjMDpZA8yfEi+/IxTswo/hvw2YNX2lmu/zt48xsnt99YYmZfayDn6fHer28EjOYNCQn6FBERKQr6z8OYuK6RHe6s0fkMrpfOrc+uoR7Xluj8t0i0qV1JDF6FnjYzM41s3PxJs975mgnhc0ofhEwGphhZqPDj3HO3eKcG+fPHP4H4DF/137gU865McCFeGVQMzv2kXqmueWVjC3IIDO5QzmpiIj0VnFJ0O+UwAswAIRijFmfO50Pj8rlx08t5+a/v8Pe2oagwxIRaVNHEqP/B7yMV3jh88C7HDrha3sOzijunKsDmmcUb88M/BnLnXMrnXOr/PebgG1AhCd/6Lr21TbwzvpdTFGZbhER6YiiUti4EBrrg46E9MQ4/vcTE7jtopE8895mpt31OuXb9gQdlojIYTpSla4JeAtYi5fsnINfrvQoOjqjOGY2EBiEl4C13leK13VvdQfu2SPN/2AHDU1O8xeJiEjHFJVCQw1sWRJ0JIA36evnPzSEBz9zOrsP1HPpXXP51xKNOxKRrqXdxMjMhpvZD8xsBV43t/UAzrmpzrm7OjmO6cBs51xjqxj6AQ8AN/gJWusYbzKzMjMr2759eyeH1HW8Xl5JQmwMEwZmBR2KiIh0B4XNE70GX4Ah3KQhOfzry2cyql86N//9HX745FLqGw97vIuIBOJILUYr8FqHLnHOneGc+wPQeITjWzvqjOJhpuN3o2tmZunAU8B3nHNvtnWSc26mc67EOVfSt2/P7Wk3t7yS04qzSYwLBR2KiIh0BxkFkFHUJQowtJafkchDnz2d6ycX85e5a5kx8022VmsSWBEJ3pESoyuAzcArZvYnv/CCHeH41o46oziAmY0EsoB5YdvigceBvzrnZh/DPXucbXtqWLFlj8YXiYjIsSkq7RIFGNoSHxvDHZeO4fczxrNsczUX//513lxTFXRYItLLtZsYOeeecM5NB0bizer9NSDXzP5oZucf7cIdnFEcvIRplju0huc1wFnA9WHlvMcd42frEeat9h4UU4aqTLeIiByDwlKoroDdFUFH0q5LT+nPE1+aQnpSLB+/5y1m/me1SnqLSGBij3aAc24f8Hfg72aWBVyNV6nu+Q6ce8QZxf31O9o472/A3452/d7g9VWVZCTFMaZ/RtChiIhId1LUPM5oPmQUBhvLEQzPS+OfX5rCrbOX8NOnV/D2ul388uqTSUuMCzo0EellOlKu+yDn3E5/XM+5kQpIWjjnmFteyeQhOYRijqUXo4iI9Hr5YyE2CSq6VgGGtqQlxvE/Hz+V73xkFC8s38q0u+aycqtKeotIdB1TYiTRtbZqP5t212h8kYiIHLtQHBRM6JIFGNpiZnz2rMH8/TMTqa5pYNpdc/nnovZqNomIdD4lRl3Y6+WVAJq/SEREjk9RKWxeDPUHgo6kwyYOzuGpr5zBSQXpfHXWIu6Ys5S6BpX0FpHIU2LUhc1dVUlBZhIDc5KDDkVERLqjolJoaoBN7wQdyTHJS0/k7589nU+fMYj73ljL9Jnz2Ly7+yR3ItI9KTHqohqbHG+sruSMoX0w0/giERE5Dgcneu0e3enCxYVi+N4lo7nr2vGs2LKHS37/Om+srgw6LBHpwZQYdVHvbdxNdU0DU4apG52IiBynlBzIGQobun4BhvZccnJ/5tw8hczkOD5xz1v88VWV9BaRyFBi1EU1jy+aPETzF4mIyAkomui1GHXjZGJobhr/vPkMLhrbj/9+dgWfe2Ah1TX1QYclIj2MEqMuam55JSPz0+iTmhB0KCIi0p0VlcL+StixJuhITkhqQix3zRjP9y4ZzcsrtjHtrrms2FIddFgi0oMoMeqCauobKVu3U9XoRETkxBWGTfTazZkZnz5jEA/ddDr7ahu47O65/PyZFbyyYhu7D6gFSUROTGzQAcjhytbupK6hSeOLRETkxPUdCQnpXne6cTOCjqZTnFaczb++cgb/b/YS/vTaGv7336sxg+G5aZQUZ3nLwGwKs5JUwEhEOkyJURf0enklcSGjtDg76FBERKS7i4mBwtOgovsWYGhLbloif7mhlP11DSxav4uydTtZsHYH/1y0iQffWg9AXnoCJcXZlAzM4rTibEbmpxEbUmcZEWmbEqMuaG55JeMHZJGSoL8eERHpBEUT4dWfQU01JKYHHU2nSo6PZfLQPkz2u583NjlWbKlm4bqdLFi7k4Vrd/DUks0ApMSHGD8giwl+ojRuQCapetaKiE8/DbqYnfvqeG/Tbm758PCgQxERkZ6iqBRwsLEMhpwTdDQRFYoxxvTPYEz/DD41qRiAjbsOULZ2x8Fk6fcvr8I5iDEY3T+dkoHZB7vf5WckBvsBRCQwSoy6mHlrqnAOpqjwgoiIdJaCCYB5BRh6eGLUloLMJArGFTBtXAEA1TX1vLN+FwvX7mDB2p08vGAD972x9uCxpxVnMaE4m/FFmQzLSyUhNhRg9CISLUqMupjXyytJTYjllMKMoEMREZGeIjEd8sZ4BRiE9MQ4PjS8Lx8a3heA+sYmlm+u9rrerdvB3NVVPLFoEwCxMcaQvqmM7p/O6H7pjOqXzqh+aeRoOg2RHkeJURczt7yS0wdna3CoiIh0rqJSeHc2NDV5BRnkoLhQDCcXZnJyYSafPmMQzjk27DjAuxt3s2zzbpZv3sO81VU8/s7Gg+fkpScwul86o/t7ydLofukU56QQE6MqeCLdlRKjLmTDjv2sq9rP9ZOLgw5FRER6mqKJUHYvbF8BeaODjqZLMzMG5CQzICeZi0/ud3D7jn11LN9czbJN1SzbXM3yzdX8Z1UljU0OgOT4ECPy0w62LI3un87I/DSS4/Xrlkh3oP+pXcgbqysBNLGriIh0vqLmiV7fUmJ0nLJT4pkytM8h44Br6hsp37aXZX7CtHxzNXMWt5QMN4NBOSmM8rviNbcy5aYlaI4lkS5GiVEX8np5FblpCQzNTQ06FBER6WmyBkFKX68AQ8kNQUfTYyTGhTipIIOTClrGBjvnqNh5wGtd8luWllTsOlg2HCA3LYHTBmVTWpx9cI4ldcMTCZYSoy6iqcnxRnklHxreV98giYhI5zODwlIVYIgCM6MoO5mi7GTOH5N/cHt1TT0rNu9h6abdvLN+FwvC5lhKT4ylxE+SSgdlMbYgk/hYjQUTiSYlRl3Eii17qNpXpzLdIiISOUWl8P5TsK8SUvS8ibb0xDhKB2VTOiibG6a0tCwtWLuD+R/sYP7aHby8YhsACbExjB+Q6bUoDcrmVE38LhJx+h/WRcwt98YXKTESEZGIKZrovVYsgBEXBRuLHNKydMWphQBU7q2lbO0O5n+wkwVrd3DXK+U0vdw8cW2636LktSxlp8QH/AlEepaIJkZmdiHwOyAE3OOc+3mr/b8FpvqryUCucy7T33cd8F1/34+dc/dHMtagvV5eyZC+KZpxW0REIqf/OIiJ87rTKTHqkvqkJnDhSf248CSvGt6emnreXr+LBX6L0gNvruPPr38AwNDc1INd704rzqYwKznI0EW6vYglRmYWAu4GzgMqgAVmNsc5t6z5GOfcLWHHfxkY77/PBn4AlAAOWOifuzNS8QaptqGR+R/s4JqSwqBDERGRniwuCfqd4hVgkG4hrdVktLUNjbxbsZv5fve7fy3exEPzvQp4BZlJnFacxdDcVOJjY4gPxRAfGyI+Noa4kJEQG+Nvb9kWHxvjbW+1rfl8jXuW3iSSLUalQLlzbg2Amc0CpgHL2jl+Bl4yBHAB8IJzbod/7gvAhcBDEYw3MO+s38WB+kZ1oxMRibIO9Gy4Hvgl0Dyz513OuXv8fd2zZ0NRqTefUWM9hOKCjkaOUUJsiJLibEqKs/ni2dDY5FixpZoFH+xgwdqdzF1dxROLNnXa/eJDhyZLfdMSmDgoh0mDcygdnE16ov4NSc8RycSoANgQtl4BTGzrQDMbCAwCXj7CuQVtnHcTcBPAgAEDTjzigLxRXkmMwelDcoIORUSk1+hIzwbfw865m1ud2317NhSVwpv/A1uWQMGEoKORE+SNPcpgTP8Mrp8yCOcc9Y2OusYm6hqaqPdfaxu817rGlm0Htzc2Ue+/1oUdV9d6W0MT63bsO9idL8ZgTP8MJg3J4fTB3rinNCVK0o11leIL04HZzrnGYznJOTcTmAlQUlLiIhFYNLxeXskpRZn61kVEJLqOtWdDuO7bs6GweaLXBUqMeiAzIz7Wa+EhITL3qKlv5J31u5i3poo311Rx39y1zPzPGmIMxhZkcPoQr0WppDibVFXSk24kkv9aNwJFYeuFtHRFaG068KVW557d6txXOzG2LqO6pp7FFbv54tlDgg5FRKS36WjPhivN7CxgJXCLc25DO+ce1rMBumDvhowCyCjyCjCc/vmgo5FuKDEuxKQhOUzye7ocqGvknfU7mbeminmrq7j39Q/4v3+vIRRjnFyYwaTBOZw+OIeS4iyS45UoSdcVyX+dC4BhZjYIL9GZDlzb+iAzGwlkAfPCNj8H/NTMsvz184HbIxhrYN5as4PGJqfxRSIiXdOTwEPOuVoz+xxwP3DOsVygS/ZuKCqF9ZroVTpHUnyIyUP7MNn/XWZ/XQML1+3kTT9RmvmfNfzPq6uJjTFOKco8mChNGJhFUnwo4OhFWkQsMXLONZjZzXhJTgi41zm31Mx+BJQ55+b4h04HZjnnXNi5O8zsv/CSK4AfNXdX6GnmlleSFBdi/IDMoEMREeltjtqzwTlXFbZ6D/CLsHPPbnXuq50eYaQUlsJ7j8LuCshQRVTpXMnxsZw5rC9nDvMq6e2rbaAsLFH6479Xc9cr5cSFjHHNidKQHMb0zyA5PkRsjKkangQiou2Zzrmngadbbft+q/U72jn3XuDeiAXXRbxeXslpg7JJiNU3JiIiUXbUng1m1s85t9lfvRRY7r/v3j0biprHGc1XYiQRl5IQe0jJ8b21DSxYu4M311Tx5uoq7nqlnN+/XH7w+BjzuuslxMaQEBsiMe7Q14Sw18R2XhNiYw5eIzEuREpCLPnpieSnJ5KeFKvES9qkjp4B2rK7hvJtezV/kYhIADrYs+ErZnYp0ADsAK73z+3ePRvyx0JsElQsgJOuCDoa6WVSE2KZOiKXqSNyAW8S2wVrd7B62z5qGxqpqW+itqGR2oYmauoPf91b20DV3jpqGhqpbT62vomahkbqG4/eWzUpLkR+RiJ56Qn0y0giLz2R/PQE8jOSyM9IpF9GIn1SEwjFKHnqbZQYBWhueSWAxheJiATkaD0bnHO3005LULfu2RCK8yrSbdA4IwleWmIc54zM45yRJ36txiZHXRsJ1Z6aerZU17Blt7/47xes3cHW6prDEqpQjNE3NYH8DK+VKT8j8eD7vHQvecrPSCQxTj1+ehIlRgGau7qS7JR4RuWnBx2KiIj0NkWl8Mbvof4AxCUFHY1IpwjFGEnxoWMq6tDU5Nixv+6QpGlrdQ2bd3uvq7fvZW55JXtqGw47NzM5jpyUeBJiQ8TFxhAfMuJCMQeX+NhW6837Y1ut+9tan58UHyI3LYG89ESykuPUBTDClBgFxDnH3PJKJg/JIUZNtSIiEm1FpdDUAJvegYGTg45GJDAxMUaf1AT6pCZwUkFGu8ftq21os9Vpx766gxPn1jc2Ud/g2FdXT31D2DZ/0t16fzLd5vVjER+KoW9awsFugLlpXutVXnoC+emJ5PrvUxM0hup4KTEKyOrte9laXcsZ6kYnIiJBODjR61tKjEQ6ICUhliF9UxnSN7VTrueco6HJHUym6hqbvOTJT6jqGpvYX9fI9j21bNldw9Y9NWyrrmVrdQ3vb9nDayvbbsVKjg+Rl55I7sEkKvFgq1OeX4AiNz1B3QDboMQoIK+v0vgiEREJUEoO5AyFDQuOfqyIdDozI87vOkf88V1jb20D26pr2Fpdy7Y9Xte/Lbtr/SSqhnfW72JLdQ11DYe3TmUkxdE/M4mCzCQKs5Lon5lIQWay95qVRJ+UhF7Xq0mJUUBeL69iYE4yRdnJQYciIiK9VdFEWPkcOAfqeiPS7aQmxJLaN5XBR2jFcs6x+0A9W/3Wpq3VNWzbU8vm3QfYtKuGDTv28+aaKva2an2Kj42hf0biweSpf2YSBVlJFPrv+2Um9rjpZpQYBaChsYk311Tx0VP6Bx2KiIj0ZkWlsOhB2LEGcoYEHY2IRICZkZkcT2ZyPCPy09o9bveBejbtOsDGnQfYtNt73bjLW/69cjvb9tQedk7ftAQK/MSpICuJ/hmJFGQlk5+eSFysETIjJsZ7DcW0vI+J4bBtoRgj5uArgYyTUmIUgMUVu9lb26DxRSIiEqzCsIlelRiJ9GoZSXFkJMUxql/b1ZJrGxrZsrvGS5Z2eq1NG3ftZ9OuGpZtruaF5Vvb7LJ3vEKtkqiYGDu47Y+fmEDpoOxOu1czJUYBmFteiRlMGpITdCgiItKb9R0JCeleAYZxM4KORkS6sITYEANzUhiYk9LmfucclXvr2LTrAFura2hocjQ2OZqc99ryHhqdo6n1fudobHQt+/xjDz/fkZN6nIOyjkKJUQDmllcypn862SmR+UsVERHpkJgYKDwNKlSAQUROjJnRNy2BvmkJQYdy3GKCDqC32V/XwNvrd6oanYiIdA1FE2HrUqipDjoSEZFAKTGKsvkf7KC+0Wl8kYiIdA1FpYCDjWVBRyIiEiglRlE2t7yS+NgYTivu/AFjIiIix6xgAsTEwis/g10bgo5GRCQwSoyi7PXyKiYMyNJswyIi0jUkpsPl/wfblsH/ToH3Hgs6IhGRQKj4QhRV7q1l+eZqvnXBiKBDEZEIqa+vp6KigpqamqBDibjExEQKCwuJi4sLOhQ5UWOvgoJT4dHPwuwbYNUL8JFfQEL7c56IiPQ0Soyi6I3VVQAqvCDSg1VUVJCWlkZxcXEgk9NFi3OOqqoqKioqGDRoUNDhSGfIHgw3Pgv//gW89itYPw+uvAcKS4KOTEQkKtSVLormrqokLTGWsQUZQYciIhFSU1NDTk5Oj06KwCvLmpOT0ytaxnqVUByc8x24/iloaoA/nw//+SU0NQYdmYhIxCkxihLnHK+XVzJ5SA6hmJ79C5NIb9fTk6JmveVz9koDJ8PnX4cxl8HLP4b7LlFhBhHp8ZQYRcn6HfvZuOuAynSLiEj3kJQJV/4ZLvtf2LIE/jgF3ns06KhERCJGiVGUvF5eCWh8kYhETlVVFePGjWPcuHHk5+dTUFBwcL2uru6I55aVlfGVr3wlSpFKt2EG42bA51+DvsNh9o3w+Begdk/QkYmIdLqIFl8wswuB3wEh4B7n3M/bOOYa4A7AAYudc9f6238BXIyXvL0AfNU55yIZbyTNLa+kf0Yig/qkBB2KiPRQOTk5LFq0CIA77riD1NRUvvnNbx7c39DQQGxs2z/2S0pKKCnRIHtpR/ZguOEZFWYQkR4tYomRmYWAu4HzgApggZnNcc4tCztmGHA7MMU5t9PMcv3tk4EpwMn+oa8DHwJejVS8kdTY5HhjdRUfHpWnPvkivcgPn1zKsk3VnXrN0f3T+cFHx3T4+Ouvv57ExETeeecdpkyZwvTp0/nqV79KTU0NSUlJ/OUvf2HEiBG8+uqr/OpXv+Jf//oXd9xxB+vXr2fNmjWsX7+er33ta2pNkpbCDEOmwmM3eYUZpt4OZ3wdYjQ3n4h0f5FsMSoFyp1zawDMbBYwDVgWdsxngbudczsBnHPb/O0OSATiAQPigK0RjDWilm2qZtf+eo0vEpFAVFRU8MYbbxAKhaiurua1114jNjaWF198kW9/+9s8+ujh40ZWrFjBK6+8wp49exgxYgRf+MIXNF+ReJoLMzz1da8wQ/nLcMVMyCwKOjIRkRMSycSoAAgvYVMBTGx1zHAAM5uL193uDufcs865eWb2CrAZLzG6yzm3vPUNzOwm4CaAAQMGdP4n6CTN44smD80JOBIRiaZjadmJpKuvvppQyPtGf/fu3Vx33XWsWrUKM6O+vr7Ncy6++GISEhJISEggNzeXrVu3UlhYGM2wpStrLsww7Hx46pteYYaP/hZOujLoyEREjlvQxRdigWHA2cAM4E9mlmlmQ4FRQCFegnWOmZ3Z+mTn3EznXIlzrqRv375RDPvYzC2vZEReGrlpiUGHIiK9UEpKy9jG733ve0ydOpX33nuPJ598st15iBISEg6+D4VCNDQ0RDxO6WbM4JTpKswgIj1GJBOjjUB4u3qhvy1cBTDHOVfvnPsAWImXKF0OvOmc2+uc2ws8A0yKYKwRs3r7Xhas3aFqdCLSJezevZuCggIA7rvvvmCDkZ4he5BXmOGsW2HJLPjfM6CiLOioRESOWSQTowXAMDMbZGbxwHRgTqtjnsBrLcLM+uB1rVsDrAc+ZGaxZhaHV3jhsK50XVlTk+O+uR/wkd+9RlJ8iKtL1AVFRIJ36623cvvttzN+/Hi1AknnaS7McP3T0NToFWb49y+99yIi3YRFsgK2mX0EuBNv/NC9zrmfmNmPgDLn3BzzSrT9GrgQaAR+4pyb5Ve0+x/gLLxCDM86575+pHuVlJS4srKu8Q3Vpl0HuHX2El4vr2TqiL7895Unk5uubnQivcHy5csZNWpU0GFETVuf18wWOudUx7kNXelZFTEHdsFT34D3ZsOASX5hhq47DlhEep/2nlMRncfIOfc08HSrbd8Pe++Ar/tL+DGNwOciGVskOOf456JNfO+f79HY5Pjp5WOZUVqkEt0iItJ7JGV6cxwNO88vzHAGnPIxyBkKOUMge4iXKHXFEt/OwYGdsHOtt6T1g6KJEBP0kGwRiYaIJka9yc59dXzniXd5+t0tTBiYxW+uOYWBOZrMVUREeqHmwgxFE73Wo0V/h7q9Lftj4iCrOCxZGuy95gyFtP6RTUQa62F3hZ/8fNCSBO1cCzvWQu3uQ49PzYfRl8Loy2DA6V0zoRORTqHEqBO8smIbtz66hF3767j1whF87qwhhGLUSiQiIr1c9iD45GNeS8zerVC1Gnashqpy//0aWPMKNIRVR4xNbEmUsoeEvQ6F1Fwv6TqaA7vaSXw+8JIiFzb2KSYOsgZ6iVphqfeaVey1alWuhKWPw9t/hfkzITUPRk9TkiTSQykxOgH7ahv48VPLeWj+ekbkpXHfDacxpn9G0GGJiIh0LWaQlu8txVMO3dfUBNUb/YTJT5aqymHbCnj/WWgKm2srPvXQ1qXsIdBYe2jis3Mt1Ow69B7JOZA1CApPg7FXtyQ/2YO87nLtJTj9ToaxV3klyFc+B8ueODRJGnUpjLlcSZJID6HE6DgtXLeDrz+ymPU79vO5swbz9fOHkxCrH4oiIiLHJCYGMou8ZfDZh+5rbIDdG1qSpuYWp03vwLJ/gmvyrxHntfBkFUPBBC/hOdjyMxAS008sxoQ0L0EaexXU7oWVz3pJ0jsPwII/hSVJl3kFJ6KZJO2rhK3vwZb3YOtS731TI+SN8Zb8kyBvLKTlRS8mkW5KidExqmto4rcvruT//r2a/plJPHzTJEoHZQcdloiISM8TivWSnOxBMPTDh+5rqINd6yE2HtILopeMJKQemiSteg6WPtEqSfqo35LUiUlSQx1UrfKSny3vtiRBe7e2HJOa7yVDMSFYNxfefaRlX0pfP1k6yVvyT4I+I7w/PxEBlBgdkxVbqrnl4cUs31zNx0qK+N5HR5OaoD9CEek6pk6dym233cYFF1xwcNudd97J+++/zx//+MfDjj/77LP51a9+RUmJqmtLNxMbD32GBhtDQiqcdKW3HJIkPQgL7oGU3JbCDQMndzxJ2rutVSvQUti+oqVbYSge+o6EIeeGtQqdBCmtJpPfv6MlgWq+3vw/ed0PAWJiveSo+fy8MZA/1hvLJdIL6bf6Dmhsctzz2hp+/fxK0pNiuedTJXx4tJqkRaTrmTFjBrNmzTokMZo1axa/+MUvAoyq6zKzC4Hf4c23d49z7uftHHclMBs4zTlXZmbFeBOPv+8f8qZz7vNRCFm6qsOSpOf97nZHSJIa6qDy/VatQEth37aW66b185KWoee2tPTkDPUm1T2a5GwYdKa3NGts8Lojhrc6ffAaLHm45ZiU3EO74eWNgT7Du2/rUlMT1O/zulyG4lV+XdqlxOgoNuzYzzceWcz8tTu4YEweP718LDmpCUGHJSLdwTO3eb98dKb8sXBRm7+7A3DVVVfx3e9+l7q6OuLj41m7di2bNm3ioYce4utf/zoHDhzgqquu4oc//GHnxtUN+ZOJ3w2cB1QAC8xsjnNuWavj0oCvAm+1usRq59y4aMQq3UxCKpx0hbfU7Wsp3BCeJKX09ZKipgbvnFAC5I705n9qbr3JOwlScjo3tlAs9B3hLWOvatm+f0erVqp34a2ZYa1LcV4rVd8RLeO3soq9in7R7MrYHue8z1BV3rKEj01rONBybHOCFBvvvYYSvEQz1n8NJRz6/uC++JYl/NzEdMgd5SWRnf33JVGlxKgdzjkeKdvAj55cRowZv776FK44tUCTtYpIl5adnU1paSnPPPMM06ZNY9asWVxzzTV8+9vfJjs7m8bGRs4991yWLFnCySefHHS4QSsFyp1zawDMbBYwDVjW6rj/Av4b+FZ0w5MeIT7l0CRp1fOwbI43r9Pw81vG/OQM9ZKWoCRnw6CzvKVZY4OXYIR3xatY4JUwb13yPKPw0GQpvPhFUlbHyqx3RO0eP9kpb6lg2LzUhM1BZaGWubIGneVVRGys95da77WhFhrrWpaG2kP311S32t/8Puz88D8HaGnha+6emD/WiyHoxFE6RIlRG7bvqeX2x5bw4vJtTBqcwy+vPpnCrOSgwxKR7uYILTuR1Nydrjkx+vOf/8wjjzzCzJkzaWhoYPPmzSxbtkyJERQAG8LWK4CJ4QeY2alAkXPuKTNrnRgNMrN3gGrgu86519q6iZndBNwEMGDAgM6KXbqj+BSvKMOYy4OOpGNCsV4rVu7IQ1uXGhugunmS3HUt5dJ3rYPlc2B/1aHXSciArAGHJktZg7wEKnOA1xoTrqHWK71+sNWnvCUZCi82AZBe6JVvP+mqljLuOUO963aku+GJampsaW072OL2njc/V3NrYGyi36LkJ0rNLYJJmZGPLxKaGmHfdtizGfZs9V+3eK97/fUDuyC9f6vWRX9Jzeu8RLmTKTFq5dn3tvDtx99lb20D37tkNDdMLiZGk7WKSDcybdo0brnlFt5++232799PdnY2v/rVr1iwYAFZWVlcf/311NTUHP1CvZyZxQC/Aa5vY/dmYIBzrsrMJgBPmNkY51x16wOdczOBmQAlJSUugiGLREcotuWX3LbU7mlJmHaFJU7bV8KqFw6d0BfzfoHOHAhxiV4CtHtDSyl2gOQ+XrIz9DzIGdyS/GQNgviAv7iOCUFqX0idCkOmtmxvHj/WnChteRfef9qrXtgsY0BYy5L/mjUouDFQTU2wv/LQRKetxGfv1kP/fgAwr3toWp7XatZnOFRv9savLZ4FhP3oi030k+PiNpaB3hcIAVFi5KuuqeeHc5bx6NsVnFSQzm+vGcewvLSgwxIROWapqalMnTqVG2+8kRkzZlBdXU1KSgoZGRls3bqVZ555hrPPPjvoMLuCjUBR2Hqhv61ZGnAS8KrfjTofmGNmlzrnyoBaAOfcQjNbDQwHyqIRuEiXlpDm/aKff9Lh+5qavF+sDyZMYYnTgZ3eJLynTPeTnyHeJL7dsWUlNt5rHcof27LNOS/BaE6UmluYVj7bkmjEp0Lu6JZEKcevvOgavZaapgZ/aWx5da23h60fcl749iaoP+BVQGxOfPZuPbxrIHiJaVo/L+nJP8l/n++9pvoTN6fmtt9C11ALuzb4f88ftPx971znlZWv23vo8Sm57SRNxf6EzJFLHJUYAe9v2cON9y1gS3UNXzlnKDefM4z4WFUsEZHua8aMGVx++eXMmjWLkSNHMn78eEaOHElRURFTpkwJOryuYgEwzMwG4SVE04Frm3c653YDB+sfm9mrwDf9qnR9gR3OuUYzGwwMA9ZEM3iRbikmBtL7ecuA04OOJrrMWj77sPNattcfgG3LD+2K9+6jUHZvJ98/5JVoj/FfQ/Fet7a0fK+rX3Oyk5YflvDknXg1wtgEr7R+W+X1m4tmHJY0rYX1b8J7sw9tnQrFe61N0+6KyL8fJUZA/8xEBvdN4a5rxzN+QFbQ4YiInLDLLrsM51q6Ltx3331tHvfqq69GJ6AuyDnXYGY3A8/hleu+1zm31Mx+BJQ55+Yc4fSzgB+ZWT3QBHzeObcj8lGLSI8TlwQFp3pLM+e8LoU71/oJTVhSczDBaU5y/PVDEp+wY5rP74rjesy8Sn4pOVA44fD9DXUtfw7h3TKTsiMSjhIjIC0xjgc+PfHoB4qISI/inHsaeLrVtu+3c+zZYe8fBR6NaHAi0nuZeQUkMnt5wZbYeL+oxpCo3E79xUREREREpNdTYiQi0snCu7D1ZL3lc4qISO+gxEhEpBMlJiZSVVXV45MG5xxVVVUkJiYGHYqIiEin0BgjEZFOVFhYSEVFBdu3bw86lIhLTEyksLAw6DBEREQ6hRIjEZFOFBcXx6BBg4IOQ0RERI6RutKJiIiIiEivp8RIRERERER6PSVGIiIiIiLS61lPqZxkZtuBdSd4mT5AZSeEE22KO7oUd3Qp7ujqjLgHOuf6dkYwPU0nPKt687+rICju6FLc0dVd44YTj73N51SPSYw6g5mVOedKgo7jWCnu6FLc0aW4o6u7xt1bdNe/H8UdXYo7uhR39EUqdnWlExERERGRXk+JkYiIiIiI9HpKjA41M+gAjpPiji7FHV2KO7q6a9y9RXf9+1Hc0aW4o0txR19EYtcYIxERERER6fXUYiQiIiIiIr2eEiMREREREen1lBgBZnahmb1vZuVmdlvQ8XSEmRWZ2StmtszMlprZV4OO6ViYWcjM3jGzfwUdy7Ews0wzm21mK8xsuZlNCjqmjjCzW/x/J++Z2UNmlhh0TG0xs3vNbJuZvRe2LdvMXjCzVf5rVpAxtqWduH/p/ztZYmaPm1lmgCG2qa24w/Z9w8ycmfUJIjY5lJ5T0afnVHTpORVZek51TK9PjMwsBNwNXASMBmaY2ehgo+qQBuAbzrnRwOnAl7pJ3M2+CiwPOojj8DvgWefcSOAUusFnMLMC4CtAiXPuJCAETA82qnbdB1zYatttwEvOuWHAS/56V3Mfh8f9AnCSc+5kYCVwe7SD6oD7ODxuzKwIOB9YH+2A5HB6TgVGz6ko0XMqKu5Dz6mj6vWJEVAKlDvn1jjn6oBZwLSAYzoq59xm59zb/vs9eD/4CoKNqmPMrBC4GLgn6FiOhZllAGcBfwZwztU553YFGlTHxQJJZhYLJAObAo6nTc65/wA7Wm2eBtzvv78fuCyaMXVEW3E75553zjX4q28ChVEP7Cja+fMG+C1wK6DqPF2DnlNRpudUIPSciiA9pzpGiZH3Q3pD2HoF3eQHdzMzKwbGA28FHEpH3Yn3j7kp4DiO1SBgO/AXv3vFPWaWEnRQR+Oc2wj8Cu9blc3Abufc88FGdUzynHOb/fdbgLwggzlONwLPBB1ER5jZNGCjc25x0LHIQXpORd+d6DkVNXpOdQl6TqHEqNszs1TgUeBrzrnqoOM5GjO7BNjmnFsYdCzHIRY4Ffijc248sI+u2Vx+CL+v8zS8B2Z/IMXMPhFsVMfHefMLdKtWDDP7Dl6XogeDjuVozCwZ+Dbw/aBjkZ5Dz6mo0nMqYHpORVakn1NKjGAjUBS2Xuhv6/LMLA7vYfOgc+6xoOPpoCnApWa2Fq87yDlm9rdgQ+qwCqDCOdf8jedsvAdQV/dh4APn3HbnXD3wGDA54JiOxVYz6wfgv24LOJ4OM7PrgUuAj7vuMWncELxfTBb7/0cLgbfNLD/QqETPqejScyr69JwKiJ5Th1JiBAuAYWY2yMzi8Qb7zQk4pqMyM8PrQ7zcOfeboOPpKOfc7c65QudcMd6f9cvOuW7xrZBzbguwwcxG+JvOBZYFGFJHrQdON7Nk/9/NuXSDwbhh5gDX+e+vA/4ZYCwdZmYX4nXFudQ5tz/oeDrCOfeucy7XOVfs/x+tAE71/+1LcPSciiI9pwKh51QA9Jw6XK9PjPxBZzcDz+H9J3zEObc02Kg6ZArwSbxvshb5y0eCDqoX+DLwoJktAcYBPw02nKPzvzmcDbwNvIv3/35moEG1w8weAuYBI8yswsw+DfwcOM/MVuF9q/jzIGNsSztx3wWkAS/4/z//N9Ag29BO3NLF6Dklx0jPqQjScyq6ov2csu7RaiYiIiIiIhI5vb7FSERERERERImRiIiIiIj0ekqMRERERESk11NiJCIiIiIivZ4SIxERERER6fWUGIlEgJk1hpWnXWRmnTbzuJkVm9l7nXU9ERHpffScEjlcbNABiPRQB5xz44IOQkREpB16Tom0ohYjkSgys7Vm9gsze9fM5pvZUH97sZm9bGZLzOwlMxvgb88zs8fNbLG/TPYvFTKzP5nZUjN73sySAvtQIiLSY+g5Jb2ZEiORyEhq1UXhY2H7djvnxuLNOH2nv+0PwP3OuZOBB4Hf+9t/D/zbOXcKcCrQPNv9MOBu59wYYBdwZUQ/jYiI9DR6Tom0Ys65oGMQ6XHMbK9zLrWN7WuBc5xza8wsDtjinMsxs0qgn3Ou3t++2TnXx8y2A4XOudqwaxQDLzjnhvnr/w+Ic879OAofTUREegA9p0QOpxYjkehz7bw/FrVh7xvReEEREek8ek5Jr6TESCT6Phb2Os9//wYw3X//ceA1//1LwBcAzCxkZhnRClJERHotPaekV1L2LhIZSWa2KGz9WedccynULDNbgvdt2gx/25eBv5jZt4DtwA3+9q8CM83s03jfuH0B2Bzp4EVEpMfTc0qkFY0xEokiv+92iXOuMuhYREREWtNzSnozdaUTEREREZFeTy1GIiIiIiLS66nFSEREREREej0lRiIiIiIi0uspMRIRERERkV5PiZGIiIiIiPR6SoxERERERKTX+/+IqhEhwoGeSQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1008x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "accuracy_loss_plot(history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "a9858563",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "82/82 [==============================] - 4s 48ms/step - loss: 0.4938 - accuracy: 0.7393\n",
      "Test set\n",
      "  Loss for model embed: 0.494\n",
      "  Accuracy: 0.739\n"
     ]
    }
   ],
   "source": [
    "accr = model_embed.evaluate(X_test,y_test)\n",
    "print('Test set\\n  Loss for model embed: {:0.3f}\\n  Accuracy: {:0.3f}'.format(accr[0],accr[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "2553265b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#evaluate\n",
    "import seaborn as sns\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "#convert probabilistic output to binary\n",
    "y_pred = (model_embed.predict(X_test) > 0.5).astype(\"int32\")\n",
    "#confusion matrix\n",
    "confusion_matrix = confusion_matrix(y_test, y_pred)\n",
    "sns.heatmap(confusion_matrix, linewidths = 1, fmt= 'g', annot=True);\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "id": "e841283a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Set Accuracy: \t 73.93\n",
      "Set Precision: \t 0.68\n",
      "Set Recall: \t 0.69\n",
      "Set F score: \t 0.68\n"
     ]
    }
   ],
   "source": [
    "#precision, recall, AUC, ROC Curve\n",
    "from sklearn.metrics import accuracy_score, precision_recall_fscore_support\n",
    "accuracy = accuracy_score(y_test, y_pred)\n",
    "metric = list(precision_recall_fscore_support(y_test, y_pred, average='macro'))\n",
    "print('Set Accuracy: \\t', round(accuracy * 100, 2))\n",
    "print('Set Precision: \\t', round(metric[0], 2))\n",
    "print('Set Recall: \\t', round(metric[1], 2))\n",
    "print('Set F score: \\t', round(metric[2], 2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "53c18638",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_embed.save('models/lstm_skill_extractor.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "585e6c4b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Layer lstm_1 will not use cuDNN kernels since it doesn't meet the criteria. It will use a generic GPU kernel as fallback when running on GPU.\n"
     ]
    }
   ],
   "source": [
    "saved_model = tf.keras.models.load_model('models/lstm_skill_extractor.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "fb309c69",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 2/82 [..............................] - ETA: 4s - loss: 0.4681 - accuracy: 0.7188 "
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-08-17 22:25:28.650426: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:112] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "82/82 [==============================] - 4s 44ms/step - loss: 0.4938 - accuracy: 0.7393\n",
      "Test set\n",
      "  Loss for model embed: 0.494\n",
      "  Accuracy: 0.739\n"
     ]
    }
   ],
   "source": [
    "#check that model is saved properly\n",
    "accr = saved_model.evaluate(X_test,y_test)\n",
    "print('Test set\\n  Loss for model embed: {:0.3f}\\n  Accuracy: {:0.3f}'.format(accr[0],accr[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e39cfcd1",
   "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": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}