10_Bag_of_N_Grams_News_Classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "KYoVrnewenmh"
   },
   "source": [
    "### Bag of N grams"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "KYoVrnewenmh"
   },
   "source": [
    "##### Let's first understand how to generate n-grams using CountVectorizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'thor': 5, 'hathodawala': 1, 'is': 2, 'looking': 4, 'for': 0, 'job': 3}"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "\n",
    "v = CountVectorizer()\n",
    "v.fit([\"Thor Hathodawala is looking for a job\"])\n",
    "v.vocabulary_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'thor hathodawala': 4,\n",
       " 'hathodawala is': 1,\n",
       " 'is looking': 2,\n",
       " 'looking for': 3,\n",
       " 'for job': 0}"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "v = CountVectorizer(ngram_range=(2,2)) # single two token ngrams\n",
    "v.fit([\"Thor Hathodawala is looking for a job\"])\n",
    "v.vocabulary_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'thor': 9,\n",
       " 'hathodawala': 2,\n",
       " 'is': 4,\n",
       " 'looking': 7,\n",
       " 'for': 0,\n",
       " 'job': 6,\n",
       " 'thor hathodawala': 10,\n",
       " 'hathodawala is': 3,\n",
       " 'is looking': 5,\n",
       " 'looking for': 8,\n",
       " 'for job': 1}"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "v = CountVectorizer(ngram_range=(1,2)) # single token and two token ngrams\n",
    "v.fit([\"Thor Hathodawala is looking for a job\"])\n",
    "v.vocabulary_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'thor': 12,\n",
       " 'hathodawala': 2,\n",
       " 'is': 5,\n",
       " 'looking': 9,\n",
       " 'for': 0,\n",
       " 'job': 8,\n",
       " 'thor hathodawala': 13,\n",
       " 'hathodawala is': 3,\n",
       " 'is looking': 6,\n",
       " 'looking for': 10,\n",
       " 'for job': 1,\n",
       " 'thor hathodawala is': 14,\n",
       " 'hathodawala is looking': 4,\n",
       " 'is looking for': 7,\n",
       " 'looking for job': 11}"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "v = CountVectorizer(ngram_range=(1,3)) # Single, Bigram and Trigrams.\n",
    "v.fit([\"Thor Hathodawala is looking for a job\"])\n",
    "v.vocabulary_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will not take a simple collection of text documents, preprocess them to remove stop words, lemmatize etc and then generate bag of 1 grams and 2 grams from it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "corpus = [\n",
    "    \"Thor ate pizza\",\n",
    "    \"Loki is tall\",\n",
    "    \"Loki is eating pizza\"\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "import spacy\n",
    "\n",
    "# load english language model and create nlp object from it\n",
    "nlp = spacy.load(\"en_core_web_sm\") \n",
    "\n",
    "def preprocess(text):\n",
    "    # remove stop words and lemmatize the text\n",
    "    doc = nlp(text)\n",
    "    filtered_tokens = []\n",
    "    for token in doc:\n",
    "        if token.is_stop or token.is_punct:\n",
    "            continue\n",
    "        filtered_tokens.append(token.lemma_) # lemma gives you the base word.\n",
    "    \n",
    "    return \" \".join(filtered_tokens) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'thor eat pizza'"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preprocess(\"Thor ate pizza\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Loki eat pizza'"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preprocess(\"Loki is eating pizza\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['thor eat pizza', 'Loki tall', 'Loki eat pizza']"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corpus_processed = [\n",
    "    preprocess(text) for text in corpus\n",
    "]\n",
    "corpus_processed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'thor': 7,\n",
       " 'eat': 0,\n",
       " 'pizza': 5,\n",
       " 'thor eat': 8,\n",
       " 'eat pizza': 1,\n",
       " 'loki': 2,\n",
       " 'tall': 6,\n",
       " 'loki tall': 4,\n",
       " 'loki eat': 3}"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "v = CountVectorizer(ngram_range=(1,2))\n",
    "v.fit(corpus_processed) # Fit Creates the vocabulary\n",
    "v.vocabulary_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now generate bag of n gram vector for few sample documents"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1, 1, 0, 0, 0, 1, 0, 1, 1]], dtype=int64)"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "v.transform([\"Thor eat pizza\"]).toarray() # creating a vector"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's take a document that has out of vocabulary (OOV) term and see how bag of ngram generates vector out of it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1, 1, 0, 0, 0, 1, 0, 0, 0]], dtype=int64)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "v.transform([\"Hulk eat pizza\"]).toarray()"
   ]
  },
  {
   "attachments": {
    "467b1a02-0d94-4efb-9c1f-9a122d2bc309.jpg": {
     "image/jpeg": "/9j/4AAQSkZJRgABAQEAYABgAAD/4RCaRXhpZgAATU0AKgAAAAgABAE7AAIAAAANAAAISodpAAQAAAABAAAIWJydAAEAAAAaAAAQeOocAAcAAAgMAAAAPgAAAAAc6gAAAAgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAERoYXZhbCBQYXRlbAAAAAHqHAAHAAAIDAAACGoAAAAAHOoAAAAIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAEQAaABhAHYAYQBsACAAUABhAHQAZQBsAAAA/+ELH2h0dHA6Ly9ucy5hZG9iZS5jb20veGFwLzEuMC8APD94cGFja2V0IGJlZ2luPSfvu78nIGlkPSdXNU0wTXBDZWhpSHpyZVN6TlRjemtjOWQnPz4NCjx4OnhtcG1ldGEgeG1sbnM6eD0iYWRvYmU6bnM6bWV0YS8iPjxyZGY6UkRGIHhtbG5zOnJkZj0iaHR0cDovL3d3dy53My5vcmcvMTk5OS8wMi8yMi1yZGYtc3ludGF4LW5zIyI+PHJkZjpEZXNjcmlwdGlvbiByZGY6YWJvdXQ9InV1aWQ6ZmFmNWJkZDUtYmEzZC0xMWRhLWFkMzEtZDMzZDc1MTgyZjFiIiB4bWxuczpkYz0iaHR0cDovL3B1cmwub3JnL2RjL2VsZW1lbnRzLzEuMS8iLz48cmRmOkRlc2NyaXB0aW9uIHJkZjphYm91dD0idXVpZDpmYWY1YmRkNS1iYTNkLTExZGEtYWQzMS1kMzNkNzUxODJmMWIiIHhtbG5zOnhtcD0iaHR0cDovL25zLmFkb2JlLmNvbS94YXAvMS4wLyI+PHhtcDpDcmVhdGVEYXRlPjIwMjItMDgtMTJUMDk6Mzg6NTYuOTk4PC94bXA6Q3JlYXRlRGF0ZT48L3JkZjpEZXNjcmlwdGlvbj48cmRmOkRlc2NyaXB0aW9uIHJkZjphYm91dD0idXVpZDpmYWY1YmRkNS1iYTNkLTExZGEtYWQzMS1kMzNkNzUxODJmMWIiIHhtbG5zOmRjPSJodHRwOi8vcHVybC5vcmcvZGMvZWxlbWVudHMvMS4xLyI+PGRjOmNyZWF0b3I+PHJkZjpTZXEgeG1sbnM6cmRmPSJodHRwOi8vd3d3LnczLm9yZy8xOTk5LzAyLzIyLXJkZi1zeW50YXgtbnMjIj48cmRmOmxpPkRoYXZhbCBQYXRlbDwvcmRmOmxpPjwvcmRmOlNlcT4NCgkJCTwvZGM6Y3JlYXRvcj48L3JkZjpEZXNjcmlwdGlvbj48L3JkZjpSREY+PC94OnhtcG1ldGE+DQogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgIDw/eHBhY2tldCBlbmQ9J3cnPz7/2wBDAAcFBQYFBAcGBQYIBwcIChELCgkJChUPEAwRGBUaGRgVGBcbHichGx0lHRcYIi4iJSgpKywrGiAvMy8qMicqKyr/2wBDAQcICAoJChQLCxQqHBgcKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKir/wAARCAFNBKsDASIAAhEBAxEB/8QAHwAAAQUBAQEBAQEAAAAAAAAAAAECAwQFBgcICQoL/8QAtRAAAgEDAwIEAwUFBAQAAAF9AQIDAAQRBRIhMUEGE1FhByJxFDKBkaEII0KxwRVS0fAkM2JyggkKFhcYGRolJicoKSo0NTY3ODk6Q0RFRkdISUpTVFVWV1hZWmNkZWZnaGlqc3R1dnd4eXqDhIWGh4iJipKTlJWWl5iZmqKjpKWmp6ipqrKztLW2t7i5usLDxMXGx8jJytLT1NXW19jZ2uHi4+Tl5ufo6erx8vP09fb3+Pn6/8QAHwEAAwEBAQEBAQEBAQAAAAAAAAECAwQFBgcICQoL/8QAtREAAgECBAQDBAcFBAQAAQJ3AAECAxEEBSExBhJBUQdhcRMiMoEIFEKRobHBCSMzUvAVYnLRChYkNOEl8RcYGRomJygpKjU2Nzg5OkNERUZHSElKU1RVVldYWVpjZGVmZ2hpanN0dXZ3eHl6goOEhYaHiImKkpOUlZaXmJmaoqOkpaanqKmqsrO0tba3uLm6wsPExcbHyMnK0tPU1dbX2Nna4uPk5ebn6Onq8vP09fb3+Pn6/9oADAMBAAIRAxEAPwD5tooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiuw0bStE0rwkviLxLaS6j9qnaCzso5TEG2/edmHIGeOP68R+KtH0k6DpviLw3DJbWV4zQzW0shcwyr2DHkgjP5e9AHJ0UUUAFFFFABRRRQAUUUUAFFFFABRWv4fg0GSWeTxLd3cMUagxxWkYZ5j3GTwPx9av/EDRrDQfFTWWlRtHbiCNwHcscsuSSTQBzNFFFABRRRQAUUUUAFFFFABRRRQAUUVs+Hn0CCS4uPEkVzcrGo8i1gbZ5rHrubsAPTmgDGortdQsfDuueC77WtB02bSLnTJYlnt2uGmSRZG2ghm5zmuKoAKKKKACiiigAooooAKKKKACinRbPOTzt3l7hv2dcd8e9egaFF4K8U6suh2vh+7sJZ1YQX32xpGBClsuh+UDjtmgDz2ilddkjLuVtpI3KeD7ikoAKKKKACiiigAooooAKKKKACiiigAooAJOBya9BuPCGlaZ8O9WkvI/M8QWSW00pEjYthNIAI8A4LbQScg/eoA8+ooooAKKKKACiiigAooooAKKKKACiit3wVp1rq/jPTbDUIvOtp5Ssibiu4bSeoIPagDCorrNY1fwr5V5Z2Pg/wCy3Clo47n+05X2EHG7aRg/SuToAKKKKACiiigAooooAKKK6PwnoNlqEd9qmuyyRaTpqK83lcPK7HCxrnuT/nnNAHOUV0Ou6l4ZvtOCaHoM2l3STj94120yyRYOchujZ29PfmueoAKKKKACiiigAooooAKK6HR9Z8N2WnLDq3hX+0rkMS1x/aMkORngbVGOK1vFdv4ei8G6Xe6foH9lX2oyNIim8km2wrxn5jj5iRjjoKAOIooooAKKKKACiiigAooooAKKKKACiiigAoruLSw8OeH/AAhpmra5pU2tT6o0hVUuWhjgVGxjK8lu+P5d83xxoun6RqNlLpAlitr+zjultpzmSDdn5T+X86AOZooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKAOx16MzfCrwpPGMpbzXcMp9HaTcB+QNF1G1r8GbFJsBrvWXnhz12LFsP4ZrL0Dxff+HrWe0hhtL2yuGDSWl7D5sZb1xkc1B4g8SX/iS6ilvzEiQp5cMECbI4l9FXtQBk0UUUAFFFFABRRRQAUUUUAFFFFABXafFb/keX/69Yf/AEAVgaFr39hSzP8A2Vpmo+aoG3ULfzQmO68jBrV8QePrrxJZzQ32i6NHLKFX7XDakTqFIIAcscdMfQmgDlaKKKACiiigAooooAKKKKACiiigAq/ouiX2v6itlpsXmSEbmZjhY1HVmPYCqFdH4c8aXfhrT7uyttP027hvGBmF5AZNwA+7wwBHsaAJvEGrWNhoy+GPD0nn2kcnm3l7jH2uUenog7evX3PLV0Wr+L/7W02Sz/4R7QbLeQfPsrLy5Fwc8NuOM9K52gAooooAKKKKACiiigAooooAK9G8IeKo9YvIfDc2iWVml+n2ZrvS0MEwG37zEE7unOeOtef2V3LYX9veW+3zbeVZU3DI3KcjI78iurk+JeqeXN9h03SNOuJlKyXdnZ7JiD/tZNAHLajafYNUurMOJPs8zxbx/FtYjP6VXoJLMSxJJOST3ooAKKKKACiiigAooooAKKKKACiiigDr/h3ZafLrE9/qN9p9vJYx+ZaxX84ijlmOdpJPZSMnAPauqsPDM8/hPxaLvxPoN1PqT2zyXSX26OIiUsfMbb8uc4HqeK8mrSstcubDQtT0qFImg1PyvOZgdy+W25dpzgcnnINAGfKnlzOm5X2sRuQ5Bx3B9KbRRQAUUUUAFFFFABRRRQAUUUUAFWtN1K60jUob/T5PKuYDujfaG2nGM4PHeqtXdG1afRNWh1C1SGSWHdhJk3owZSpBHfgmgDvtC8SX/jrTNZ0zxSIbuK3sJLqO8MKo0Dr0OVAGOf0NeaV1GqePtT1HSZNNt7TT9KtJjmaPTrfyvN9m5PFcvQAUUUUAFFFFABRRRQAV6H4T1VtJ+Fms3dnDHNd21/E48xd6xhgFVyvfHzYzxnB7V55Wt4e8S6h4avJJ9PMbLMmyaCZN8cq+jCgDrLzWLnxj8NNUv/EEcUt7pdxCLa9EYRn3thk+UAcDn8vSvPa39e8Zahr1jDYPBZ2FjC29bSwh8qPd/eIyef8AGsCgAooooAKKKKACiiigDQ0HSJde1+z0yDh7mUISP4V6sfwAJ/CtHxxqsWqeKZ/sQC2Vmq2dqo6COPgY9icn8ao6Hr114eubi4sI4TPNA8AkkUkxBurLgj5vc5rMoAKKKKACiiigAooooAKKKKACiiigAooooA9Hj1geAvCOkQCyj1STVEW/db3LQwjIwEX+9j+LscVkfEC2SabTPEEUlxt1m3MxhuJC7QspAIDH+HkYqvpXj7UtN0iLTJ7PTtUtYG3QJqNv5vk/7vIrL1/xFqHiXUBd6nIpZECRxxrtSNR2UdqAMuiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACp3s5U06G9YDyZpZIUOedyBC3H0kWoKvTQ3C+HbOd5t1s91OkcP9x1SIs34hkH/AAGgCjRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFWr7StQ0vyP7Ssbmz+0RiaH7RC0fmoejLkcqfUcVVoAKKKKACiiigAooooAKKKtS6XqEGmwajPY3MdjcsywXTwsIpSpwwVyMMR3weKAKtFFFABRVvUdK1HR7hbfVrC6sZmQSLHdQtGxU9GAYA4PrSadpWoaxd/ZdIsbm/uNpbybWFpX2jqcKCcCgCrRRRQAUUUUAFFW7vStR0+3tbi/sLq1hvE8y2kmhZFnX+8hIww5HIqpQAUU6ON5ZFjiRndyFVVGSxPQAVNf6fe6VfSWWqWc9ldxECSC4iaORMjIyrAEcEH8aAK9FWrHStQ1Tz/7Nsbm8+zxmab7PC0nlIOrNgcKPU8VVoAKKKKACiiigAooooAKKKKACirVnpeoajFcy6fY3N1HaRGa4eCFnEKDq7kD5V9zxVWgAooqz/Zt9/ZX9p/Y7j+z/O+z/a/KbyvN27vL34xu284znHNAFaiirWo6VqGj3f2XV7G5sLjaG8m6haJ9p6HDAHBoAq0UUUAFFFW7TStR1C3uriwsLq6hs08y5khhZ1gX+85Awo4PJoAqUVZ/s2+/sr+0/sdx/Z/nfZ/tflN5Xm7d3l78Y3becZzjmq1ABRRRQAUUUUAFFT2NheanfR2em2k95dTHEcFvGZHc4zgKOTwKjmhltp5ILiN4pY2KPG6lWRgcEEHoQe1ADKKKKACiiigAooooAKKKKACiiigAoqexsLzU76Oz020nvLqY4jgt4zI7nGcBRyeBUsOj6ncz3cFvp13LLYxvLdRxwMzW6IcOzgD5Qp6k4A70AU6KKKACiiigAooq1daXqFjaWt1e2Nzb294pa2mlhZEnA6lGIwwGR0oAq0UUUAFFFFABRRRQAUUUUAFFFWtR0vUNHuza6tY3NjcBQxhuoWifB6HawBwaAKtFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFXZbZV0G1uROWeS5mjMOeECrEQ2PfeR/wGqVW5Y7QaJbSRvm8a4mWVM9Iwsew49yZPyoAqUUUUAFFFFABRRRQB1Hgr4e6547muf7HW3gtbNd1zfXkvlQQDtubB9OwNW/F3wv1rwjotvrL3emaxpFxJ5a6hpFz9ohV/7pbAx37Y4r0Xw43hSP8AZehHiI639hl1thf/ANheUZPNA+QS+ZwE2hPx2VhTeM/hvpXwy8R+F/DEXiudtY8qRBqq25jhljYMGBjYEZwM8HO0e9AGJoPwc1vWvD9nrN1rGgaJaX5xZ/2tqAha45wNoAbqQeDg8dKfpHwt1Gx+LFt4X8UXOl6c8Lx3Dm+udkN3F5ijbE2PmZgSAOOhHBFdLrWh+Afh1BpWn+PLLWfE+uTafFcMsV35VtboxbEaHIbAwfbvxnAt/Fhwvxh+Hjy2babjStMZ7WWYyG2/0h8o0h5Yr0LHrjNAGN8cPh/Y+H/Feq6npOqeHbe082GOPRbS6AuoQY1BJhC8DILZz/ED3ryWvRPj2MfHHxFn/npD/wCiI687oAKKKKANLw/4e1PxTrlvpGhWrXV7cHCRqQOBySSeAAO5rt9V+BviPT9Mv7uy1PQtZl01d95ZaXfedcW4HUsm0dOeOvFbv7PDWEcnjOa5N2LqPRHMf9n7TdCLP7wwhuN/3Me+KXwX4x+EXgTxNFrejr45knjjeNoblbNopVZSCGCkEjoevUCgDhPBnw51jxta3t5ZXOnafp9jj7TfaldCGGMnkAnBPQE9Mcdat+JPhVrPhiTSJbzUNIudM1eZYbfV7S78y0VicfNJgYAGSTjGAeuDXRaB4a8JaX8Oj4x8aNq93o1/qjwadpFjKFGVDfvJST1wpHBz7nPGt45u9Ivv2adPuPDeg3Gh6W3ibFvDPdtcNKBbyBpAW+6N25do4+QnvQBb+IPwm0u38A+FJNN1zwjp1xb2Esl1cPfCP+1Hwh3RNtzL0OOn3h614LXrnxcYP8K/haykMP7LmGQe48oEV5HQB6NafBHxJc6bp+py3+jWWmXtjHffb727MMMCv91HZl++eeFyOOvSsfxt8NNe8CQ2d1qbWd7p98P9H1DT5vNgkP8Ad3YBzjnpyOmcHHZfGR7/AP4Vv8M0Pmf2f/YiEYztMuxM57Z24x9TTLUtD+yTejVd22fxAv8AZgl56Ku8p6DiXpxndQBjeK9O8V69rnhPT/GXiXTSdQ0eG6sLu8m8qG1t5AxUSvsGG+TBPzZOOTWhpvwF1HWbh4NI8b+Cb+ZIzK8drqzSsqDqxCxkgDI596rfGCN5pfAEUKNJI/g7TlREGSxPmYAHc1f8TSJ8J/AB8GWTofE2uxLLr0ykE2sJHyWoI7kElvqezDABz2j/AAwGs2L3K+OfBtkFmkiEd9qphdtjFd4Upna2MgnqCOBR4x+FOqeC7LSrm61jRNS/tdwtnHpty8rSqR98Eoo2cgZz3FO+GHg6z1y+u9f8TsbfwvoSi4v5SP8AXH+CBfUseMDt6EipH1DVfjJ8YrJWnXTnvZ1gtQpO2ygTJCrjGSqgntlvTPABqJ+z74neX7H/AGx4cGseT5v9jnUh9rxjpsC49s5x715hc201ndS211G0U8LtHJG4wUYHBBHqCK9+8MTfDrS/i9YaJoHhzXdY1+DVfJbU73UDHiRZMSTbVPzAAO2CBuAwepryH4jf8lT8V/8AYZvP/R70Ac3RRRQAV6n4q/5Nq8Cf9f8Aff8Aoxq8sr1rXfs5/Z5+Hn20kW39p3fmkddnmtn9M0AY2jfBnX9T8P22tanqWieHbK7x9lfW777ObgHoVG08HtnGevTFYXjTwHrngLUobXXYYylynmW11bv5kNwvGSjd8ZGQQDyOxBPZftHC+HxeuBc5FiLO3/s7+75OwZ2+3meZU/iNp/8AhlTwsupbvOOtymzEn3vs+2TOM87dxH6e1AFr416HqHiX4w6LpGjW7XN7daTapHGv/AsknsAOSewFZHhHwr4v8NfFufw/4N8S6VFrkNrKs91AzSQIANzxEtEcsMDopAPfIOPcPFKC5TWbXwTJDD49fQ7bY0uRLJZ4O5YGz8r5zz67c9iPFf2e45YfjEI7hHSVLK6V1cEMrBDkEHvmgDypmZ2LOSzMckk5JNJRRQAUUUUAer/Fv/knfwy/7Arf+yVSs/gZ4ik02zutZ1XQPDz3wBtrXWNQ8iaXPTChTyfTr7V1XiJbdh8D1vQptjFbiUN02edFnPtjNcZ8dPt//C59f/tPzd3mr5HmZx5WwbNvtj075oAyr/wdrXgj4gafpXiC18mf7TE8bo26OZC4w6N3B/P1wa7H4neFNV8a/tKa1ougwCW6mkiJZiQkSCCPLueyj1+gGSQKveL2m/4Vv8IV1rf/AGrvkKbuotvNTy8/8A8vH416N8SEkvdO8c2nw5wviYXETa7GoP2qe18lNvkkH7uCAQOT8w6kZAPEvCuneK9D1rxdpng/xJYKdP0m4uNSnt2ZoriGLG9Iy0eS2WwDhe+GxyfO69J+Df8AzPv/AGJmo/8AtOvNqAOs8GfDfXPHEF3d6e1nY6bZf8fGo6jP5NvEf7pbBOcHPA474yMr4y+G+teCrOyv7yex1HTL7It9R0y48+3duu3dgc4GenPOM4OPTrZvCUf7MfhZfFp14WEmoXBl/sPycmcSSbfN8zj7mMf/AKq5vWPGvw/tfhFqfg7wjB4kkkvLyO8ik1dLcrE4KhsGNuMqpHQ9aAM7Rvgnr2q6Pp2oXWs+HtHGqIr2NvqWo+XLcK33SqqrZzngdfan+D/hjdp8XoPD3im40nT5NOvbdrm11G5Ci+jMi/u4eD5jOp4XjOe1dJ4j0T4d/DvVofD/AIp0rXPF3iFLeHzXF0YYVJQBIo9p3YAwB17D2Gn8Tzt/ak8Gl08n95phZGbOz9/0J749aAOL+M3ga18NeKtSv9O1Xw+bWa+MUWk6fcg3Fou0n54goCAYx16keteZ12vxiBX4x+Jsgj/TmPP0FcVQAUUUUAep/B7/AJFb4j/9i3N/I1y3gr4c6/47a5fSI4ILKzGbm/vZfKgh4zgtzz7AHHfFdT8Hv+RW+I//AGLc38jVzVFkH7JeiHSg/lNrcp1Pyc43fPs8zHbHl9eM7fagDnPEPwg17Q/Dcmv2l/o+v6VC22e60W8+0LD7twOPcZx3q7/zaz/3Of8A7ZVqfABnW78YG+3f2EPD8/8AaGf9X22599vmY/4FXS/B248MW3wjsJPGYUWY8Zn7O8v+pjuPsQ8tpf8AY6+2cZ4zQB5TrvgDWfCHhvStd1i6tbGbUf3tpYeY4uwo5EhULhR0PLA89M5AtfFmDxVbeODD461SDVNUS1i/f24wnlkZVfuL0ye3fvTfizbeKbb4j6j/AMJu5k1CR90cqg+U8WTsMWf4McAduQec1v8A7RX/ACVuX/rwtv8A0CgCn8EPCFl4r+IVp/atzpP2S1kDS2F/NtkvAVf5Ykx85BUEjjAxUfxD+HraT40lh0fVtA1BtT1WWC103SrwPLbFpCEjkTAEeMhcZwCMdqg+CLKnxq8Nl2Cj7SwyTjkxsBVu08MzeL/2jr3Q4L2TT5J9cu3+0x53xCN5JCV9GwhwfXFAFtv2f/E5M9va6v4du9Vt4jLJpFvqO67AHbZtxnt1x71N8JlZPh78T1cFWXRlBBGCD89dj8M7v4bw/FnTdL8IeGdZutTgmlQate3hXYFRg0hRTggjIwQOo4B4rG8KcWnxrx/z7zf+jZaAOY/5tZ/7nP8A9sq82r0n/m1n/uc//bKvNqACiiigC7omlT69r+n6RZtGlxqF1HaxNKSEVnYKCxAJxk84Br0GD4C+JGu3tdQ1bw/pVwbmW3todQvzFJeFGKl4k2lmUkHBIBI5xXK/Dj/kqfhT/sNWf/o9K0vjC+pS/GLxJ/awfzxfOsQOf9SDiLHts20AWPDvhPxp4Y+MljoGnSxaP4ljlZbeebDRDMbfNnawZWXI6Hrg98VrLwZe+K/E2vjVvFXh3S7+0vGFzNq999mF1KzvvaP5fmGVJPAxuXjmvY70yL8efhXBqe461FocC6gXOW8zy5OGPru3/nXkVv4OvfHXxl1bRrF1hVtRuZbm5f7tvCsp3yH6dvUkDvQBY1/4O3/h/wAHTeJpPFPhi/06KQQq2n3zzGWQ/wDLNMR4Ld8Z4AJq5afA+51C8itLDx94FurmZgkUMOsl3dj0AUR5J+lZfxP8Y2euX9roPhhTb+F9CU29hED/AK4/xzt6ljzk9vQk1v8AhqJPhR8Px4yvUQ+JtdiaLQYWAJtYSPnuiD3IOF+o7McAHNxfCvW7r4hXfg/TrzTL2+sUMl1cQ3BFvAFxu3Oyg/KWAIAODx2OJ/Efwg1vw/4Yn8QW+qaJrunWsgiupdHvfP8As5JA+f5R3IHGcZqb4a+GdO16x8UeI/Fmo6imk6PBHJfQ2Lfvr0yOSqEnjBaPJz3wcjGa9F8MX/hDUPhV8RD4H8MXuk28elgT3N3dtL57Ycqu0kgEcnj1oA+d6KKKACtHQNI/t7XLbTf7QsNN88sPtWoz+TBHhSfmfBxnGBx1IrOooA9x+N3wz0vSbmDUNF1PwzpcFrpUO7TFuhHdXT7mzIke3585HzE87T6VymjfBPXtV0fTtQutZ8PaONURXsbfUtR8uW4VvulVVWznPA6+1a37RrB/HmjOhDK2gWxVgcgjdJyK0fEeifDv4d6tD4f8U6Vrni7xClvD5ri6MMKkoAkUe07sAYA69h7AAxPhp4b1Pwl+0Zoui65B5F7a3TB1DblIMTEMD3BBBFaHgz/kevi1/wBi9rP/AKNWu48SA/8ADX3hPdGYj9iiyhbcVPly8Z749a4fwZ/yPXxa/wCxe1n/ANGrQB47RRRQAVZ02z/tHVbSx+029p9pmSH7RdSbIotzAb3b+FRnJPYA1WooA93+Jvws0yw8B+GrvTNY8KWM1npDy3TC8CPqrqitugO3M2cHGccsPWm+IPBOp+MfhX8OfsM1nZWlnp1y91f6hcCG3twZE27nPc4OAAeh9KxvjIc/D/4YEcj+wgMj1Cx5pvxNvJ1+CvwysllYW72tzK8YPDOGUKT7gM2P940Acp45+GmueAY7G41SSzvbHUFLW19p8xlhcjtuIHOCD05HTODjQ0P4O6/q3hyHXtQ1DRfD2m3P/HtPrd79nE47FeDx6ZxnqOK29Vdpf2StE8wlvJ8SvHHn+FTFKxH5nNSftJLcL8RrJVDjSxpkA08D/V+Xznb26/09qAOI8a/D7XfAV3BFrcUMkF0u+2vLWTzIJxxna3HTPQgH8Oa6W++BfiDSleXWNa8O6baiJHjur2/MUczMpbyk3JlnAwSMY5HJ5xfvWmH7JtgNZ35PiJv7Kz/zy8tt3/Ad3m/jinftHXc83jrSLZ5GMNvotv5ce47VLFiSB6njn2HpQB5FRRRQAV0Pgfw7H4o8V22nzalpunx7hI76lceTHIAwygbByxzwO9c9V3RJEi8QafJK6oiXUbMzHAUBhkk0AepfHL4eWGg+K9X1TSNV8N2lmht1j0O2ugt3HmJFJ8gLxk5fOeQc966X4ufDPVvF3xKl1BNQ0nR7Fre2t4bnV7zyFuJfLHyR8EsRkdu9cL+0IrL8ddfLAgMLYgkdR9miH9K0P2kLye4+KUcEshaK2063SJc8KCCx4+poA4Xxh4J1rwP4mOh63Cn2sorxGBt6zK3AZTjJBII6ZyDXXQfAbxH5Np/a+seHdEvLxd0Gn6nqPlXEn0QKeenGe/Neg3v2O58V/A651lg/m6fCHklOS7hUMeT3+cj8TXjnxT/tD/hbHiX+1/N+0/2jL/revl7v3ePbZtx7YoAr3vgLXtM8fWvhDVLdbLU7q6itojKx8tvMYKjhgDlMnqAeh4yMVqeI/hTq/hPRr2/1zVNFtpbWZo47BrzF1dKsvleZFEVyyZy2Tg7QTivS/E7SHxT8D11Pd/bAWyN153+t2+dFs355znf175968q+K+oT6n8XPE81025o9SmgXnosbmNR+SigDkQCSABknoBXp1n8BPE88NmL7U9B0q/vk322l39/5d1KPZAp59s8d8VzPwzW0b4p+GV1HH2c6nBuz0zvG3PtuxmvSviLcfDD/AIWdrUvieTx8msx3jea1v9kVFK/cMefmCbQu0nnGCeaAPLLDwRr+oeOF8Ix2Jh1kytE1vO4TYVUsxLE4wFBbIzkdM5FdPrHwT17TNB1HVbPWNA1qPS1L30OlX/nS2yjqXUqMYwc854NdXcHw78YPinrHie3n1TR9D0zRTc6m6oiXU4RCjKNrMPmTgk5yARjmt34Yaj4Gv5vE0fgbwnqViY9BuRcajeXrPlSBhCmSuWIz6/Icd6AMLwX8L9M1P4Ha/e3Gr+FhqF4bV7e/ubwAaYvmKWSZtv7p2GVwM5zivHdc0r+w9budO+32Oo/Z2C/atPm82CTgHKPgZHOOnUGvSPBjCT9mv4ixIQ0i3Ni7IDkhfOTkj0+U8+x9K8ooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACrcslodEto40xeLcTNK2OsZWPYM+xEn51Uq7Lcq2g2tsICrx3M0hmxw4ZYgFz7bCf+BUAUqKKKACiiigAooooA6vwT8R9e8Bm7j0k21zZXqgXNjfQ+bBLjoSuQc/QjPfNWfE/xO1DxLoP9jR6JoGiWDSiaSHR9PEHmMOm45J/LFcXRQB6TbfHPxNDo9jZ3en6HqNzp0Qis9Tv7Dzrq3AGAVcnGRxyRzjnNcx4t8cav401DTb/WGjF3p1jFZRzQhg0ixliJHJJy5LEkjA9hXO0UAdx4x+Kuq+OtKjtdd0jQzdoEDapDZbbtwucAybjwc8gACuHoooAKKKKANXw14l1Xwjr1vrOg3RtryAna2AQwPBVgeCCO39a7W/8AjhrN5Fdm38N+FtPvbyNo5tQtNLC3DBhgneWPP4V5rRQB23hP4q634T8PT6CtjpWs6TNL5v2LV7X7RGj9yo3DHTOOmee5y3xR8VPEHi/w0dD1dLL7GL9b6IQQmPyCsXlLEgB2iMAk4xnJJzXF0UAdzpvxY1ex8DR+FLzStF1fT4A/2ZtTs/Oktt+cmM7gARuODgkfTiuGoooA988ZeP8AUvB3g/wHawWWmarp974bt2nsNVtfPgZlxtfbkYYZI6/yFeW+NviNrvj2S1XVzbW9pZLttrGyi8qCH1IXJ5+prn7zVNQ1GK2i1C+ubqO0iENuk8zOIUHREBPyr7DiqtAHaQfE/VIvE3h7XJdN0u4ufD2mxadZRzRSGMpEGCSMA4JcbicggZwccVt3fxwudQvJbu/8A+Bbq5mYvLNNoxd3Y9SWMmSfrXmFFAHpen/G3UNP8PtoieD/AAjPpzXL3RtrjTnkTzGYtnaZMcZwOOAAO1Zmq/FK9vptPuNK8OeG/Dt5p90t1Dd6Lp32eUsoI2sdx3Ic8qRg4rh6KAPUZ/j94mkv0v7XSfD9jfGaOW4vLSwKTXYRg2yR9xJUkAEAjI4zXnet6rPr2v6hq94kaXGoXUl1KsQIRWdixCgknGTxkmqVFABRRRQAVv6h4x1DUvA+k+Fp4bZbHSZpZoJEVhKxkYltxLYI54wBWBRQB6HpHxp1/T/DtromqaZofiKzswBajWrH7QYABgBTuHA7Zye1c74x8ca3461SO816dCIE8u3toE8uG3T+6idvryTgc8CueooA63XPiTr+ueMbDxPvh0/U9PhjhgkslZQAmcEhmbOckEdCOMVpf8Lg1hfiEfGcGkaNBqj27QTCKGURzbl2l2XzPvY7gj3BrgKKACiiigAooooA6DxF4z1HxNouhaZfw2scOh2ptbZoUYM6nHL5Ygn5R0Arq7X46+Ixptna61pPh/xDJYgC3u9YsPPnjx0IbcORxzjPHJNeaUUAb+veNdb8TeKk8Qa3dfaryN1aNSuI41U5VFUdFHoPc9STWjN8T/EL/EyTxzbPBZ6rI6s6W6sIXAVVKFSxJUhRkZ+mOMcfRQB2TfEvUR4i8Qaxa6TpNnN4g06bT7yKCKQR7Zcb5FBc4c7Qc5IznjmuNoooA6/wX8Tdd8D2V3YWEdjf6ZeHdPp+o2/nQO2AN23IOcADrg/gKTxb8RtQ8WaTb6W2kaJo+nwTeeLbSLEW6NJgjceSejN371yNFAHpr/HnxTNZ24uLDQ7jU7aHyYNZm08PeRDGMq5OAf8AgNcz4p+IGs+LPFVn4hvRb2uoWcUUcL2qsADGcq3zM3zZ5/pXMUUAdn44+JuoeP7eIaxo2h292kgd7+yszHcTYUqFdyxyvPTjoPSuMoooAKKKKAN/w14x1DwtYa3aafDbSR61YtY3BnViURupTDDDfXI9qu+CfiTr/gP7VFpLW9zY3gxcWF9F5sEpxjJXI5xxwRnvnArk6KAO58SfFnXfEHh86FbWWk6DpLtvms9Fs/s0cx/2+ST9M4PfPFYP/CV33/CA/wDCI+Vb/wBn/wBp/wBp+btbzfN8rysZzjbt5xjOe/asSigDrNX+I2ra/wCB7DwzrVtZXsenNm0v5Uc3UK/3A+/BXHGCp4A7gEUvGnjHUPHXiJtZ1aG2huGhSEraqyphBgcMxOfxrAooAt6Vql3omsWmp6bKYbuzmWaGQfwspyPqPau11n4x67q3iTT/ABBBpeh6Vq9jMZvt2n2WyW4JXaRKzM28bcjHua8/ooA9TH7QPiaHU0v9P0bw7p87SiW6e008xtekA4Erb9zDJzwRyPTiuW0/4harpsfihILezYeJ0ZLzejnywzMx8vDccseu7tXK0UAbf/CV33/CA/8ACI+Vb/2f/af9p+btbzfN8rysZzjbt5xjOe/asSiigAooooA6T4cf8lT8Kf8AYas//R6V6J4o+NOvaF8QNctH0nQdVl0zV7yOxvdSsfNuLVPOcBUcMMADge3FeN21zPZ3UV1ZzSQXELiSKWJyrxsDkMpHIIIyCKLm5nvLqW6vJpJ7iZzJLLK5Z5GJyWYnkkk5JNAHR2fxC122+IkfjW4liv8AV0lMubpSUY7SuCFIwADgAEYwK2PDPxf1Xwv/AG59m0PQb3+3Z3mvft1rJIHDZJjx5gGz5m4OepyTXAUUAegX/wAWBf2bWzfD/wADwBipL2+jmN+GBwGD5GcYOOxNampfHvUtZuFuNY8EeCb+ZEEayXWktKyoOigtISAMnj3ryuigDvtM+MGs6L4kvtV0fSNBsYNQgSC60q3sdtnKq5wTHuznk857mrF58bfEV1oupaPFpuiWOl39o1r9isrMwxQBiSzxhW+8eOW3dBx1z5zRQAUUUUAFFFFAHc6n8WNX1rwVb+HNY0rRb9bW2Ftb6jcWe+8hjGMBZC2AcADOMnvWo/x58UzWduLiw0O41O2h8mDWZtPD3kQxjKuTgH/gNeZUUAdxd/FnXr34iab4zmttP/tLTYVhijEcnlOFVgCw37ifmPRhWZpnjrU9K1bxFqFvBaNL4hsrmyulkRiqJOwZymGGCCOM5HqDXNUUAFFFFABRRRQB3EXxX1j/AIQFPCWoaXo2qWUMMkNrcX9mZbi0DgjMb7sKwzwcZGB6Vja74x1DxD4d0HRr2G2S30KGSG2aJWDuHIJ3ksQT8o6AVgUUAdDN401Gf4d2/gx4bUadb35v0lCN5xkKsuCd2NuGPbPvW/onxk1/SvDUGgajpui+ItOteLaHW7L7R5A7KvzDgds5x0HGBXn9FAHR+MvHmueOr+G41yaMRW6bLa0to/Lgt17hE7Z9Tk9OcAUzxn4z1HxzrUOp6tDawzQ2sdqq2qMqlUzgncxOefWufooAKKKKACiiigDufFnxY1fxroUdhr2k6HLcoiJ/aq2WLwhTnHmbuAe4AA5NYvjTxjqHjrxE2s6tDbQ3DQpCVtVZUwgwOGYnP41gUUAdL4i8dar4l0zw/ZXkdvbp4ftRa2klsrK5UbcMxLH5vlHIxXVx/HvxHJFaHV9F8Oa1e2a7YdR1HTfMuE9wwYAH3Ary+igDobjxzrt947tfF2qXIv8AVLW6iuUM4+TMbBlXauMLkdBjv9ay9b1WfXtf1DV7xI0uNQupLqVYgQis7FiFBJOMnjJNUqKAFVmRgyEqynIIOCDXpg+O2v3Ftarrmg+GddurVdsV9qmmCWdfQ7gwGeB27V5lRQB2Ok/FLxLo/ju78WWk1ub6+BS5heAeRLGcfIUGOBtXGCDx1653B8d/EcP2iLTtJ0DTbO4t5YZbKxsTDCxkABlIVslwM4JJHJ4rzKigDq/A/wAQ9W8BPqA0y3sL211KIRXdnqEBlhlAzjIBB43N3xyawtb1T+29audRNjZWH2ht32awh8qGPgDCJzgcVRooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACr001w3h2zgeDbbJdTvHN/fdkiDL+AVD/wKqNXpvtf/CO2fmbfsf2qfysfe8zZFvz7Y8vH40AUaKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACpLdYGuEF3JJFCT87xRh2A9lJAP5io6KAL3k6T/AGhs+23v2PbnzvsaeZu9Nnm4x77vwohh0lryZbi9vY7Yf6qRLNGd/wDeUygL+DGqNFAF22i0to5jd3l5E6k+SIrRXDjtuJkXb+GabHHpp09nlu7pbwZ2wraq0Z9Mv5gI/wC+TVSigC3JHpo09Xiu7prw43QtaqsY9cP5hJ/75FOuYtLWOE2l5eSuxHnCW0VAg77SJG3fjiqVFAF6aHSVvIVt729ktj/rZHs0V0/3VEpDfiwq/NZ2C6dZyPql62lPcTrH/oS+YkipEWOzzcYIZBnd/D0rCq7LbKug2tyJyzyXM0ZhzwgVYiGx77yP+A0ALDDpLXky3F7ex2w/1UiWaM7/AO8plAX8GNJbRaW0cxu7y8idSfJEVorhx23EyLt/DNUqKALccemnT2eW7ulvBnbCtqrRn0y/mAj/AL5NEkemjT1eK7umvDjdC1qqxj1w/mEn/vkVUooAu3MWlrHCbS8vJXYjzhLaKgQd9pEjbvxxSzQ6St5Ctve3slsf9bI9miun+6olIb8WFUaKAL3k6T/aGz7be/Y9ufO+xp5m702ebjHvu/CiGHSWvJluL29jth/qpEs0Z3/3lMoC/gxqjRQBdtotLaOY3d5eROpPkiK0Vw47biZF2/hmmxx6adPZ5bu6W8GdsK2qtGfTL+YCP++TVSigC3JHpo09Xiu7prw43QtaqsY9cP5hJ/75FOuYtLWOE2l5eSuxHnCW0VAg77SJG3fjiqVFAF6aHSVvIVt729ktj/rZHs0V0/3VEpDfiwo8nSf7Q2fbb37Htz532NPM3emzzcY9934VRooAvQw6S15Mtxe3sdsP9VIlmjO/+8plAX8GNJbRaW0cxu7y8idSfJEVorhx23EyLt/DNUqKALccemnT2eW7ulvBnbCtqrRn0y/mAj/vk0SR6aNPV4ru6a8ON0LWqrGPXD+YSf8AvkVUooAu3MWlrHCbS8vJXYjzhLaKgQd9pEjbvxxSzQ6St5Ctve3slsf9bI9miun+6olIb8WFUaKAL3k6T/aGz7be/Y9ufO+xp5m702ebjHvu/CiGHSWvJluL29jth/qpEs0Z3/3lMoC/gxqjRQBdtotLaOY3d5eROpPkiK0Vw47biZF2/hmmxx6adPZ5bu6W8GdsK2qtGfTL+YCP++TVSigC3JHpo09Xiu7prw43QtaqsY9cP5hJ/wC+RTrmLS1jhNpeXkrsR5wltFQIO+0iRt344qlRQBemh0lbyFbe9vZLY/62R7NFdP8AdUSkN+LCjydJ/tDZ9tvfse3PnfY08zd6bPNxj33fhVGigC9DDpLXky3F7ex2w/1UiWaM7/7ymUBfwY0ltFpbRzG7vLyJ1J8kRWiuHHbcTIu38M1SooAtxx6adPZ5bu6W8GdsK2qtGfTL+YCP++TRJHpo09Xiu7prw43QtaqsY9cP5hJ/75FVKKALtzFpaxwm0vLyV2I84S2ioEHfaRI278cUs0OkreQrb3t7JbH/AFsj2aK6f7qiUhvxYVRooAveTpP9obPtt79j25877GnmbvTZ5uMe+78KIYdJa8mW4vb2O2H+qkSzRnf/AHlMoC/gxqjRQBdtotLaOY3d5eROpPkiK0Vw47biZF2/hmmxx6adPZ5bu6W8GdsK2qtGfTL+YCP++TVSigC3JHpo09Xiu7prw43QtaqsY9cP5hJ/75FOuYtLWOE2l5eSuxHnCW0VAg77SJG3fjiqVFAF6aHSVvIVt729ktj/AK2R7NFdP91RKQ34sKPJ0n+0Nn229+x7c+d9jTzN3ps83GPfd+FUaKAL0MOkteTLcXt7HbD/AFUiWaM7/wC8plAX8GNJbRaW0cxu7y8idSfJEVorhx23EyLt/DNUqKALccemnT2eW7ulvBnbCtqrRn0y/mAj/vk0SR6aNPV4ru6a8ON0LWqrGPXD+YSf++RVSigC7cxaWscJtLy8ldiPOEtoqBB32kSNu/HFLNDpK3kK297eyWx/1sj2aK6f7qiUhvxYVRooAveTpP8AaGz7be/Y9ufO+xp5m702ebjHvu/CiGHSWvJluL29jth/qpEs0Z3/AN5TKAv4Mao0UAXbaLS2jmN3eXkTqT5IitFcOO24mRdv4ZpscemnT2eW7ulvBnbCtqrRn0y/mAj/AL5NVKKALckemjT1eK7umvDjdC1qqxj1w/mEn/vkU65i0tY4TaXl5K7EecJbRUCDvtIkbd+OKpUUAXpodJW8hW3vb2S2P+tkezRXT/dUSkN+LCjydJ/tDZ9tvfse3PnfY08zd6bPNxj33fhVGigC9DDpLXky3F7ex2w/1UiWaM7/AO8plAX8GNJbRaW0cxu7y8idSfJEVorhx23EyLt/DNUqKALccemnT2eW7ulvBnbCtqrRn0y/mAj/AL5NEkemjT1eK7umvDjdC1qqxj1w/mEn/vkVUooAu3MWlrHCbS8vJXYjzhLaKgQd9pEjbvxxSzQ6St5Ctve3slsf9bI9miun+6olIb8WFUaKAL3k6T/aGz7be/Y9ufO+xp5m702ebjHvu/CiGHSWvJluL29jth/qpEs0Z3/3lMoC/gxqjRQBdtotLaOY3d5eROpPkiK0Vw47biZF2/hmmxx6adPZ5bu6W8GdsK2qtGfTL+YCP++TVSigC3JHpo09Xiu7prw43QtaqsY9cP5hJ/75FOuYtLWOE2l5eSuxHnCW0VAg77SJG3fjiqVFAF6aHSVvIVt729ktj/rZHs0V0/3VEpDfiwo8nSf7Q2fbb37Htz532NPM3emzzcY9934VRooAvQw6S15Mtxe3sdsP9VIlmjO/+8plAX8GNJbRaW0cxu7y8idSfJEVorhx23EyLt/DNUqKALccemnT2eW7ulvBnbCtqrRn0y/mAj/vk0SR6aNPV4ru6a8ON0LWqrGPXD+YSf8AvkVUooAu3MWlrHCbS8vJXYjzhLaKgQd9pEjbvxxSzQ6St5Ctve3slsf9bI9miun+6olIb8WFUaKAL3k6T/aGz7be/Y9ufO+xp5m702ebjHvu/CiGHSWvJluL29jth/qpEs0Z3/3lMoC/gxqjRQBdtotLaOY3d5eROpPkiK0Vw47biZF2/hmmxx6adPZ5bu6W8GdsK2qtGfTL+YCP++TVSigC3JHpo09Xiu7prw43QtaqsY9cP5hJ/wC+RTrmLS1jhNpeXkrsR5wltFQIO+0iRt344qlRQBemh0lbyFbe9vZLY/62R7NFdP8AdUSkN+LCjydJ/tDZ9tvfse3PnfY08zd6bPNxj33fhVGigC9DDpLXky3F7ex2w/1UiWaM7/7ymUBfwY0ltFpbRzG7vLyJ1J8kRWiuHHbcTIu38M1SooAtxx6adPZ5bu6W8GdsK2qtGfTL+YCP++TRJHpo09Xiu7prw43QtaqsY9cP5hJ/75FVKKALtzFpaxwm0vLyV2I84S2ioEHfaRI278cUs0OkreQrb3t7JbH/AFsj2aK6f7qiUhvxYVRooAveTpP9obPtt79j25877GnmbvTZ5uMe+78KIYdJa8mW4vb2O2H+qkSzRnf/AHlMoC/gxqjRQBdtotLaOY3d5eROpPkiK0Vw47biZF2/hmmxx6adPZ5bu6W8GdsK2qtGfTL+YCP++TVSigC3JHpo09Xiu7prw43QtaqsY9cP5hJ/75FOuYtLWOE2l5eSuxHnCW0VAg77SJG3fjiqVFAF6aHSVvIVt729ktj/AK2R7NFdP91RKQ34sKPJ0n+0Nn229+x7c+d9jTzN3ps83GPfd+FUaKAL0MOkteTLcXt7HbD/AFUiWaM7/wC8plAX8GNJbRaW0cxu7y8idSfJEVorhx23EyLt/DNUqKALccemnT2eW7ulvBnbCtqrRn0y/mAj/vk0SR6aNPV4ru6a8ON0LWqrGPXD+YSf++RVSigC7cxaWscJtLy8ldiPOEtoqBB32kSNu/HFLNDpK3kK297eyWx/1sj2aK6f7qiUhvxYVRooAveTpP8AaGz7be/Y9ufO+xp5m702ebjHvu/CiGHSWvJluL29jth/qpEs0Z3/AN5TKAv4Mao0UAXbaLS2jmN3eXkTqT5IitFcOO24mRdv4ZpscemnT2eW7ulvBnbCtqrRn0y/mAj/AL5NVKKALckemjT1eK7umvDjdC1qqxj1w/mEn/vkU65i0tY4TaXl5K7EecJbRUCDvtIkbd+OKpUUAXpodJW8hW3vb2S2P+tkezRXT/dUSkN+LCjydJ/tDZ9tvfse3PnfY08zd6bPNxj33fhVGigC9DDpLXky3F7ex2w/1UiWaM7/AO8plAX8GNJbRaW0cxu7y8idSfJEVorhx23EyLt/DNUqKALccemnT2eW7ulvBnbCtqrRn0y/mAj/AL5NRzpaLbwG2nnkmYHzkkhCKh7bWDEt36hagooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACrcsdoNEtpI3zeNcTLKmekYWPYce5Mn5VUq3LJaHRLaONMXi3EzStjrGVj2DPsRJ+dAFSiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiit3wf4uv/AATr39raVDbTT+U0W25VmXDYzwrA549aTvbQDCor6T8CfFjXfFHhTxdqd/aafHNolkLi3WCNwrtslbD5ckjMY6Ed6831f4x+IPGcFto+qWemxW8l3C5a3ikV8q4I5ZyP0qFKTdrEKTvseaUV7B8dtJ1LWPilFDpOn3V9KumxMY7aFpGA3yc4UGvJr2xu9OumttQtZrSdfvRTxlGH1B5qoy5lcpO6IKKu6bouqaw7JpGm3l+yfeW1gaUr9doNRX2n3umXJt9Ss57OcDJiuImjYD1wRmqGV6Kux6Lqsunx30WmXj2csgijuFt2MbuTgKGxgnPGOtJqOkalo8qRavp91YSSLvRLqBoiy+oDAZHvSugKdFFaVh4b1zVbY3Gl6NqF7Aucy29q8ijHXkAimBm0U6WKSCVopkaORDtZHGCp9CK04PCviG5s/tdtoOpy22N3nR2cjJj13AYouBlUVqeHtCvPEOuQafZWlzcFnXzvs8Rdoo9wVnIAOAMjk8V1nxK+Gdx4S16WPQrDVrvSYrdZXvZYS6oTndl1UKAOKnmSdhXV7Hn9FFXYtF1SfTxfw6beSWZkEQuUgYxlycBd2MZyRx1qhlKitO+8Na7pdqLnU9F1GztzjEtxaPGp/EgDvXR/DDX9S8O6xqFzpPhqTX3lsmhkijjZjEpI+Y7Vb5eMEcZ9RSb0uhX0OJorR03w7rWsxPLo+j39/HG213tbV5Qp9CVBwau/8IL4u/6FbWv/AAXS/wDxNF0F0YNFath4W8QaraC60vQtSvbdiQJrezkkQkdRlQRUs/gzxRa28lxdeG9XhhiQvJLJYSqqKBkkkrgADnNF0F0YtFW9P0nUdXmaLStPur6ReSltC0hH4KDSahpeoaTcCDVbG5spiMiO5haNseuGANO4yrRRRQAVdluVbQbW2EBV47maQzY4cMsQC59thP8AwKqVXpprhvDtnA8G22S6neOb++7JEGX8AqH/AIFQBRoop8MMtxPHDbxvLLIwRI0UszMTgAAdST2oAZRWhqfh/WdFSN9Y0i/09ZCQjXVs8QcjqBuAzWfRe4WaCiiigAooqe6sbuxaNb21mtjKgkjE0ZTehJAYZ6jIPPtQBBRRRQAUUUUAFFFFABRRRQAUUUUAFFFOjjeaVY4UaSRzhVUZLH0AoAbRUt1a3Flcvb3kEtvOn34pUKsvGeQeRUVABRRRQAUUUUAFFFFABRRRQAUUUUAFFXtN0PVtZLjR9Lvb8p98Wtu8u367QcVVuLaezuHt7uGSCaM4eOVCrKfcHkUXHZkdFFFAgooooAKKKKACiiigAooooAKKK07Tw1rt/Ym8sdF1G5tRyZ4bR3Qf8CAxSukOzexmUUUUxBRRRQAUUUUAFFFFABRRRQAUUUUAFFad14a12ysBfXui6hb2hwRcS2rpHz/tEYrMovcdmtwooooEFFFFABRRRQAUUUUAFFFFABRRSqCzBVBJJwAO9ACUVPd2V1YT+Tf201tLtDbJoyjYPQ4PY1BQAUUUUAFFFFABRRRQAUUVLbWtxe3KW9nBJcTyHCRRIWZj7AcmgCKir2paJqujMi6xpl5YNJ9wXVu8Rb6bgM1Ro3DYKKKKACiiigAooooAKKKnuLC8tIYZbu1ngjnBaJ5IyokHqpPXr2oAgooooAKKKKACiiigAooooAKKKKACiilVWdgqKWZjgADJJoASipruzubC5a3vreW2nUAtFMhRhkAjIPPIIP0NQ0AFFFFABRRRQAUUVo6b4d1vWYXl0fR7+/jjba72tq8oU+hKg4NF7Ds3sZ1FK6NG7JIpVlOGVhgg+lJQIKKKKACiiigAooooAKKKKAPWvhD/AMk6+JX/AGCh/wCirivMNJ/5DVl/18R/+hCu5+HHinR9B8F+N7DVrz7Pc6rp4hs08p281/LmGMqCBy69cda4PT5Ug1O1llO1I5kZjjOAGBNQk7slbs97+NXxE1Xwp4jt9N8MullczWyz3V2IlZ3G5giAsDgDDHp/F9ayfiZI3i34V+DNdvljGp3c627zKuM7g2eB2ygOPc1yPxl8TaR4r8cRahoF39rtVso4jJ5bp8wZyRhgD3FWfE/izR7/AOC3hnQrG+L6pYT754RE6+WMSc7iNp+8OhrKMbKLsSlZI674m+Lr34Zf2X4Q8EbNNhitVnmuFiVnlJJX+IEZO0knqc+3MenatJ8VPg34hbxOiTanoEbXFverEEYgIXAyBjJ2MCB2I471n3HjLwL8SNHsV+IMt5o+sWMflm+tYy6zL36Kx5POCOCTg8mqfiDx14X8P+Bbjwj8OluJ477/AI/dRuVKmQHggAgHJAx0AAPHJoUdErait0tqdP4R8Q/8Ir+ziusxW8c91bXcn2USjKpKzlA34Bia8a8ReK9a8WXUVx4gvjeSwoUjYxom1Sc4+UDvXY/8JZov/DPv/CNfbf8Aib/bPN+zeU/3fMzndjb09682q4R1bZUVq2dH8P8AQ7bxJ4/0nSb7P2a4m/egHG5VUsVz77cfjXvXinUIbXXBZab8TbDwvb2CrEmmR2CN5eAPvEuM/TAGMfU/OGiaxdaBrlnqunsFuLSUSpnocdj7EcH2Neu6xrfwn+ITRax4iur/AEHVmRRcrbxs3mEcckI4PHAPBxjPTFKonzJ9BSWpN8QrjRtc8YeE9V8JXFlr+upcxpdW9r8ouihVwWHO0cMCTnAPJwK6UXvi3Tdfhl8V/EXQ9LkeQMdIhgSX5T0TBw/Prz9TXm7+MvBnhbxp4evfBGl3L22lSTG7uJuJLsSIE4zz8o3YBwMnoOtamtax8Ll8RT+MrO+1HUtTaX7VHpLxskfn5yCzlfug84DH8elRyuyVhWLHj+ebwj8fbWbw7K1g2qR24uvKAxIHlwwx2zsUnHemfHbxZrtj4wuNDtNSli0y4so/NtgBtbdnPbPOKo/E/wAW+Etf1zR/FGg6hcXWpQPCJrJ4WjVI0LP94rjdkgcEirHxK1n4c+M45/EFvreoLrP2QRw2It2VSwzjcSmO/OG7U4rZtAulzx6ve/AfiN/Cf7Ot9rMEEc9xb3riBZBlVkZkUMfpnP4V4JXvHga50S1/Z2u38U2sl1pbagUnSI4cZdAGHI5BIP4VdXZFT2K3ws+Juu+KvGH/AAjniyWPVbHVIpV2yQIPLIQsfugfKQpGDnnHTmrHwl0qPQvid410u3YNDaQyxRnOflEny598Yz71S0rxL8MPhzFc6p4RuL/W9YliKQi5RlEOexJRQBwMkZNYnwo8b6ZovibX9T8VX/2eTUrd/wB55Lv5krPuPCA471DV07IlrexynhLxJ4s024XSfCN7cRS30wAggCkyOeB1Fe1+JvHOofDLwVFp2p6w2seLb5N5L4KWoPfAxwOcZ+8cnoMVxvwY8R+CPClje6j4hvxa63JI0ULPbyyhIdqnI2qQCW3A85wO3dmqab8Lta1S41HVPiHqdxd3D75JXsJMsf8Av1wAOAOwFOVnLVDer1RzXgbX/HNxdWvhrwdqU8QkclIlC7I88s7Eg4A6mvSvih4+fwt4VPgu21N9X1i4gMepXswB2I4IZcDgMQcAdl9zmqnw58UfDzwj4Uu4U15rLWLx5UkuzZyyOiB2ERXCFfubWx6nnpiuavtC+Fc/2m5Pj3Uri7k3SZkspSZHOTyTH3Pek7OWq/AN3sdB4Gj8ax/Dey+wanpnhHRY3d21G82+ZdFmJBAYYC9hnBOOCRWv48jh1r4EX97fa9aeJrvTLqPytRt4BFtJkjUqcEgna/UYzkcZGa5iz8W+CfGHw30rw740v7zR7nSAFimgiaRZABtB4Vv4euQORwa0rHxj8N7nwFqfgiW7vtK0tZE8i7aFpJbrDK5c4U7SWXGCAMY6dAmne9uotb3PDqKKK6TUKvTfa/8AhHbPzNv2P7VP5WPveZsi359seXj8ao1emhuF8O2c7zbrZ7qdI4f7jqkRZvxDIP8AgNAFGtjwh/yPGhf9hG3/APRi1j1seEP+R40L/sI2/wD6MWlLZjjujd+K+rajefEPWbO81C6uLW1vXEEEszMkX+6pOF/CuKrqvid/yVDxD/1+vXK1MPhRU/iYUUUVZBveFvGGoeD57i40iGzNzOgUT3EAkaHB6pngH8D0rrvjlcS3fiTw/cXDb5ZvD9tI7YxuYvISfzrzOvSPjV/yGvDX/YuWv/oUlZNLnTNYt+zaPN6KKK1MgooooAKKKKACiiigAooooAK7r4ceN9W0HU9P0bSltYY73U4vPuPs6tMysyKU3nOFwOwzyea4Wtjwh/yPGhf9hG3/APRi1MknHUqDalobPxa/5Kxr/wD18D/0Ba46ux+LX/JWNf8A+vgf+gLXHUofAh1PjYUUUVZAUUUUAFFFFABRRRQAUUUUAereO9d1LwJpnh/wr4ZvZtNSLT47y8ltX8t7ieTJJLDnAx0z0OOwqt8QZ28T/DTwt4uvQp1N3lsLyYKB52wnYxwOuFJ/Gk+L1pNqj+HvE9lE8un6jpUMYkUFgkq5DIT2PI+pB9KZ4zSTQfg/4T8PXytFqE002oywNw0SEkJuHUEhv0Nc8bWi+p0yveSex5tRRRXQcwUUUUAFFFFABRRRQAUUUUAa/hPSF17xhpOlS/6u7u44pMHHyFhu/TNdt4x+J3iDTfiFcxeHtQlsNN0e4Nra2MR2wFIztwyDhgSp69AeMVyHgXUodH8faJf3JCwQ3sRkYnG1SwBP4A5/Ctbx74S1W3+Kepadb2U88t/evLZhEJ85ZG3DaehxuwT2waylZz17GsW1D3e5a+MunWln49F5p8Qhh1Wziv8AYowAz5DEfUqT9Sa4GvQfjRdQv45g06CXzf7J06CxdgcjcoLEfhux9c159VU/gQqnxsKKKKszCiiigAooooAKKKKACu7+EFjZT+MrjUtTiWa30XT59SMbgFWMeMZz6Fs/UCuEr0L4NeTdeKNV0aaTy21nRrmxiycZdgp/krVFT4GaU/jRf8CfErxBqnxHtbbxDqUl9p+sTfZLmzmO6HEnygKnReSOnXvXBeJ9MXRfFmraZGrLHZ3ksKbs52q5Cnn2ANdJ8N/DGoz/ABU062urSa3Ol3a3N4XXaIBGd/zE9ASoHvmuf8W6muteM9Y1KJt8V1eyyRn/AGC52/pipikp6dhybcPe7mPRRRWpkFFFFABRRRQAUUUUAFFFFABXR+GPHOreD7W6j0NLSKe5IJu5LdZJo8AjCluAOfSucopNJqzGm07o9I+PX/JWLv8A694f/QK83r0j49f8lYu/+veH/wBArzeopfAi6v8AEYUUUVoZhRRRQAUUUUAFeleDbw+EvhLrfirTyser3V8ml2twVBMI2iRyuehIJ/Ielea16ToFk/iH4C6zptgjz32l6umomCPljG0QjyB3xhzx6VnU21NKe7sXfCOu6h448D+LPD/iO6m1SS0099TsXuGMksbx8kBjknOVGPcjvXlNek/DaCXRfCPjDxRch4Lf+yZNPtpWGFkllIAC56kFV/OvNqIWUnYc7uMbhRRRWhkFFFFABRRRQBs+F/FF54S1STUNNgtJbhoTErXUIkEeSDuUHo3y4z6E13Pxh1C61XQPA9/qEvnXVzpXmyybQNzHaScDgfhXltek/FH/AJE74f8A/YGH8krKSXOmaxb5JI82ooorUyCiiigAooooAKKKKACiiigArrvBHjjVfC8q2WjpaRNd3Kb7prcPMASBtVj0H4d65Grek/8AIasf+viP/wBCFKSTVmVFtO6Oy+Nv/JYtb/7d/wD0njrgq7342/8AJYtb/wC3f/0njrgqmn8C9Cqnxv1CiiirMwooooAK9B+C2q6hb/EzSdPt7+5isriVzNbJMyxykRNgsoOD0HX0rz6u2+Dn/JXNC/66Sf8Aop6ip8DNKfxo5TVv+Q1ff9fEn/oRqpVvVv8AkNX3/XxJ/wChGqlUtiHuFFFFMQUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFd5beONNh+Cl34PaC6OoTXYnWUIvlBd6NgndnOFPauDopNJiauFFFFMYUUUUAFFFFABRRRQAUUUUAFXZbZV0G1uROWeS5mjMOeECrEQ2PfeR/wGqVW5Y7QaJbSRvm8a4mWVM9Iwsew49yZPyoAqVpeHLuGw8VaVeXb+XBb3sMsr4J2qrgk4HJ4Has2ih6jWjO2+JreG7/AMRXOt+HPEX9qPqFy8ktt9hkh+zg9Pmf734AVxNFFKKsrDk+Z3CiiimSbfhrxjr3g+aeXw5ffY3uFCynyY5NwByPvqcde1btz8ZfHl5aTW1xru+GZGjkX7HAMqRgjITPQ1w9FS4Rbu0UpySsmFFFFUSFFFFABRRRQAUUUUAFFFFABXTeG/iJ4p8I6bJYeHtU+yW0kpmZPs8UmXIAJy6k9FH5VzNFJpNWY02ndHWa78T/ABf4l0iXS9b1f7TZylS8X2aFMlSCOVQHqB3rk6KKEktgcnLVhRRRTEFFFFABRRRQAUUUUAFFFFAHSeHviF4q8KWT2mg6xLa2zEsYiiSKCepAcHH4YrH1XVr/AFzUpdQ1e6ku7qU5eWQ5J9vYewqnRS5Unew+ZtWuFFFFMQUUUUAFFFFABRRRQAUUUUAFddZfFTxtp+kpptp4guEtY02ICiMyr6BypYfnxXI0UnFPcak1sx0kjyyNJK7O7kszMcliepJptFFMQUUUUAFFFFABRRRQAUUUUAFSW9xNa3MdxbSvDNEweORGwyMOQQR0NR0UAdZq3xQ8Za5pL6ZqeuzS2ki7XjWNELj0ZlUEj6nmuToopJJbDcm92FFFFMQUUUUAFFFFABRRRQAUUUUAFdD4a8eeJPB8M8PhzUfsaXDBpR5Ecm4gYH31OOvaueopNJqzGm07o7LU/i1421nS7jTtS1rzrS5Qxyx/ZIV3KeoyEBH4VxtFFCio7IHJy3YUUUUxBRRRQAUUUUAFaOh+INV8NakL/Qr2SzuQpXfHg5B7EHgj2IrOoo33Gm1qje8SeOPEfi4RL4h1SW8SE7kj2qiKemdqgDPvisGiikkkrIG23dhRRRTEFFFFABRRRQBr+G/FOs+EdSkv/D159kuZIjCz+UkmUJBIw4I6qPyrpH+NXxAkjZH1/KsMEfYrfkf9+64SipcIt3aKU5JWTCiiiqJCiiigAooooAKKKKACiiigAro/DXj/AMTeELSa28O6l9jhnfzJF8iOTc2MZy6k9K5yik0mrMabTujsNX+K/jTXdJuNM1XWfPs7ldssf2WFdwznqqAjp2NcfRRQoqOyByct2FFFFMQUUUUAFd98Kbnwzo3iS28QeIvEX9nTWMzeXZ/YZJfPUoVzvTIXlumD0rgaKmS5lYqMuV3LGoypPql1NEdySTOynGMgsSKr0UVRIUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABVuWS0OiW0caYvFuJmlbHWMrHsGfYiT86qVdluVbQbW2EBV47maQzY4cMsQC59thP/AqAKVFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFSW8DXN1FBGQGlcIpboCTio6t6T/wAhqy/6+I//AEIUAa/jXwVqPgTWodM1ea1mmlt1uFa1dmUKWZcHcqnOUPb0qDwl4VvvGXiCPR9Llt4riRGcNcMyphRk8qCf0r2X41fDvxT4t8aWl/4e0v7XbR6ekLP9oijw4kkJGHYHow/OqPwl+Gfi7wx8QbbUtc0n7LaJDKrSfaYnwSuBwrk/pWPtFyXvqZ83u3PFtSsZdL1W7sLhkaW1meFyhJUsrFTjOOMiq1ekaX4C/wCE9+I3iy2/tL7B9juLifd5Hm7/AN8RjG5cfWuL8MaL/wAJH4o0/R/tH2b7bOsXnbN+zPfGRn860UkXcd4W8OXfi3xJa6JpskEVzdb9j3DEINqM5yQCeintWhpHgPU9a8dXPhS1ntEv7aWaJ5JHYRExEhsEKTjjjj8q7HwJ4d/4RL9o+y0P7V9r+yNKvneXs35tXb7uTjrjrV7wD/yc5rH/AF+6h/6G9RKb1t2Jcux5HqVjLpeq3dhcMjS2szwuUJKllYqcZxxkVWr0mz+Hd3408ceJ7qe8TStIs765a41CZNyKfMY7QMgE9zyMD8M8DqsFla6pPBpd699axttjuXh8oyD125OB6c/l0q1JPQpO5UoqW2tpr27itrSJpp5nCRxoMlmJwAK9NHwbstOWC38V+NtL0bVJlDrZMokKg9NzF1x+WOuCaHJLcG0jy2iuu1r4ba5ovja18NTCKSe+dRaXAYiKZScBs9sdx1Hvxnq4PgtpUtz/AGc3xB0j+2M7PsUahvn/ALud4Of+A59qTnFC5keTUV3XhvwSln8TjoXizU4dHlsHSUM6eYty4dNsa8j7wbIPt0rvPjh4a0S81y81i68U29pqMNknl6U0GXlxnGG3Drn0PSlzq9g5lex4TRRXo/g/4P3XjLwcut2WrQ27/ajC8E0WFRFxucvu7A9Nv41bkoq7G2luecV0Pg7wTqvji/ubTRTbrJbQGdzPIVBGQABgHkk/T1IrtpPghFf6bczeEPGGna/d2ozJawqqnPoGDt15xkAH1rC+GXhDUPE2sapb2Wt3OhT2loxleFG3SLkBozhlIHqDnp0qHNNNpi5lbQ4Wiuv8IeEvD/iDTZrjW/GNroM0c2xIJoPMLrgHdnevckfhXaWnwM0jUNFuNV07x5BcWVuGLzrYYjG0ZPzGXFNzitwckjxyiu58M+CPDOtaDDfat48s9HupCwezltt7JhiAc+YOoGeneug1b4K6fp/gm78TWnjKK8tIYGliP2HYs5HAUMZO54HB5oc4p2DmR5NRXfeGfhhDq3hy113X/FGm6BYXjOtsbghpJCrFW+Uso6qe5qHxt8MrjwppNvrOn6rba3o9wwRby2GArHOMgEjHHXPXinzxvYfMr2OHoooqhhV6aa4bw7ZwPBttkup3jm/vuyRBl/AKh/4FVGr032v/AIR2z8zb9j+1T+Vj73mbIt+fbHl4/GgCjRRWl4ctIb/xVpVndp5kFxewxSpkjcrOARkcjg9qHoNaszaK7b4mr4b0/wARXOieHPDv9lvp9y8ctz9ukm+0AdPlf7v4E1xNKLurjkuV2CiiimSFFb3ha08MXE9w/i/UryzgiQGKKyhDyTnPIBPC49/Wtr4q+HdG8N65pMPh2CaG1u9JhuyJpC7szu/J7A4C8Djip5lzcpfK+XmOHoooqiAooooAKKKKACiiigAooooAKKK7PwLpPgzUXt08U32pG+uL1LaCwsowAysVAdpG4AyxBA5wvvSk7K5UVzOxxlFdJ8QtIstB+IGraXpcRhtLaUJEhYtgbQep5PJrm6E7q4mrOwUUUUxBRRRQAUUUUAFFFFABRRRQAUV6V/Y/hHwN4b0ifxdo8+u6vq8AuxaLdtbpbQE/KSV5LHn8QRxjnL8f+GNJ0/T9H8ReFDMNF1mJikMzbmt5UOGQnv8A/WPNQppuxo6bSucTRRRVmYUUUUAFFFFABRRRQAUUUUAFFWLCym1LUraxtFDT3UyQxAnGWYgD9TXpupW/w08Ia6vhnVtEvNWmt9sd/qy3jx+XIQCSkSnDAZ6HnjvUylbQuMbq55VRXR+PfCh8G+L7nSkmNxbgLLbTHGZImGVJx36g/Sucpppq6JaadmFFFFMQUUUUAFFFFABRRRQAUUV0fgTwofGXiyDS3mNvbBWmupxjMcSjLEZ79B+NJtJXY0m3ZHOUV6vo9t8NPGOuN4Z0vRL3Sp5wyWOqtePIZJACRvjPCg47fpXl99ZzadqNzZXQAmtpWhkAOcMpIP6ipjK7sVKNlcgoooqyAooooAKKKKACiiigAooooAKKK6Twta+EHt7m58YahqEbRMBDZ2EILzDByd7fKOwpN2Q0ruxzdFdp8V/Dmm+FfHs+laLC0NrHBEwVnLkkrknJri6IvmV0OUXF2YUUUUyQooooAKKKKACiiu38D+GdIuND1XxV4tE76PpZWJbaBtr3UzdEz1AGQT9evBpSkoq5UYuTsjiKK9Kn0Xwn4z8FatqvhHSZtC1PRUE89m921xHND3YM3OQAT29Oc5HmtKMrhKPKFFFFUSFFFFABRRRQAUV13w+8K2fiK/v73XJpIdF0e2N3etF99wOiL7tg/lXS6ZpPgj4hW2paZ4Y0O50HWba3e4s2e9edLoJ1Vg33Sc9B9c8YqHNJmkabaPLKKKKszCiiigAooooAKKKKACiiigAooruvCHhzQ4PCN94w8YxT3OnwXC2dpYwyeW11MRk5bqFA549D6YMykoq5UYuTscLRXo2qaH4X8UeAb7xH4O02fR7zSJEF9pz3LTqY34Dqzc9c+nQ8dK85ojLmCUeUKKKKokKKKKACiiu++FNt4Z1nxJbeH/EXh3+0Zr6ZvLvPt0kXkKELY2JgNyvXI61MpcquVGPM7HA0VY1GJINUuoYhtSOZ1UZzgBiBVeqJCiiigAooooAKKKKACrek/wDIasv+viP/ANCFVKkt52trqKeMAtE4dQ3QkHNAHsvx/wBd1fS/H1lDpuq3tnE2mRuY7e4eNS3myjOARzgDn2rO+CfiHWtR+J1rb6hq9/dQmCYmOe5d1JCHHBOK4nxr411Hx3rUOp6vDawzRW626raoyqVDM2TuZjnLnv6VB4S8VX3g3xBHrGlxW8txGjIFuFZkwwweFIP61lye5bqRy+7Y9c+ExD/FzxxApHmyC52Lnk/v8f1H51wXwm0PULz4raZHHbSKdPuPNutykeSEzkN6HIx9awtO8Y6vpPjCXxJpsqW99NNJK4VcxtvJLKQTyvPTPpznmus1D47+L7+W2ZV0+0WGdJnS2gZRcFSCFclidpxyARScZK9uoWfQ6m1/5O9b/ro//pEapeAf+TnNY/6/dQ/9DeuFj+IerR/Ec+NVt7P+0ixbyij+TzF5fTdu+6f73X8qi0jx5qei+OrnxXawWj39zLNK8ciMYgZSS2AGBxzxz+dHI7fIXK/wPZb7UdO+J+n+KfBGmwf2XeabcvNZIr7FuWRjuLgHnL5Jz/eU9RXzxPBLa3MlvcRtFNE5SRGGCrA4II9c1r2HizUtM8ZnxNZGOO+Ny9wVAPlkuSWXGc7TkjGc470zxT4kn8WeIJtYvbS0tbmcDzVtFZUcgY3YZm5xj8vrVRi4u3QqKsbnwgntrf4s6FJesqxea6gt03tG6p/48Vr0f4j+M4vDPjC7h1n4daLeiRg0OoXMCsblccHcUOSAMEZ4xXgqsyMGQlWU5BBwQa9L0z4+eMdO0+O0lGn6h5YAWa7hZpOOmSrKCfcjNKcW5XFKOtzo5rnVvij4s8KWOtaNJ4Yt1+0PbzwyMkkkaqu5UBAx91QDjGCcdMU+01bwzpfxEh8KeF/h5BPc296IGvL5jJIpDYMoDAkAfeByOOwrzXWfiP4n1zxFaa3d6iY7yyYtamFAqwZ6gDuDjnOc963dV+OHi3VdOktQLCyklUJJd2kBWZ19CxYjH0AqeSQuVm58XP8Akvelf9un/oys39oEEfE/kdbGLHvy1YXi34n674yt7KPU4rKGSzkEiT2sTJIzAcEksfXPGOa0dZ+NfibXvDs+kahb6YyXEXkyXC27CUqevO7Az7CnGMlYaTVjzyvbfDcOo3H7L2qw6PHNLcSXbBkgUlmj3pvGBzjbnPtmvEq9w8Ka7qHhv9m261XR5/Iu7fUwUfaCMGSMEEHqCCRVVNl6hI574CafqcvxJivbVJFsreGVbyTb8pBUgIT67tpx/s11vwwuLW7+LXju40/H2aRZ2jKnIYGXqPY9fxriNd+OHi/XNJk08tZ2EcqlZZLKJkdweoyWOM+2K57wb441LwPd3dxpMFrM93B5EguUZgFznI2sOalxlK7YnFu5e+HHw9u/Het7W3W+lWxDXl102j+6p6bj+g5Pvu/FD4g2d7axeEPBoWDw9YgRs0XS5Zf5qDzn+I8+lZfhH4ua54M8Pto2m2OmT2zSNIxuoXZiWABBw4GOPStT/he+uf8AQveG/wDwCf8A+OU2pOV7Dad7lT4X/DpfEs0uueImFp4c0/LzyyNtExXkqD/dH8R/Ac9IPif8RW8Y6glhpS/ZtAsTttYFXaHxwHI7ccAdh9TVvSfjh4h0fQodIg0zRprWLOFmt5Gzli3OJAOp9KLz43a1e2M9rJoPh1EnjaNmSzcMAwxkHzOvNK0ua7Qa3uzfbSPC/gH4ceH9d17RpvEt9qcAeBZpmWC3Dr5mzHIH3j2JJ3HitrxPeDUf2a7q9TQrbQop543itLdNq7POUB+g5OM5x0xXnHhr4v8AiXwzoa6REtlqFlH/AKqO/hMnlDOcAhhwD0BzirNn8bfFtquorcGyv0v5fNZLuEsITgDCAMMABRwc9M9SSU4SuLlZ55RRRW5oFXpobhfDtnO8262e6nSOH+46pEWb8QyD/gNUauy2yroNrcics8lzNGYc8IFWIhse+8j/AIDQBSrY8If8jxoX/YRt/wD0YtY9bHhD/keNC/7CNv8A+jFpS2Y47o0vid/yVDxD/wBfr1ytdr8V9J1Gz+Ies3l5p91b2t1euYJ5YWVJf91iMN+FcVUw+FFT+JhRRRVkBXpHxq/5DXhr/sXLX/0KSuS8Nf8ACK+dP/wmP9seVtHkf2V5W7dnnd5nbp0ruPGPir4b+LBbzyxeKYrux09bK2CrbiMhNxQvyT1bnGOOgrKTfMtDWKXI9TyyiiitTIKKKKACiiigAooooAKKKKACtjwh/wAjxoX/AGEbf/0YtY9df4KvvBWlyQX/AInXX21K1vFmgGnCEwlV2su7fhs7g2cdsVMtio7jvi1/yVjX/wDr4H/oC1x1d94/13wL4mutQ1jSE8Qx6zdyK4W6EAtx0B4UlvujjnrXA0ofCh1PidgoooqyAooooAKKKKACiiigAooooA9G+NoP/CYaaw/1baPbGIjoV+bp+Oadrh2/s7+GVk+++qXDR5P8I3g/rUdv4z8KeIPC+maZ4+07VHutJj8i1vdLdN0kXZHDnHGMZ5/DmsPxt4uh8SzWFnpVkdP0bSoTBY2rNuYAnLOx7sxAz9PqThFPSNtjeUlrK+5y9FFFbmAUUUUAFFFFABRRRQAUUUUAdR8NGRfif4eMpUL9vjA3Hvnj9cVT8bpLH4/8QLOCJP7SuC2e/wC8bmsi0uprG9gu7VzHPBIskbj+FlOQfzFel33jT4e+INVi8Q+IvD+qnWcK1zbW0sf2S5dQBlsncAcDgD65rOV1K9jWNnHluVvjXkeJ9FRwRMmhWqzZOTvy+f6V5xWv4p8R3fizxLd6zqAVZblgRGn3Y1Awqj2AA/nWRTgmopMmbUpNoKKKKsgKKKKACiiigAooooAK9C+ECyPfeKo7b/j5fw1eLBzzvymMe9ee1ueDvFNz4O8UW2sWkazeXlJYGOFmjYYZT/nqBUTTcWkXBpSTZP8ADtJn+JXh0W2d/wDaUBODj5Q4Lf8Ajuab8QHhk+JHiJrZcRnUp++cnzDk/icn8a6y38Z+AvDeoT674P0PVBrTK32aO/eM21mzDBZApLNgE8H9K81mmkuJ3mmcvJIxZ2PViTkmlG7lzWHKyjy3GUUUVoZhRRRQAUUUUAFFFFABRRRQAUUV0Phr/hDPJn/4TH+3fN3DyP7K8nbtxzu8zv06Um7IaV3Y6f49f8lYu/8Ar3h/9ArzevTfiD4q8AeM7281iGLxJFq8kKpCrLAtvuUYG4ZLY9cGvMqinfkSZdW3O2gooorQzCiiigAooooAK9CgWR/2cLnyOieJQ0+D/B9nAH4biK89rsfBHjKx0Ox1LRPElhJqOg6oqmeGJsSROv3XTPGfy6D0wYmnbQuDV7M1PhGdkfjKW4Vmsl8N3QmUNjcTtwPqQGxXnVd5qnjDw7pfhO80DwDY6hCmpMv2++1NkM0iKciNQnygZJ59PrXB0RvdscrWSQUUUVZmFFFFABRRRQB6F4IWR/hJ8Q1tv9d5VixwedglYv8A+O5pPgcJf+FuaY8ZIjSOdpjnACeS459txWsXwJ4wHhLVbg3lp9v0vUIGtb60Jx5kZ7j/AGh2+pHGc1vHxl4R8M6ZqI+H+m6qupajE1u13qjxn7NE33ljCH6cnn3NYyT95W3N4te677Hn1w0TXMrW6FIi5KKTyq54H5VHRRWxgFFFFABRRRQAUUUUAFFFFABXo2rgn9nbQCn3F1iYSY7NtbGfwrzmu18HeMtL0/w9f+GfFthcX2iXsgnBtXCzW8owN6ZIB4A6+nuRUTT0a6GkGtU+pp/DQ7PAnxBeT/Vf2WinnA3EuF/WvN67jX/F+hQeEX8MeB7G9trC6mWa+utQZWnuCv3VwvyhQQDx3/HPD0RTu2+oTaskugUUUVZmFFFFABXbfBz/AJK5oX/XST/0U9cTXoPwW0rULj4maTqFvYXMtlbyuJrlIWaOImJsBmAwOo6+tRU+BmlP40cTq3/Iavv+viT/ANCNVKt6t/yGr7/r4k/9CNVKpbEPcKKKKYgooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAK2I/FetReFJPDSXmNIlk817byk5bIOd2N3UDvWPRRa4BRRRQAUUUUAFFFFABRRRQAUUUUAFW5Y7QaJbSRvm8a4mWVM9Iwsew49yZPyqpVuWS0OiW0caYvFuJmlbHWMrHsGfYiT86AKlPhmlt545reR4pY2DpIjFWVgcggjoQe9MooA0NT8QazrSRprGr3+oLGSUW6uXlCE9SNxOKz6KKNgu2FFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAVo6b4i1vRoXi0fWL+wjkbc6Wt08QY+pCkZNZ1FG47tbCu7SOzyMWZjlmY5JPrSUUUCCiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKuy3KtoNrbCAq8dzNIZscOGWIBc+2wn/AIFVKr001w3h2zgeDbbJdTvHN/fdkiDL+AVD/wACoAo0UUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABV6b7X/AMI7Z+Zt+x/ap/Kx97zNkW/Ptjy8fjVGr00Nwvh2znebdbPdTpHD/cdUiLN+IZB/wGgCjRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFWHEP9mQFZ3afzpN8BB2ou1NrA+pO4H/dFV6KACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooA/9k="
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![thor_hulk.jpg](attachment:467b1a02-0d94-4efb-9c1f-9a122d2bc309.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### News Category Classification Problem"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Okay now that we know basics of BAG of n grams vectorizer 😎 It is the time to work on a real problem. Here we want to do a news category classification. We will use bag of n-grams and traing a machine learning model that can categorize any news into one of the following categories,\n",
    "\n",
    "1. BUSINESS\n",
    "1. SPORTS\n",
    "1. CRIME\n",
    "1. SCIENCE"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Dataset\n",
    "\n",
    "Dataset Credits: https://www.kaggle.com/code/hengzheng/news-category-classifier-val-acc-0-65\n",
    "\n",
    "- This data consists of two columns.\n",
    "        - Text\n",
    "        - Category\n",
    "- Text is a news article\n",
    "- Category can be one of these 4: 'BUSINESS', 'SPORTS', 'CRIME', 'SCIENCE', to keep things simple I trimmed additional categories from the original dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(12695, 2)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>category</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Watching Schrödinger's Cat Die University of C...</td>\n",
       "      <td>SCIENCE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>WATCH: Freaky Vortex Opens Up In Flooded Lake</td>\n",
       "      <td>SCIENCE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Entrepreneurs Today Don't Need a Big Budget to...</td>\n",
       "      <td>BUSINESS</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>These Roads Could Recharge Your Electric Car A...</td>\n",
       "      <td>BUSINESS</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Civilian 'Guard' Fires Gun While 'Protecting' ...</td>\n",
       "      <td>CRIME</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                text  category\n",
       "0  Watching Schrödinger's Cat Die University of C...   SCIENCE\n",
       "1     WATCH: Freaky Vortex Opens Up In Flooded Lake    SCIENCE\n",
       "2  Entrepreneurs Today Don't Need a Big Budget to...  BUSINESS\n",
       "3  These Roads Could Recharge Your Electric Car A...  BUSINESS\n",
       "4  Civilian 'Guard' Fires Gun While 'Protecting' ...     CRIME"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "df = pd.read_json('news_dataset.json')\n",
    "print(df.shape)\n",
    "\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "category\n",
       "BUSINESS    4254\n",
       "SPORTS      4167\n",
       "CRIME       2893\n",
       "SCIENCE     1381\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.category.value_counts() # gives category count"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Handle class imbalance**\n",
    "\n",
    "As you can see above, SCIENCE category has almost 1/3rd data samples compared to BUSINESS and SPORTS categories. I initially trained a model without handling the imbalanced I saw a lower f1-score for SCIENCE category. Hence we need to address this imbalanced. \n",
    "\n",
    "There are various ways of handling class imbalance which I have discussed in this video: https://www.youtube.com/watch?v=JnlM4yLFNuo\n",
    "\n",
    "\n",
    "Out of those techniques, I will use **undersampling** technique here. \n",
    "\n",
    "In undersampling, we take a minor class and sample those many samples from other classes, this means we are not utilizing all the data samples for training and in ML world - Not using all the data for training is considered a SIN! 😵 In real life, you are advised to use a technique such as SMOTE so that you can utilize all of your dataset for the training but since this tutorial is more about bag of n-grams then class imbalance itself, I'd go with a simple technique of undersampling. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Basically we are using science sample number as our base for all samples. called undersampling only doing it for this case we do not generally waste data. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "min_samples = 1381 # we have these many SCIENCE articles and SCIENCE is our minority class\n",
    "\n",
    "# change random state according to you for every random state the samples picked are different\n",
    "\n",
    "df_business = df[df.category==\"BUSINESS\"].sample(min_samples, random_state=2022)\n",
    "df_sports = df[df.category==\"SPORTS\"].sample(min_samples, random_state=2022)\n",
    "df_crime = df[df.category==\"CRIME\"].sample(min_samples, random_state=2022)\n",
    "df_science = df[df.category==\"SCIENCE\"].sample(min_samples, random_state=2022)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "category\n",
       "BUSINESS    1381\n",
       "SPORTS      1381\n",
       "CRIME       1381\n",
       "SCIENCE     1381\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_balanced = pd.concat([df_business,df_sports,df_crime,df_science],axis=0) #concate adds the data frames.\n",
    "# axis = 0 --> at row level , axis = 1 --> at column level.\n",
    "df_balanced.category.value_counts() # gives value counts."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Convert text category to a number**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "target = {'BUSINESS': 0, 'SPORTS': 1, 'CRIME': 2, 'SCIENCE': 3}\n",
    "\n",
    "df_balanced['category_num'] = df_balanced['category'].map({\n",
    "    'BUSINESS': 0,\n",
    "    'SPORTS': 1, \n",
    "    'CRIME': 2, \n",
    "    'SCIENCE': 3\n",
    "}) # adding classification category."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>category</th>\n",
       "      <th>category_num</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>11967</th>\n",
       "      <td>GCC Business Leaders Remain Confident in the F...</td>\n",
       "      <td>BUSINESS</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2912</th>\n",
       "      <td>From the Other Side; an Honest Review from Emp...</td>\n",
       "      <td>BUSINESS</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3408</th>\n",
       "      <td>Mike McDerment, CEO of FreshBooks, Talks About...</td>\n",
       "      <td>BUSINESS</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>502</th>\n",
       "      <td>How to Market Your Business While Traveling th...</td>\n",
       "      <td>BUSINESS</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5279</th>\n",
       "      <td>How to Leverage Intuition in Decision-making I...</td>\n",
       "      <td>BUSINESS</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    text  category  \\\n",
       "11967  GCC Business Leaders Remain Confident in the F...  BUSINESS   \n",
       "2912   From the Other Side; an Honest Review from Emp...  BUSINESS   \n",
       "3408   Mike McDerment, CEO of FreshBooks, Talks About...  BUSINESS   \n",
       "502    How to Market Your Business While Traveling th...  BUSINESS   \n",
       "5279   How to Leverage Intuition in Decision-making I...  BUSINESS   \n",
       "\n",
       "       category_num  \n",
       "11967             0  \n",
       "2912              0  \n",
       "3408              0  \n",
       "502               0  \n",
       "5279              0  "
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_balanced.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Build a model with original text (no pre processing)**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    df_balanced.text, \n",
    "    df_balanced.category_num, \n",
    "    test_size=0.2, # 20% samples will go to test dataset\n",
    "    random_state=2022,\n",
    "    stratify=df_balanced.category_num # creates equal number of samples for train and test split form every class\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4419,)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "7589     Ovulating Women Prefer Images of Penetration O...\n",
       "10442    Scientists Discover Spooky Influence On Baby N...\n",
       "8792     Olympic Race Walker Steps Up To Propose To His...\n",
       "1733     Beloved Bipedal Bear Named Pedals Believed Kil...\n",
       "2526     Elizabeth Smart Gave Birth To Baby Girl, Fathe...\n",
       "Name: text, dtype: object"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(X_train.shape)\n",
    "X_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "category_num\n",
       "3    1105\n",
       "2    1105\n",
       "0    1105\n",
       "1    1104\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "category_num\n",
       "1    277\n",
       "0    276\n",
       "3    276\n",
       "2    276\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Attempt 1 : Use 1-gram which is nothing but a Bag Of Words (BOW) model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.75      0.87      0.81       276\n",
      "           1       0.93      0.80      0.86       277\n",
      "           2       0.83      0.90      0.86       276\n",
      "           3       0.90      0.80      0.85       276\n",
      "\n",
      "    accuracy                           0.84      1105\n",
      "   macro avg       0.85      0.84      0.84      1105\n",
      "weighted avg       0.85      0.84      0.84      1105\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.metrics import classification_report\n",
    "\n",
    "#1. create a pipeline object\n",
    "clf = Pipeline([\n",
    "     ('vectorizer_bow', CountVectorizer(ngram_range = (1, 1))),        #using the ngram_range parameter \n",
    "     ('Multi NB', MultinomialNB())         \n",
    "])\n",
    "\n",
    "#2. fit with X_train and y_train\n",
    "clf.fit(X_train, y_train)\n",
    "\n",
    "\n",
    "#3. get the predictions for X_test and store it in y_pred\n",
    "y_pred = clf.predict(X_test)\n",
    "\n",
    "\n",
    "#4. print the classfication report\n",
    "print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3716     African Nation Slaps Exxon With Fine Nearly 7 ...\n",
       "608      These Cringe-Worthy Stories Show It Can Be Har...\n",
       "11172    LISTEN: The Accidental Discovery That Proved T...\n",
       "1346     Build Loyalty -- The Cost -- $00.00 Remember y...\n",
       "1356     Man Killed By Michigan Police Wasn't Targeting...\n",
       "Name: text, dtype: object"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test[:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- BUSINESS: 0\n",
    "- SPORTS: 1\n",
    "- CRIME: 2\n",
    "- SCIENCE: 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 3, 3, 0, 2], dtype=int64)"
      ]
     },
     "execution_count": 72,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3716     0\n",
       "608      3\n",
       "11172    3\n",
       "1346     0\n",
       "1356     2\n",
       "Name: category_num, dtype: int64"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test[:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Attempt 2 : Use 1-gram and bigrams**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.69      0.90      0.78       276\n",
      "           1       0.95      0.74      0.83       277\n",
      "           2       0.82      0.88      0.85       276\n",
      "           3       0.92      0.78      0.84       276\n",
      "\n",
      "    accuracy                           0.82      1105\n",
      "   macro avg       0.85      0.82      0.83      1105\n",
      "weighted avg       0.85      0.82      0.83      1105\n",
      "\n"
     ]
    }
   ],
   "source": [
    "#1. create a pipeline object\n",
    "clf = Pipeline([\n",
    "    ('vectorizer_1_2_gram', CountVectorizer(ngram_range = (1, 2))),        #using the ngram_range parameter \n",
    "     ('Multi NB', MultinomialNB())         \n",
    "])\n",
    "\n",
    "#2. fit with X_train and y_train\n",
    "clf.fit(X_train, y_train)\n",
    "\n",
    "\n",
    "#3. get the predictions for X_test and store it in y_pred\n",
    "y_pred = clf.predict(X_test)\n",
    "\n",
    "\n",
    "#4. print the classfication report\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "# Performance little lower than Bago of Words so sometimes Bag of words is preferred, Depends on the problem."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Attempt 3 : Use 1-gram to trigrams**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.67      0.91      0.77       276\n",
      "           1       0.96      0.73      0.83       277\n",
      "           2       0.83      0.87      0.85       276\n",
      "           3       0.93      0.76      0.83       276\n",
      "\n",
      "    accuracy                           0.82      1105\n",
      "   macro avg       0.84      0.82      0.82      1105\n",
      "weighted avg       0.84      0.82      0.82      1105\n",
      "\n"
     ]
    }
   ],
   "source": [
    "#1. create a pipeline object\n",
    "clf = Pipeline([\n",
    "    ('vectorizer_1_3_grams', CountVectorizer(ngram_range = (1, 3))),        #using the ngram_range parameter \n",
    "     ('Multi NB', MultinomialNB())         \n",
    "])\n",
    "\n",
    "#2. fit with X_train and y_train\n",
    "clf.fit(X_train, y_train)\n",
    "\n",
    "\n",
    "#3. get the predictions for X_test and store it in y_pred\n",
    "y_pred = clf.predict(X_test)\n",
    "\n",
    "\n",
    "#4. print the classfication report\n",
    "print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Use text pre-processing to remove stop words, punctuations and apply lemmatization"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You may wonder, we have not done any text-processing yet to remove stop words, punctuations, apply lemmatization etc. Well we wanted to train the model without any preprocessing first and check the performance. Now we will re-do same thing but with preprocessing of text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_balanced['preprocessed_txt'] = df_balanced['text'].apply(preprocess) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>category</th>\n",
       "      <th>category_num</th>\n",
       "      <th>preprocessed_txt</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>11967</th>\n",
       "      <td>GCC Business Leaders Remain Confident in the F...</td>\n",
       "      <td>BUSINESS</td>\n",
       "      <td>0</td>\n",
       "      <td>GCC Business Leaders remain Confident Face Reg...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2912</th>\n",
       "      <td>From the Other Side; an Honest Review from Emp...</td>\n",
       "      <td>BUSINESS</td>\n",
       "      <td>0</td>\n",
       "      <td>Honest Review employee wake morning love impor...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3408</th>\n",
       "      <td>Mike McDerment, CEO of FreshBooks, Talks About...</td>\n",
       "      <td>BUSINESS</td>\n",
       "      <td>0</td>\n",
       "      <td>Mike McDerment ceo FreshBooks Talks give build...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>502</th>\n",
       "      <td>How to Market Your Business While Traveling th...</td>\n",
       "      <td>BUSINESS</td>\n",
       "      <td>0</td>\n",
       "      <td>market business travel World recently amazing ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5279</th>\n",
       "      <td>How to Leverage Intuition in Decision-making I...</td>\n",
       "      <td>BUSINESS</td>\n",
       "      <td>0</td>\n",
       "      <td>Leverage intuition decision making feel safe r...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    text  category  \\\n",
       "11967  GCC Business Leaders Remain Confident in the F...  BUSINESS   \n",
       "2912   From the Other Side; an Honest Review from Emp...  BUSINESS   \n",
       "3408   Mike McDerment, CEO of FreshBooks, Talks About...  BUSINESS   \n",
       "502    How to Market Your Business While Traveling th...  BUSINESS   \n",
       "5279   How to Leverage Intuition in Decision-making I...  BUSINESS   \n",
       "\n",
       "       category_num                                   preprocessed_txt  \n",
       "11967             0  GCC Business Leaders remain Confident Face Reg...  \n",
       "2912              0  Honest Review employee wake morning love impor...  \n",
       "3408              0  Mike McDerment ceo FreshBooks Talks give build...  \n",
       "502               0  market business travel World recently amazing ...  \n",
       "5279              0  Leverage intuition decision making feel safe r...  "
      ]
     },
     "execution_count": 84,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_balanced.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Build a model with pre processed text**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    df_balanced.preprocessed_txt,  # Instead of Text\n",
    "    df_balanced.category_num, \n",
    "    test_size=0.2, # 20% samples will go to test dataset\n",
    "    random_state=2022,\n",
    "    stratify=df_balanced.category_num\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4419,)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "7589     ovulate woman prefer Images Penetration Images...\n",
       "10442     scientist Discover Spooky Influence Baby Choices\n",
       "8792     Olympic Race Walker step Propose Boyfriend Rio...\n",
       "1733     Beloved Bipedal Bear name Pedals believe kill ...\n",
       "2526       Elizabeth Smart give Birth Baby Girl Father say\n",
       "Name: preprocessed_txt, dtype: object"
      ]
     },
     "execution_count": 87,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(X_train.shape)\n",
    "X_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "category_num\n",
       "3    1105\n",
       "2    1105\n",
       "0    1105\n",
       "1    1104\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "category_num\n",
       "1    277\n",
       "0    276\n",
       "3    276\n",
       "2    276\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.80      0.88      0.84       276\n",
      "           1       0.92      0.82      0.87       277\n",
      "           2       0.83      0.92      0.87       276\n",
      "           3       0.90      0.81      0.85       276\n",
      "\n",
      "    accuracy                           0.86      1105\n",
      "   macro avg       0.86      0.86      0.86      1105\n",
      "weighted avg       0.86      0.86      0.86      1105\n",
      "\n"
     ]
    }
   ],
   "source": [
    "#1. create a pipeline object\n",
    "clf = Pipeline([\n",
    "    ('vectorizer_bow', CountVectorizer(ngram_range = (1, 2))),        #using the ngram_range parameter \n",
    "    ('Multi NB', MultinomialNB())\n",
    "])\n",
    "\n",
    "#2. fit with X_train and y_train\n",
    "clf.fit(X_train, y_train)\n",
    "\n",
    "\n",
    "#3. get the predictions for X_test and store it in y_pred\n",
    "y_pred = clf.predict(X_test)\n",
    "\n",
    "\n",
    "#4. print the classfication report\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "# The acccuracy increased for preprocessed text."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If you compare above classification report for (1,2) gram with the one from unprocessed text, you will find some improvement in the model that uses preprocessed cleaned up text. Hence we can conclude that for this particular problem using preprocessing (removing stop words, lemmatization) is improving the performance of the model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Plot confusion matrix**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[243,   8,  17,   8],\n",
       "       [ 10, 228,  29,  10],\n",
       "       [ 15,   2, 253,   6],\n",
       "       [ 36,   9,   7, 224]], dtype=int64)"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "cm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(95.72222222222221, 0.5, 'Truth')"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x700 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "import seaborn as sn\n",
    "plt.figure(figsize = (10,7))\n",
    "sn.heatmap(cm, annot=True, fmt='d')\n",
    "plt.xlabel('Prediction')\n",
    "plt.ylabel('Truth')"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "BOW_exercise.ipynb",
   "provenance": []
  },
  "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": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}