notebooks/first_module_intro/03_lexical_disambiguation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "9b8c190e",
   "metadata": {},
   "source": [
    "# Word Sense Disambiguation (снятие лексической неоднозначности слов) и Word Sense Induction (нахождение значений слова)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2cf62b83",
   "metadata": {},
   "source": [
    "WSD и WSI  - это связанные задачи про определение значений, в которых употребляются слова. WSI занимается определением возможных значений слова, а WSD нужна для того, чтобы определить, в каком из имеющихся значений слово употреблено в тексте. \n",
    "\n",
    "Основное предположение, на котором строятся все модели для снятия и определения неоднозначности - значение определяется в контексте. \n",
    "\n",
    "На практике WSD/WSI применяются достаточно редко. Современные модели (вроде BERT) используют фиксированные эмбединги для отдельных слов и преобразуют их в зависимости от контекста. Преобразованные (контекстуализированные) эмбединги удобно использовать для предсказания (именованных сущностей, частей речи, перевода), но из них нельзя вытащить информацию о том, сколько значений есть у слова. А это может требоваться в специфичных лингвистических задачах.   "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b8956f34",
   "metadata": {},
   "source": [
    "Для нас WSD и WSI удобны тем, что на их примере можно познакомится с важными методами и инструментами."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e189015",
   "metadata": {},
   "source": [
    "# WSD"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e778c0f9-ba43-43da-98f0-5e55d8ddb1bc",
   "metadata": {},
   "source": [
    "## WordNet"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "377c238e",
   "metadata": {},
   "source": [
    "Выводить значения просто из текста тяжело, поэтому можно попробовать воспользоваться словарями (wsi сделанное людьми). Самый известный такой словарь - WordNet (https://wordnet.princeton.edu/), который есть в nltk. Для русского есть https://ruwordnet.ru/ru , но он конечно попроще и данные нужно запрашивать."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6b2c58bf",
   "metadata": {},
   "source": [
    "WordNet - это лексическая база данных, где существительные, прилагательные и глаголы английского сгруппированы в синонимические множества. К тому же между ними установлены связи (гипонимия, гипоронимия, миронимия и т.п.)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "9dcc6d61",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package wordnet to\n",
      "[nltk_data]     /Users/mnefedov/nltk_data...\n",
      "[nltk_data]   Package wordnet is already up-to-date!\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# запустите если не установлен ворднет\n",
    "import nltk\n",
    "nltk.download('wordnet')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "139f468e",
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk.corpus import wordnet as wn"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c49e0165",
   "metadata": {},
   "source": [
    "Синонимичные множества в WordNet'е коротко называются синсеты. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "1733465f-4e2e-44e4-ba0a-b54c7a71f034",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "117659"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# всего синсетов\n",
    "len(list(wn.all_synsets()))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "06a2bf67",
   "metadata": {},
   "source": [
    "Из них существительных:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "a554f130",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "82115"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(list(wn.all_synsets('n')))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a1e72af6",
   "metadata": {},
   "source": [
    "А глаголов:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "a114a04d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "13767"
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(list(wn.all_synsets('v')))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a17fcffb",
   "metadata": {},
   "source": [
    "А прилагательных:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "id": "3859a55b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "18156"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(list(wn.all_synsets('a')))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "db754c80",
   "metadata": {},
   "source": [
    "Синсеты для слова можно достать вот так:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "id": "6aeb8b88",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Synset('go.n.01'),\n",
       " Synset('adam.n.03'),\n",
       " Synset('crack.n.09'),\n",
       " Synset('go.n.04'),\n",
       " Synset('travel.v.01'),\n",
       " Synset('go.v.02'),\n",
       " Synset('go.v.03'),\n",
       " Synset('become.v.01'),\n",
       " Synset('go.v.05'),\n",
       " Synset('run.v.05'),\n",
       " Synset('run.v.03'),\n",
       " Synset('proceed.v.04'),\n",
       " Synset('go.v.09'),\n",
       " Synset('go.v.10'),\n",
       " Synset('sound.v.02'),\n",
       " Synset('function.v.01'),\n",
       " Synset('run_low.v.01'),\n",
       " Synset('move.v.13'),\n",
       " Synset('survive.v.01'),\n",
       " Synset('go.v.16'),\n",
       " Synset('die.v.01'),\n",
       " Synset('belong.v.03'),\n",
       " Synset('go.v.19'),\n",
       " Synset('start.v.09'),\n",
       " Synset('move.v.15'),\n",
       " Synset('go.v.22'),\n",
       " Synset('go.v.23'),\n",
       " Synset('blend.v.02'),\n",
       " Synset('go.v.25'),\n",
       " Synset('fit.v.02'),\n",
       " Synset('rifle.v.02'),\n",
       " Synset('go.v.28'),\n",
       " Synset('plump.v.04'),\n",
       " Synset('fail.v.04'),\n",
       " Synset('go.a.01')]"
      ]
     },
     "execution_count": 144,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# это все значения, которые соответствуют такой словоформе\n",
    "wn.synsets('go')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "79b7dabf",
   "metadata": {},
   "source": [
    "Можно ограничить часть речи:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "id": "2ddbe807",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Synset('travel.v.01'),\n",
       " Synset('go.v.02'),\n",
       " Synset('go.v.03'),\n",
       " Synset('become.v.01'),\n",
       " Synset('go.v.05'),\n",
       " Synset('run.v.05'),\n",
       " Synset('run.v.03'),\n",
       " Synset('proceed.v.04'),\n",
       " Synset('go.v.09'),\n",
       " Synset('go.v.10'),\n",
       " Synset('sound.v.02'),\n",
       " Synset('function.v.01'),\n",
       " Synset('run_low.v.01'),\n",
       " Synset('move.v.13'),\n",
       " Synset('survive.v.01'),\n",
       " Synset('go.v.16'),\n",
       " Synset('die.v.01'),\n",
       " Synset('belong.v.03'),\n",
       " Synset('go.v.19'),\n",
       " Synset('start.v.09'),\n",
       " Synset('move.v.15'),\n",
       " Synset('go.v.22'),\n",
       " Synset('go.v.23'),\n",
       " Synset('blend.v.02'),\n",
       " Synset('go.v.25'),\n",
       " Synset('fit.v.02'),\n",
       " Synset('rifle.v.02'),\n",
       " Synset('go.v.28'),\n",
       " Synset('plump.v.04'),\n",
       " Synset('fail.v.04')]"
      ]
     },
     "execution_count": 145,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wn.synsets('go', pos='v')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b06590ca",
   "metadata": {},
   "source": [
    "У многих из них есть определения:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "d4aa4f59",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "car - a motor vehicle with four wheels; usually propelled by an internal combustion engine\n",
      "car - a wheeled vehicle adapted to the rails of railroad\n",
      "car - the compartment that is suspended from an airship and that carries personnel and the cargo and the power plant\n",
      "car - where passengers ride up and down\n",
      "car - a conveyance for passengers or freight on a cable railway\n"
     ]
    }
   ],
   "source": [
    "word = 'car'\n",
    "for synset in wn.synsets(word):\n",
    "    print(word + ' - ' + synset.definition())\n",
    "#     print()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6e088c53",
   "metadata": {},
   "source": [
    "И даже примеры:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "45f98c07",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "car - he needs a car to get to work\n",
      "car - three cars had jumped the rails\n",
      "car - \n",
      "car - the car was on the top floor\n",
      "car - they took a cable car to the top of the mountain\n"
     ]
    }
   ],
   "source": [
    "word = 'car'\n",
    "for synset in wn.synsets(word):\n",
    "    print(word + ' - ' + ' | '.join(synset.examples()))\n",
    "#     print()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "31d72334-b1d7-4d67-919c-257c3bfa1f2c",
   "metadata": {},
   "source": [
    "Каждый синсет объединяет словоформы, которые являются взаимозаменяемыми во многих контекстах. Достать словоформы для значения можно вот так:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "id": "00027625-023b-4b02-9470-9f966d28e9d2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[['auto', 'automobile', 'machine', 'motorcar'],\n",
       " ['railcar', 'railroad_car', 'railway_car'],\n",
       " ['gondola'],\n",
       " ['elevator_car'],\n",
       " ['cable_car']]"
      ]
     },
     "execution_count": 147,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# длина совпадает с длиной wn.synsets(\"car\")\n",
    "# то есть каждый список соответствует значению \n",
    "wn.synonyms('car')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "id": "56716fd0-762d-4208-9a38-e091834bcae5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[['badness'],\n",
       " [],\n",
       " ['big'],\n",
       " ['tough'],\n",
       " ['spoiled', 'spoilt'],\n",
       " ['regretful', 'sorry'],\n",
       " ['uncollectible'],\n",
       " [],\n",
       " [],\n",
       " ['high-risk', 'risky', 'speculative'],\n",
       " ['unfit', 'unsound'],\n",
       " [],\n",
       " [],\n",
       " ['forged'],\n",
       " ['defective'],\n",
       " ['badly'],\n",
       " ['badly']]"
      ]
     },
     "execution_count": 150,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# другой пример\n",
    "# пустой список означает, что для этого значения\n",
    "# взаимозаменяемых синонимов нет\n",
    "wn.synonyms('bad')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0dfa4a80-36d3-4427-a7db-7f9649fff1dd",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "cf64ddbd",
   "metadata": {},
   "source": [
    "Посмотрим на другие отношения (гипонимы, антонимы, синонимы, меронимы)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "id": "16191e3f",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Synset('accelerator.n.01'),\n",
       " Synset('air_bag.n.01'),\n",
       " Synset('auto_accessory.n.01'),\n",
       " Synset('automobile_engine.n.01'),\n",
       " Synset('automobile_horn.n.01'),\n",
       " Synset('buffer.n.06'),\n",
       " Synset('bumper.n.02'),\n",
       " Synset('car_door.n.01'),\n",
       " Synset('car_mirror.n.01'),\n",
       " Synset('car_seat.n.01'),\n",
       " Synset('car_window.n.01'),\n",
       " Synset('fender.n.01'),\n",
       " Synset('first_gear.n.01'),\n",
       " Synset('floorboard.n.02'),\n",
       " Synset('gasoline_engine.n.01'),\n",
       " Synset('glove_compartment.n.01'),\n",
       " Synset('grille.n.02'),\n",
       " Synset('high_gear.n.01'),\n",
       " Synset('hood.n.09'),\n",
       " Synset('luggage_compartment.n.01'),\n",
       " Synset('rear_window.n.01'),\n",
       " Synset('reverse.n.02'),\n",
       " Synset('roof.n.02'),\n",
       " Synset('running_board.n.01'),\n",
       " Synset('stabilizer_bar.n.01'),\n",
       " Synset('sunroof.n.01'),\n",
       " Synset('tail_fin.n.02'),\n",
       " Synset('third_gear.n.01'),\n",
       " Synset('window.n.02')]"
      ]
     },
     "execution_count": 154,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wn.synsets('car')[0].part_meronyms()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "15ff7c54",
   "metadata": {},
   "source": [
    "Антонимы только достаются не очень удобно:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "id": "29ba9bd9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Lemma('bad.n.01.bad')]"
      ]
     },
     "execution_count": 155,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wn.synsets('good')[2].lemmas()[0].antonyms()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "id": "6647bbea",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Synset('good.n.01'),\n",
       " Synset('good.n.02'),\n",
       " Synset('good.n.03'),\n",
       " Synset('commodity.n.01'),\n",
       " Synset('good.a.01'),\n",
       " Synset('full.s.06'),\n",
       " Synset('good.a.03'),\n",
       " Synset('estimable.s.02'),\n",
       " Synset('beneficial.s.01'),\n",
       " Synset('good.s.06'),\n",
       " Synset('good.s.07'),\n",
       " Synset('adept.s.01'),\n",
       " Synset('good.s.09'),\n",
       " Synset('dear.s.02'),\n",
       " Synset('dependable.s.04'),\n",
       " Synset('good.s.12'),\n",
       " Synset('good.s.13'),\n",
       " Synset('effective.s.04'),\n",
       " Synset('good.s.15'),\n",
       " Synset('good.s.16'),\n",
       " Synset('good.s.17'),\n",
       " Synset('good.s.18'),\n",
       " Synset('good.s.19'),\n",
       " Synset('good.s.20'),\n",
       " Synset('good.s.21'),\n",
       " Synset('well.r.01'),\n",
       " Synset('thoroughly.r.02')]"
      ]
     },
     "execution_count": 156,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wn.synsets('good')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7d8bc4fa",
   "metadata": {},
   "source": [
    "Можно засунуть это все в граф и нарисовать:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "65b873de",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: networkx in /Users/mnefedov/miniforge3/envs/tf2/lib/python3.9/site-packages (3.0)\n",
      "\u001b[33mWARNING: Error parsing requirements for pybind11: [Errno 2] No such file or directory: '/Users/mnefedov/miniforge3/envs/tf2/lib/python3.9/site-packages/pybind11-2.10.1.dist-info/METADATA'\u001b[0m\u001b[33m\n",
      "\u001b[0m"
     ]
    }
   ],
   "source": [
    "!pip install networkx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "0497a6b8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[Synset('accelerator.n.01'), Synset('air_bag.n.01'), Synset('auto_accessory.n.01'), Synset('automobile_engine.n.01'), Synset('automobile_horn.n.01'), Synset('buffer.n.06'), Synset('bumper.n.02'), Synset('car_door.n.01'), Synset('car_mirror.n.01'), Synset('car_seat.n.01'), Synset('car_window.n.01'), Synset('fender.n.01'), Synset('first_gear.n.01'), Synset('floorboard.n.02'), Synset('gasoline_engine.n.01'), Synset('glove_compartment.n.01'), Synset('grille.n.02'), Synset('high_gear.n.01'), Synset('hood.n.09'), Synset('luggage_compartment.n.01'), Synset('rear_window.n.01'), Synset('reverse.n.02'), Synset('roof.n.02'), Synset('running_board.n.01'), Synset('stabilizer_bar.n.01'), Synset('sunroof.n.01'), Synset('tail_fin.n.02'), Synset('third_gear.n.01'), Synset('window.n.02')]\n"
     ]
    }
   ],
   "source": [
    "import networkx as nx\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "words = wn.synsets('car')\n",
    "hyponyms = words[0].part_meronyms()\n",
    "print(hyponyms)\n",
    "\n",
    "G=nx.Graph()\n",
    "\n",
    "for w in hyponyms: \n",
    "    G.add_edge(words[0].name().split('.')[0],\n",
    "               w.name().split('.')[0], \n",
    "               weight=words[0].wup_similarity(w))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "f783af3f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAxoAAAMWCAYAAAB2gvApAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAA9hAAAPYQGoP6dpAAEAAElEQVR4nOzdd3gU5fbA8e/M9ppC0wAJgigoVsSCXHtXREREUMECiqAIFkBBVIpiQUSUpqKComKh2a/1d7l4FQuK9SrShEAISbbX2fn9kXKT0BJIsrvJ+TxPHiU7s3O2ZHfOvO97jqLruo4QQgghhBBC1CE12QEIIYQQQgghGh9JNIQQQgghhBB1ThINIYQQQgghRJ2TREMIIYQQQghR5yTREEIIIYQQQtQ5STSEEEIIIYQQdU4SDSGEEEIIIUSdk0RDCCGEEEIIUeck0RBCCCGEEELUOUk0hBBCCCGEEHVOEg0hhBBCCCFEnZNEQwghhBBCCFHnJNEQQgghhBBC1DlJNIQQQgghhBB1ThINIYQQQgghRJ2TREMIIYQQQghR5yTREEIIIYQQQtQ5STSEEEIIIYQQdU4SDSGEEEIIIUSdk0RDCCGEEEIIUeck0RBCCCGEEELUOUk0hBBCCCGEEHVOEg0hhBBCCCFEnZNEQwghhBBCCFHnJNEQQgghhBBC1DlJNIQQQgghhBB1ThINIYQQQgghRJ2TREMIIYQQQghR5yTREEIIIYQQQtQ5STSEEEIIIYQQdU4SDSGEEEIIIUSdk0RDCCGEEEIIUeck0RBCCCGEEELUOUk0hBBCCCGEEHVOEg0hhBBCCCFEnZNEQwghhBBCCFHnJNEQQgghhBBC1DlJNIQQQgghhBB1ThINIYQQQgghRJ2TREMIIYQQQghR5yTREEIIIYQQQtQ5STSEEEIIIYQQdU4SDSGEEEIIIUSdk0RDCCGEEEIIUeeMyQ5ACCHSwvr14PcnO4qG53TCIYckOwohhBBpSBINIYTYl/XroVs3iMWSHUnDM5lg9WpJNoQQQtSaJBpCCLEvfn9pkqGqYGxCH5vxeOnjboojOUIIIQ5YE/rGFEKIA2Q0ll7hb0qi0WRHIIQQIk3JYnAhhBBCCCFEnZNEQwghhBBCCFHnJNEQQgghhBBC1DlJNIQQQgghhBB1ThINIYQQQgghRJ2TREMIIYQQQghR5yTREEIIIYQQQtQ5STSEEEIIIYQQdU4SDSGEEEIIIUSdk0RDCCGEEEIIUeck0RBCCCGEEELUOUk0hBBCCCGEEHVOEg0hhBBCCCFEnZNEQwghhBBCCFHnJNEQQgghhBBC1DlJNIQQQgghhBB1ThINIYQQQgghRJ2TREMIIYQQQghR5yTREEIIIYQQQtQ5STSEEEIIIYQQdU4SDSGEEEIIIUSdk0RDCCHqSZ/iYib4fA12vFXRKDnbt+NNJBrsmPvSp08fJkyYkOwwhBBCJIEx2QEIIURj9XxGhnzICiGEaLJkREMI0aTEYrEGO1amquJUG//HrK7rxOPxZIchhBAixTT+b0AhRJPWp08fxo0bx4QJEzjyyCPp378/OTk5/PzzzxXbeL1ecnJyWLVqFQCrVq0iJyeHlStXcsEFF9D+wgvpGQyyTtMq9pnm93POzp28GQpxYmEhhxcUMNTjwV9p2lL1qVMnFhbyVCDAKK+XjgUFnFBYyMvBYJV4v4lGOWfnTtoVFHBBUREfhMPkbN/Oz7VIkL6OxTi77D4uKSrit2pJwLvhMGfs3EleQQEnFhYyJxCocvuboRAXFBXRsaiIY/x+hk2eTGFhYcXt5c/Pp59+yvnnn0+7du34+uuvCQaDjBgxgkMPPZRjjz2WOXPm1DhmIYQQjY8kGkKIRm/x4sWYzWaWL1/OI488UuP9pk6dyv33388Hs2djBEZVOyHfqGl8EImwIDOTBZmZ/Cca5elqiUN1c4JBjjEa+Wd2NtfZbIz1+VhXlgj4EgkGejx0Nhr5KDub0Q4Hk/3+Wj/eSX4/9zudvJ+VRTNVZVBJCTFdB+DHWIybPR56WSx8lp3NnQ4HjwYCvB4KVewfB0Y7HHyckcF8m43N27YxcuTIXY4zZcoUxo0bxxdffMERRxzBpEmT+PLLL3nxxRd59dVX+fLLL1m7dm2t4xdCCNE4yPRhIUSjd8ghhzB+/HgANm/eXOP9xo4dyymnnAJr13Kr2cy14TARXceiKAAkgCfd7orpUVdYrayMRvd6n2ebzVxntwMw3GBgXjDIv6NROhiNLAmHUYDH3W4sisJhRiPbEgnu8npr9XjvdDg4zWIBYIbbzfGFhbwfiXCp1crcYJAeZjOjnE4A2huN/BGPMzsYpJ/NBsBVZf8lFiPPYGDybbdx4YgRBAIBHA5HxXFGjx7NaaedBkAgEGDRokU8/fTT9OjRo/TYM2Zw/PHH1yp2IYQQjYeMaAghGr2jjz56v/br3Llzxf+3KksuCitNjWprMFRZg9HSYKhy+27v0/i/6zuKotBCVSv2WadpHGE0ViQyAEcbDLWOu6vJVPH/mapKB4OBP8pGTf6Ix+lW6XaAbmYz6zUNrdKox6CSEk4oKaGj38/lo0YBsGXLlir7HXPMMRX/v3HjRmKxWJXEIjMzkw4dOtQ6fiGEEI2DJBpCiEbPXjaCAKCWJQZ62Uk17HmBuKnSCXn5qX/lNKL6kLBS7fbd3melJKJ8H333mwIQDAZJNGC52qCu07+kBKei8IzDwft2O/MnTgR2fZ4qP69CCCFEdZJoCCGalGbNmgGwffv2it9VXhieTB0MBn6Nx4lWSoL2J7JvKyUEnkSCvzSNjmUjKR2NRlZXSxhWR6O0NxgwKAp/xuMUJxKMczo5yWTiUFWlsLh4n8fMy8vDZDLx3Xff/e/YHg9//fXXfjwCIYQQjYGs0RBCNClWq5WuXbvy9NNPk5ubS2FhYa0WiNen3lYrUwMB7vZ6udXhYIum8XxZUrC3UY/qpgcCZKsqzVWVqX4/2arKhWVrNoba7VxYVMR0v59eVivfxGLMD4V42OUCoLXBgElRmB8Mcq3JxO/xONNffnmfx3Q4HPTv359JkyaRlZVF8+bNmTp1asUIkhBCiKZHvgGEEE3OE088QTwe5/zzz2fChAmMGTMm2SEB4FJVFmRk8FM8zjlFRUwNBLi9bGG2sRbTp+51OrnP5+P8oiJ2JBIsyMiomLJ1lMnE3IwMlkUinFlUxGOBAHc7HBULwZupKjPcblZEIpzu8TAzGuX+oUNrdNz77ruPk046iUGDBtGvXz9OPPHE/V4fI4QQIv0peuWJykIIIXa1di306AFmM1RbSF3f3gwGGenx8FNWFplWa4Mem1gMolFYuRKOOqphjy2EECLtydQpIYRIIW+EQuQZDBxkMPBLLMZDgQDnKQqmBlwQLoQQQtQFSTSEECKF7EgkeCwQoCCRoJWq0tNqZXAkgqZpjPF6eSsc3u1+faxWHnG7GzhaIYQQYs8k0RBCiBQyzOFgWKWmeADFmoamadztdnPLHkrKOmXRtRBCiBQjiYYQQqQ4g8FANBqleVklKSGEECIdyDeWEEKkOIPBgKZptSpxK4QQQiSbJBpCCJHijAYDuq43aIdwIYQQ4kBJoiGEECnOYDAAoGlakiMRQgghak4SDSGESHGSaAghhEhHkmgIIUSKUxQFVVUl0RBCCJFWJNEQQog0UL4gXAghhEgXkmgIIUQaMBoMaPF4ssMQQgghakwSDSGESAMyoiGEECLdSKIhhBBpwGA0oiUS0ktDCCFE2pBEQwgh0oBUnhJCCJFuJNEQQog0IImGEEKIdCOJhhBCpAFVVVEURRaECyGESBuSaAghRBpQAIOqoiUSyQ5FCCGEqBFJNIQQIk0YpMStEEKINCKJhhBCpAmD0ShrNIQQQqQNSTSEECJNlPfSkBK3Qggh0oEkGkIIkSYMBgMJXUfXJdUQQgiR+iTREEKINCElboUQQqQTSTSEECJNVCQasiBcCCFEGjAmOwAhhBA1oyoKqqLIiIYotX49+P3JjqLhOZ1wyCHJjkIIUQOSaAghRBopXxAumrj166FbN4jFkh1JwzOZYPVqSTaESAOSaAghRE2lwJQlk6qix2INc4KZAo9X7IHfX/oeUFUwNqGv8ni89HE3xZEcIdJQE/p0EkKI/eR0ll5FjcUgGk1qKBZdJ6ZpDReHyVT6+EVqMhpLX6OmJMl/g0KImpNEQwgh9uWQQ0qnaqTAVdQPli3j6aef5oMPPqhYHF6vZD68EEKI/SSJhhBC1ESKnGxn7NzJz7Nmkd+8OW3atEl2OEIIIcQeSXlbIYRII7m5uQBs3LgxyZEIIYQQeyeJhhBCpJE2bdqgKAqbNm1KdihCCCHEXkmiIYQQacRsNnPwwQdLoiGEECLlSaIhhBBpJjc3V6ZOCSGESHmyGFwIIdJMXl4ef/75Z8MfWDpRCyGEqAVJNIQQIs20bduWTz75pGEPKp2oJdkQQohakkRDCCHSTF5eHoWFhQQCARwOR8McVDpRJzsSIYRIO03o20IIIRqHvLw8ADZv3kynTp0a9uDSiVoIIUQNyWJwIYRIM23btgWQylNCCCFSmiQaQgiRZlq2bInFYpFEQwghREqTREMIIdKMoihS4lYIIUTKk0RDCCHSUF5enoxoCCGESGmSaAghRBpq27atJBpCCCFSmiQaQgiRhspHNHRdT3YoQgghxG5JoiGEEGkoLy+PUChEYWFhskMRQgghdkv6aAjRmK1f3zQbjTmdjb6Lc25uLlBa4rZFixZJjkYIIYTYlSQaQjRW69dDt26lXY2bGpMJVq9u1MlGeS+NjRs30rVr1yRHI4QQQuxKEg0hGiu/vzTJUNXSbs5NRTxe+rgb+UiO0+kkOztbFoQLIYRIWU3o7EOIJspoLL3C35REo8mOoEHk5eVJLw0hhBApSxaDCyFEmsrNzWXz5s3JDkMIIYTYLUk0hBAiTUl3cJGKVkWj5GzfjjeRSHYoQogkk0RDCCHSVF5eHlu3biXWFBf8i5TVzWRiTfPmuBQl2aEIIZJMEg0hhEhTubm56LrOli1bkh2KSEOxemr2aFIUWhoMKHtINDRdJ7GbY+9vPJJoC5G6ZDG4kF4LQqSp8l4aGzdupF27dskNRjSIRCLBnDlzeHnePLb4fLQwGLjWbud2h4PJPh/vRyLkJxK0VFUut1oZ5XBgKjvhn+b3834kwg12OzMCAf7WNLa0arXX4/UpLqaTwYBBUVgcCmFWFMY4nfS2Whnn8/FOOEwLVWWyy8VZFgtQOnXqiuJifmvRAreq8nooxP0+H09lZDDF7+eveJxVzZvTp7iY/lYr6zWNDyIRLrJYeDIjg3fDYR4LBFivabRSVW6w2RjqcFTEdKLfT/+FC1kfDvPBBx9w0UUX8eSTT9bbcy6E2H+SaDR10mtBkg2RtnJycjAYDFLitgl56KGHeOWVV3jwxhs58cEHKTAY+LMskXCqKjPcbloZDPwWj3OX14tTURhW6SR9g6bxbjjM8xkZGGp4zDfCYYY5HLyfnc2ySISxXi/vh8NcaLUywm5nXjDIbV4v3zRvjm0Poxgh4OlAgGkuF1mqSnO1dELFnGCQUQ4Hd5TF+GMsxs0eD3c6HPSyWlkdi3GPz0eWqtLPZqu4vzmLFzNq7FjuuOOO2j+JQogGI4lGUye9FpIdiRD7zWg00rp1a1kQ3kT4/X6ee+45HnroIa486iiYNIl2JhMnlpWvHlkpoWhrMDDUbmdZJFIl0YgBT2Vk0Eyt+czpI4zGivu+zWDg6UCAbFXl6rIT/1EOBy+FQvwSj9N1D6W0Y7rOVJeLI6rdfqrZXGW0YrjHQw+zmVFOJwDtjUb+iMeZHQxWSTROPe44hg4dWuPHIIRIjiZ0Zin2SnotCJGWcnNzZUSjifjjjz+IRqP06NEDPJ5dbl8eDvNcMMhGTSOg68RhlwXZrVW1VkkGlCYa5VRFwa3rdKx0vy3K7m/nXqpMmRSFzru5mHVMte+dP+Jxzi+bglWum9nMs6EQmq5XjMIcc9hhtXoMQojkkMXgQgiRxiTRaDqsVuseb/s2FmO4x8PZFgsLMjP5Z3Y2t9vtuyywtu9HJShjpX3C4TB6IlGx7gOoWPS9uwXe5WyVtjvQeADse3kuhBCpQxINIYRIY9IdvOlo3749VquVlStX7nLb6miUNgYDtzscHGMycYjRyN913MdCp3T6lqooGAw1XeFROx2NRlZXWzO4OhqlfdmCdCFEepGpU0IIkcby8vLweDx4vV7cbneywxH1yGKxMHz4cCZNmoRp8GC6JRLsjMf5bzxOe6ORLZrGsnCYY4xGPolGeT8SqdPjh0IhNE1DrackA2Co3c6FRUVM9/vpZbXyTSzG/FCIh12uejumEKL+SKIhhBBprG3btgBs2rSJLl26JDkaUd9GjRqF0WjkseeeY3sgQEuDgYF2O/1tNm6y27nX5yOq65xtsTDK4WBaHRW80IGA34/VakWpxyqFR5lMzM3I4LFAgCeDQVqqKnc7HFUWggsh0oei6/XUsUekh7VroUcPMJub1mLwWKx0MfjKlXDUUcmOpn7Ia9t4X9tKioqK6NKlC88++ywXX3xx/R2oFu+nmK5XmcPfUOrluKn6fmrgv+9AMIjf56NZ8+YY63FEY59S9fUQQuyWrNEQQog0lpWVhcPhSOqC8D7FxYzzepng83Hkjh30Lynht3icq4uLObSggKN37OA2j4eisjUDLweDHLdjxy6Lh68rKWGU11vx7w8jEc7buZN2BQWcXFjINL+feKV9crZv56VgkEElJXQoKOCpQABPIsFwj4cuO3ZwSEEB3QsLeT0Uqthnq6Zxs8dDp4ICOhcUcF1JCZs1rZ6fofSW0HUCgQBWmy25SYYQIu3I1CkhhEhjiqKkROWpxeEwg2w2lmdl4dF1+hYXM8Bm40GXi7CuM9nv52aPhzeysuhptTLO72dVLEYPsxmAkkSCz6JRXsnMBOCraJQRHg+TXC5OMpnYqGnc7fMBcGdZjwWAaYEA45xOJjqdGBWFR/1+/huP80pmJtmqygZNI1yWnMR0nf4lJXQ1mViSlYVRUXgyEGBAcTGfNmuWlFGYZNqiaZy+c+ceb/+iWTNaGwyEgkF0XcdZqd+FEELUhCQaQgiR5vLy8pKeaBxiMDC+bMHuk4EAXYxG7qmUEEx3u+laWMhfZQuXzzKbeTscrkg03gmHyVYUupdNA5oWCHCrw8GVZXPz84xGRicSTPb7qyQava3WKvP3/04k6GIyVfRnaFvpCvzycJiErjPN5aootfqk283hO3awKhrl9Gr9Gxq7VqrKx9nZe709oesEgkFsNlu9VZoSQjRekmgIIUSay83N5eOPP05qDEdXWifwSyzGqliMQwsKdtlug6bR3mjkcquVu71eprpcmBWFt8NhLrNaUcsSgF/icVbHYswIBNB1HUVR0ICIrhPSdWxl2x1TrQncIJuNwR4Pa2MxzjCbucBi4YSyZOaXeJwNmkbHHTuq7BPRdTZoGqfX5ROSBoyKQrvdNNGrzF/2/DtkNEMIsR8k0RBCiDSXl5fH5s2bSSQSqLXs+lxXKjdeC+g655rNjK808lCuZdlV8fMsFu4CPo5EONZk4qtYjAcrlTAN6Dp3OxycbzZTtHMnbrcbS1kyU3ncoXrDt7MsFlY3b84nkQj/F43St6SE6202JrhcBHSdo00mntlNGeDsJD1vqSyRSBAIBLDb7Rjk+RFC7AdJNIQQIs21bduWWCzG9u3bOfjgg5MdDkeZTLwbDtPGYKjSVboyi6JwkdXK2+EwGzSNDkYjR1UaFTnKaORPTWMw4FAUmplMmPZx9b1cM1XlSpuNK202TgoGmeT3M8Hl4iiTiWXhMM1UFZecOO9TIBgEwGG3JzkSIUS6kk9aIYRIc3l5eQAp0yH8epuNEl1nmMfDmliMDfE4n0cijPJ60SpVjbrcauXjaJRXQyEut1qr3McdDgdvhkI8EQjwp67zl66zLBzmkX30hXjM7+fDSIQN8Ti/x+N8HI3SsSxBudxqJVtVud7j4atolE2axqpolPE+H/lSeaoKLZEgGAzisNuTNkomhEh/8ukhhBBprnLTvlTQymBgeVYWGnBVcTFnFRUxwe/HrShVvnRONZnIUhTWaRq9qy3EPsNiYUFmJv+Kxbha17m0uJh5wSBt9rEg2aQoPOT3c1ZREb2Li1GB2RkZANgUhSVZWbRWVW70eDht507u9HqJ6DrOJlZxal8CgQAKYJe1GUKIAyBTp4Rogqb5/bwfifBxs2Z73Gakx4NH13mhrNxon+JijjQamVg2j/7EwkKG2O0MSdK0itdDIe73+fitZcukHD+VWK1WWrVqlbRE462srF1+d4jRyPNl7509URWF71u02OPtZ1gsdLXZiITDNG/efJfbt7ZqtcvvRjocjNzLyXFLg4EZZYmH2D1N0wiFQjgcjorF+UIIsT8k0RAijY0cORKPx8MLL7xQ5/c9sdLC3N15Pzt7l4W4Inlyc3NTZupUXdI0DVXKqjYofyCAoijYZW2GEOIASaIhhNgt9z7mZTer53nbMV1PSgO1WCyGqdKi5HSRCk376kMikZCKRw0ormmEQyGcLpeMZgghDph8eguRBt555x3OOussDjnkEI444giuvPJKJk2axOLFi/nwww/JyckhJyeHVatWATB58mROHTiQ9j4fJ5eU8KjfT6zSItxyC4NBuu7YQfuCAm72ePAmEhW3jfR4uL6kZI8xnVhYyLNlVWleD4XI2b59l59plRbuLgqFOG3nTtoVFPCPwkJeLNsXYLOmkbN9O8vDYS4vKqJdQQFvh8M1em4+CIfpXlhIu4IC+hcXs7XSot4NGzZw3XXXcfTRR3PooYdy4YUX8q9//avq4zjxRKZPn86IESM47LDDuPvuu2t03FSTCk376oOMaDSsgN+PqqrYKzVBFEKI/SUjGkKkuO3btzNs2DDGjx/PhRdeiN/v56uvvqJv375s2bIFn8/H9OnTAcgqmyvvdDqZMWYMrYYM4TdV5a5gEKeiMKzS3PUNmsaKSIQFmZn4dJ07vF7u8fl4Zj/mr/eyWjmzrCkawJexGLd5PHQrGxl4OxTiMb+fKS4XXUwmforFuMvnw64oFZ2fAab4/dzvdNLFZKImPZpDwIxgkJkZGZiAe3w+hgYCLC+rYBQMBjn77LMZO3YsFouFN954g4EDB7Jy5Upat25dcT9z5sxh1KhR3HHHHbV+7KkiNzeX7du3Ew6HsVar4JSudGREoyHF4nFC4TBut7uic7oQQhwISTSESHEFBQXE43Euuugi2rRpA0Dnzp2B0kXAkUiEltUWRI8cORLWrgVVpa3ZzFBgWSRSJdGI6DpPud0cVHa1eIrLxbUlJdzvdFY0Vaspq6JgLdtnQzzOPT4fY51OTiurJPR4IMD9LhcXlZ0A5xoM/FfTWBgKVUk0htjtFdvUREzXmeJycXxZQjPD7ea0nTv5XtM4DjjiiCM44ogjKrYfPXo077//Ph999BHXX399xe9PPfVUhg4dWqvHnGpyc3MB2Lx5Mx07dkxyNHVDTyTQdV1GNBpIIBDAYDBga2qjGevXwz7KJjdKTiccckiyoxCNnCQaQqS4I444gh49enDWWWdxxhlncPrpp3PJJZeQsZeRh+XLl/PcE0+w0e8nAMQVBVe1K5StDYaKJAOgq8lEAlinabVONMp5EwkGejycYzZzi8OBDoR0nQ2axh1eL3d5vRXbxgF3tZiOqWFDtnJG4NhK+xxqNOJWFP5IJDiO0hOnadOm8fHHH1ckbOFwmC1btlQ97jHH1PKRpp7KvTQaS6KhlU3lkxGN+hcr+9vIyMigSY1lrF8P3bpBLJbsSBqeyQSrV0uyIeqVJBpCpDiDwcDrr7/ON998wxdffMH8+fOZOnUq77777m63//bbbxk+fDh3DRjAGd98g9tsZmk8ztxKayLqg6brDPV4cCkKj7hcFBcX47DbCZQlAo9XGnkoV32xaV1XsZo4cSL/93//x4QJE2jXrh02m43BgwcTjUarHrcRVNdp1aoVJpOJzZs3JzuUOpMoW28jIxr1z+/3YzQaG820uxrz+0uTDFWFWl7oSGvxeOnjboojOaJBNaG/qvRw4oknMmTIEIYMGZLsUEQKURSFbt260a1bN0aNGkW3bt14//33MZlMJCot4AZYvXo1bdq04fZrroE5c8Bg4O9qJ9YAWzSN7ZpGq7KTuO9iMVSgw36e1N3v9/NrPM4H2dlYVZUQ4PP7ad6sGa1UlY2axuV1PCUjDvwQj3NcWQKzLh7Hq+t0LLsCvnr1aq688kouvPBCoHSE4++//67TGFKFqqq0bdu2UZW41RIJFJDO1PUsGosRiUTIbGqjGZUZjaVX+JuS3XwvCFHXJNEQTc6JhYUMsVgYkiInL/vqhfHdd9+xcuVKTj/9dJo3b853333Hzp076dixI5FIhM8//5x169aRlZWFy+Wiffv2bNmyhWWffcYxiQSfhMO8H4nscr8WRWGE18v9Tic+XWe8z0dPq3W/pk29HgrxUjDI/MxMFGCHphFzOIgUFWEPhbjb6WS8z4dbVTnTbCai6/wYj1OSSHDzAXQeNikK430+JrlcGIFxPh9dDQaOK3sMhxxyCO+99x7nnXceiqLwyCOP7JKYNSaNrfJUQtNQVbXpnvw2AJ3/jWZYmtpohhCi3kmiIZqMZPVlOFAul4v//Oc/PPvss/h8Ptq0acMDDzzAWWedxTHHHMOqVau44IILCAQCvPnmm5x33nncdNNN3PvUU0SDQc42mxnlcFQpNQvQzmDgIouFq0tKKNF1zjWbmbqPJn178mU0igYMqlYOd5jJxE1+P1c1b45NUZgVCDDR78cOdDYaD7iruA0Ybrcz3OMhP5HgJJOJJ5xOKJty88ADD3DHHXfQs2dPsrOzGT58OP5GPFUgNzeXr7/+Otlh1BktkZBpU/UsGo0SjUbJLLtIIIQQdUnR9d0U1xf1xu/3M2bMGD744ANcLhfDhg3jww8/5Mgjj2TixIm7TJ3asmUL48aNY+XKlaiqyplnnsnkyZNp0aIFf/31Fz169OD//u//OPTQQyuOMW/ePF544QW+/PJLAH777TcmTZrEV199hd1u5/TTT+fBBx8kOzu7tDJRjx5gNu922Dih68wJBnk5FGJLIkELVeVam43by65C/xqLMcHv55tYDJuicLHFwgNOJ46y0YKRHg8eXec4k4nngkGius5NdjsjHA4e9vtZFAphUxTGOJ30K5tWs1nTOKmwkNkZGTwXDLI2HucQg4GHXC5OKSuhquk6d/t8/DsapSCRoLWqcp3dzuBKJ67lxz7WZOLFYBCzotDGYODL8uFiXYfmzdlaUMDrr7/O/fffz8yZM3nwwQfZunUrZ599Nk899RQrVqzg8ccfx+v1csUVV/Dggw9iKDv5iUajTJ06lSVLluD1eunUqRPjxo2je/fuABX3O2fOHCZMmMDWrVsr+ja0atWKadOmMW3atCrP+Ztvvlmx/wHZx2vbELREgsLCQhx2O06ns2EOGouVTglYuRKOOqphjpkiZs+ezRNPPMF///vfui9PmoT3U3FxMYqqkrkfJZfrTKq+n+rg9dCBoqIiALKzs9Mj0ajr1yMFPieTIlXf16LRSY25I03IAw88wOrVq3nppZd47bXX+Oqrr1i7du1ut00kElx33XWUlJTw9ttv89prr7Fx48aKMpzt27fnmGOO4e23366y39tvv03v3r0B8Hq99O3bly5duvDBBx+waNEiduzYwc0331yjeB/y+5kZCDDS4eCL7Gxmud20KEsigrrOgJISMhSF97OzmZeRwb+iUcb5fFXu49/RKNs1jSVZWTzgcvF4IMDAkhIyVJX3srMZaLcz2ucjv1KjNYCJPh9D7Xb+mZ1NV5OJQSUlFJdNe0kAOarKvIwMvmjWjDvKEpfl1Zq8rYxGWReP81pWFgsyM3k+I4ODVZW7bTbWOBysWby4YttQKMTzzz/PnDlzWLRoEatWreKGG27g008/5eWXX2bmzJksXLiwyiLse++9l2+++YY5c+bwySefcMkllzBgwADWr19f5X5nz57NzJkzWbJkCVu2bGHixIkADB06lEsvvZQzzzyTNWvWsGbNGrp161aj1yYdGFQVu91OIBisqCAk6k9ubi6BQIDi4uJkh1IntERC1mfUo2gkQiwWw+l0pkeSIYRIOzJ1qgH5/X7eeOMNnnnmGXr06AHA9OnTOe6443a7/cqVK/ntt9/46quvyMnJAeCpp57ijDPOYM2aNRx77LFcfvnlvPDCC4wePRqAv/76ix9//JGnn34agPnz59OlSxfuueeeivudPn06Xbt25a+//qL93uJNJHguFOIhl6ui10E74MSy25eEw4R1nacyMiqqBU1xuRhUUsI4p5MWZVf9M1WVSS4XqqLQwWjkmUCAkK4zomxU5DaDgacDAb6OxehVaZrEDXY7F5fNGZ7qcvFZJMKroRDDHA5MisJdla6Q59psfBuLsSIc5tJK84ztisI0t7vKlCmDouBUFFqqKjRrVvH7WCzG1KlTadeuHQCXXHIJb775Jj/++CMOh4PDDjuMU089lX//+99ceumlbNmypaIaVKtWrQC45ZZb+Oyzz3jttdcqnvNYLMYjjzxScb/XX399RYM9h8Oxx14YjYXD4SAUCuH3+8lwu2u839XFxXy1h5KTIxyOiveP+J/KJW6zs7OTHM2BS2haxeihqFvlazPMZjPmSs02hRCiLkmi0YA2bdpELBarkli43W46dOiw2+3/+OMPcnJyKpIMgMMOOwy3280ff/zBscceS69evZg4cSLffvstXbt25e233+aoo46qmEr1yy+/sGrVqipTq8pt2LCB9i1a7DHePzSNqK7TYw9fQn/E4xxpMlUpSdqtUi+G8kTjcKOxShnTFqpKJ6ORhK4Ti8WwmM1kqSqF1a54d600jG1UFI4xmfij0qjHC8Egr5VN6QrrOjHgyGrlCTsZjTVel2Gz2SqSAYAWLVrQtm1bHJVOaFu0aEFhYSEAv/76K5qmceqpp1a5n2g0WuUkr/r9tmrVquI+mgJVUXA6HHh9Pux2O6YalpCc5nYT3sPMzky5yr1blZv27ekCRrpI6DoJXZcRjXoSiUSIxePpM2VKCJGWJNFIcy1btuTUU09lyZIlFYnGoEGDKm4PBAKce+65jB8/frf7sm7dHu+7ruqPVH+TKYqCUVEIBAIEAwFcbjcKpdOhampZOMxEv5/7nU5OMJlwKgqzgkG+q3YFvDZ9GUzV5ucqioKx2kmxoigVVYuCwSAGg4EPP/xwl6uulfsy7O5+m9rSKJvdTjAYxO/zkZWVVaN9DpIr2bXmdrvJyMhoFCVuy//OZESj7pWPZlgsFsxNaV2CEKLBSaLRgHJzczGZTKxZs4bWrVsDpWso/vrrL04++eRdtu/YsSNbt25l69atFaMa//3vf/F6vRx22GEV211++eVMnjyZ3r17s2nTJi677LKK24466ijeffdd2rRps8tJ8760NxqxKgoro1EG7Kb/QUejkddDIYK6XnFCv7oWvRicTie6ruP1eklQ+uVX2XexGCeXjabEy8qh3lAWx9fRKN1MJq6rdEK/odoajz0xUbqY/EB16dIFTdMoLCzkpJNO2u/72V0vjMZGAZwuFyUlJUSiUSwyVaPe5OXlNYpEQytv1icjGnsXj9d6l0g0CrEYTrs9/Tpi78fjFUIkj3yCNyCn00nfvn2ZNGkS//73v/n999+58847S+vE7+bK+z/+8Q86derE8OHDWbt2Ld9//z0jRozglFNO4ZhjjqnY7qKLLqqoZtW9e/eK9QJQuh6gpKSEYcOGsWbNGjZs2MDnn3/OqFGjKr7I98SiKAy325nk8/FGKMSGeJxvYzFeDYUAuNxqxaoo3O7x8Fs8zr+jUcb7fFxhtVZMm9obhdLSrS6nEy2RIBwKVUk2XgiFeD8c5s94nHt9PjyJBFeVJRrtjUZ+iMX4PBLhr3icR/1+fqjhF2Zbg4H/xONsSyQo8nhqtM/utG/fnssvv5wRI0bw3nvvsWnTJr7//ntmzpzJxx9/XOP7adu2Lb/88gvr1q2jqKiIWLp98ddQ+dVTn8+3S1Ip6k5ubm6j6KVRMaIhicbuOZ2lVZISidLqQTX80SMR4oEADqMRUy33TYmfRKL0cTdUFTshxAGREY0G9sADDzBmzBgGDhxYUd5269atWCyWXbZVFIUXX3yRcePG0bt37yrlbStzOp2cd955LF++vGKRcblWrVqxfPlyJk+ezFVXXUU0GqVNmzaceeaZNbpSOMrhwKgoPBYIsD2RoKWqMrDsZN+mKCzKzGSC38+FRUVVytvWlELpYmFDIEAsFqOkpIRE2f7jnE6eDgb5qay87YuZmWSXxXytzcZP8ThDyxKF3lYrg+x2Pt1NY7rq7nY6Ge3xcHIgQPTyy9laUFDjeKubPn06M2bM4MEHH2Tbtm1kZ2dz/PHHc84559T4Pq6++updemHUSXnbFFM+qlFUVEQ4FMJWx13CRanc3Fx+/PHHZIdxwLTyZn1p2PumQRxyCKxeDbXsC/PuO+8wffp0nn32Wdq331s5kBTmdJY+fiFEypM+GkkWDAY5/vjjuf/+++nfv3/DB5BCNcQj0SiekhK2qSoXaBr/zM7myPqKqSnUEE+h17ayEo+HWDRKs+bNqxQJqDNN4bXdiwULFjBu3DjWr19f6+mSe9XA7yevz0c0GqV5pcpwSdGI3k+RSITu3btz4oknMnv27GSHkxpS9HOy3jWi97VIbTIm3cB++uknli5dyoYNG1i7di3Dhw8H4Pzzz09yZMlnMZvJys4mkUiQSCSIN/J1C02V0+kkoesEg8Fkh9Io5eXloWkaW7duTXYoB0RK29a9hQsXsn37du6+++5khyKEaCJk6lQSzJ49m3Xr1mEymTj66KNZsmRJUmreb9myhdMvvhh8PlCU0p9KvmjWjNYN/EVvMhrJzMqCwkI8Hg/RrCypitLIGA0G7DYbgUAAm80mc/DrWHkvjU2bNlWUu01HWiJRtyMyTVwwGOSpp56ib9++6TtlSgiRduRTvIF16dKFDz/8MNlhAKXrNz6eNw/69i0dMq72pd4qSSeA7Uwmth50ECUlJRQXF5ORkYF1N2tYRPpyOJ2EwmECfj/uWjTxE/vWunVrFEVJ+wXhCU3DIH/3dWb+/Pl4PB7uuOOOZIcihGhCJNFowoxGI+1atwZVBYNhl0QjmVRFISsrC4/HQ0lJCS6XC0elUrYivamKgsPhwFfWxE+uXNcdk8lETk5OWpe41Skd0ZDStnXD6/XyzDPPcPXVV9O2bdtkhyOEaELkU1ykLAXIyMioOCGVsqiNi91ux2Aw4Ktl1Ryxb+le4laa9dWtZ599lnA4zO23357sUIQQTYwkGiKlKYDL6cTtchEIBvF4PJJsNBLlr20kEiEajSY7nEYl3RMNadZXd4qLi5k7dy7XX399lR5LQgjREGS+gkgLdrsd1WDA4/FQXFREZlZW/ZRGFQ3KYrViCgbx+f1kZ2cjr2jdyMvLq1XTyFppgM7MejSKUdcxJBLJ71yd5p2on3nmGRKJREWFQyGEaEiSaIi0YbVYULOyKCkpoaioiKzMTJlakebKu8MXFRURDoexWa3JDqlRyM3NZefOnfj9fpx11UG5vBN1ef39eqREo1gAJdlJRrk07URdUFDA/Pnzufnmm2mW7H4kQogmSRINkVbMJhPZ2dkUFxezs6iIrKwsTLKQOK2ZTSasFgt+vx+rxSKdoOtAeYnbzZs307lz57q50/3sRL0/Xpg7l3/961+8/PLL9X6sGknTTtRPPfUUZrOZW265JdmhCCGaKDlDE2nHaDCQnZ1dMbKRmZmJxWxOdljiADhdLnYWFhIMBnE4HMkOJ+2VVxbatGlT3SUa0GAn2z8kEoQOPVQ6Fh+ALVu2sHDhQu644w4pIS2ESBpZaSfSkkFVycrKwmw2U1JcTCgcTnZI4gAYDQZsdjuBQKCi4pDYfy1atMBqtaZtidtt27Zx8MEHJzuMtDZ9+nRcLheDBw9OdihNwuuhEJ0KCg74fvoUFzPB56uDiIRIDZJoiLSlKgqZmZnYbDY8Hg/+QEAqUqUxZ9lIhj8QSHIk6U9RFPLy8tK28lR+fj4HHXRQssNIWxs2bOD111/ntttukxHCBtLLamVl8+bJDkOIlCOJhkhrCuByu3E6nfj9frxeryQbaUpVVRwOB6FgkHhZeVOx/9K1xK2u6+Tn55OTk5PsUNLWtGnTaN68OYMGDUp2KE2GVVFoLuWYhdiF/FWItKdQejU8IyODcChESUkJCV3SjXRUXsbYL038Dlhubm5aTp0qLi4mGo3KiMZ++v3333n77bcZOXIkVqnidkD+GYnQqaAArez75OdYjJzt25lSaWrTnV4vt3o8u0ydmub3c87OnbwZCnFiYSGHFxQw1OPBX2lqaFDXGeHxcGhBAcfu2MGc3YzmehIJRng8dCoooH1BAVcXF7O+rOSyrut02bGDdypNHT5n506O3bGj4t9fR6PkFRQQku9EkSSSaIhGw2a1kpmVRTQapbi4GE3m+qcdRVFwOp2Ew2GiqVLaNE2Vj2joaXaCsW3bNgBJNPbT448/TuvWrRkwYECyQ0l7J5lM+HWdn8pO7L+MxchWVb6s9Nn0ZTRK9z0UI9moaXwQibAgM5MFmZn8Jxrl6WCw4vZJPh9fxmK8mJnJq5mZfBmLsbba597tXi8/xOO8lJnJO1lZ6MDVJSXEdB1FUTjZZOLLsnLTnkSCPzSNsK7zZ6WYjzUasUk1P5EkkmiIRsViNpOdnU1C0ygqKpIpOGnIarViMhrx+XwyDe4A5OXlEYlEKKiDBaoNKT8/H0CmTu2HtWvX8u6773LnnXdiMpmSHU7ac6sqRxqNrCo7kV8VjXKT3c7aeJxAIsE2TWODpnHKHp7rBPCk200no5GTzGausFpZWXZfgUSCReEwE5xOepjNdDaZmOF2U7k95Pp4nI8iEaa5XJxkNnOEycQzGRlsSyT4IBIBoLvZzKqy5OQ/sRhHGY2cYjZXifkUqcookkgSDdHomIxGsps1Q1EUioqK5Mp4mlEoLXcbi8WIlH2ZitrLzc0FSntppJP8/HxUVaW5LKyttUcffZT27dvTp0+fZIfSaJxiNvNlLIau63wVi3GRxUJHg4GvYzG+jMVopaocsodeTm0NBpyV1m20NBgoLBtp36hpRBMJjq50e6aq0qFSE9o/NA0jcFylRCarbJs/ykYsTjGZ+G88zs5Egi+jUU4xmSqSj5iu800sRndJOkUSSR8N0SgZVLWi10ZxcTEZGRlYLZZkh5Uc8fi+t0kxFkXBZjIR9HqxZGZSq0H/NHy89aG8l8bGjRs54YQTkhxNzW3bto2WLVtilEactfLNN9/wySefMHv2bHnu6lB3s5nXPB5+iccxAYcajaUn8tEoHl3f62hB9VdBoXSUAyAYDqPreunFlAMYcehkNJKpqnwZjfJlNMpYp5OWqsozgQA/xGLEgRNkREMkkXwaiUZLVRSysrLweDyUlJTgcrlw2O3JDqvhOJ1gMkEsBmXD6OnEaTIRjEaJ+f2Ya/tFaTKVPv4mzOFw0Lx587RbEL5161bpobEfHnnkETp37kzPnj2THUqjUr5OY14wyMlln0PdTSaeDgYpSSQYuh/fKT6/n+xAALOi8JvJxGFlv/ckEvylaZxS9u+OBgNx4PtYrCJZKE4kWKdpHFaWTCqKwkkmEx9GIvyuaZxoMmFTFCK6zsJQiKONRuyyPkMkkSQaolFTgIyMDAwGAz6fj4Sm4XS5aneFPF0dcgisXg1pWsHJADz7xBN8/vnnvPLKK7hcrprv7HQ2WBfrVJabm5t2U6e2bdsmC8FraeXKlfz73//mhRdeQJUSq3UqQ1XpbDTydjjMlLLPoJPNZm72eontY0SjOh3QNI1AIEArt5sB8TiT/X6yVZXmqspUv7/KfPZDjEbOt1i4y+fjUZcLp6Iwxe/nIFXl/Eoj9N3NZh70+TjGZMJR9vqfbDbzdjjMLdJHRSSZJBqi0VMAl9OJQVXx+nxoiQQZdnvTSTbS2DWPPMKs7t2Z/vHHTJgwIdnhpJ10LHG7bds2unfvnuww0oau6zzyyCMce+yxnHfeeckOp1E6xWzm53i8orpUpqrSsWy9RYcaTlPTgVAoRCKRICMrC5vVyn2JBAGfj0ElJTgVhZvtdnzVqsQ96XZzn8/HwJISosDJJhOvZGZiqjRKcYrJhFb233LdzWY+jERkfYZIOkk0RJNR3qPB4/HgjcdxWyxNI9lIYy1btmTYsGHMmDGD6667rmKBs6iZvLw8vv7662SHUSsydap2PvnkE7799lteffVVFJkiUy8mulxMrDai+nGzZlX+3c9mo5/NVvHvO51O7iybvpnQdTwlJVyladycnV2xXtChqszMyGBmpfsZVm0EIkNVeSojY6/xHWkysbVVqyq/G2K3M6QpTRUWKUvGWEWTYrVYyMrKIq5pBIPBipr9InUNHTqU7OxsHn744WSHknZyc3PJz88nmiZrdILBIF6vVxKNGkokEjz66KOcdNJJnHbaackOR+xGQtdLm1DGYmRlZjbdoiSiyZJEQzQ5ZpOJjIwMdODWW2/lp59+SnZIYi/sdjujR49m2bJlfP/998kOJ63k5uai6zp///13skOpEWnWVzvvvfceP/30E2PGjJHRjBSkJRIUFRWhaRrZ2dm1L2ohRCMgiYZokoyqit1mo3nz5vTu3Zsvvvgi2SGJvbjyyivp1KkTEydOTLtO18lUPtVs06ZNSY6kZsqb9cmIxr5pmsZjjz3G6aefzsknn5zscEQ18bKmsbquk52djUlKDosmShINUSoeLy2D2lR+4nFUVeWJJ57g5JNP5tprr+X1119P9qsg9sBgMHDffffx1Vdf8dFHHyU7nLRx8MEHYzQa02ZBeHmiISMa+7Z06VL++OMPRo8enexQRDWxeJyioiIURSE7OxtjpSZ8QjQ1kmI3dWnea+GAmEzYW7bkhRde4J577mHUqFFs3bqVkSNHyjSEFHTGGWfwj3/8g0mTJnHWWWdhkmoq+2Q0GmnTpk3alLjdtm0bGRkZ2CotqhW7isViPP7445x//vkcd9xxyQ5HVBKNxSgpLsZgNJKVmSnlhkWTJ4lGU5fmvRYOSFmvBSPw6KOP0rp1ax599FG2bNnC1KlTpbtuilEUhQkTJnDeeefxyiuvcN111yU7pLTQtm3btBrRyMnJSXYYKW/x4sVs3LiR+fPnJzsUUUkkEqHE48FkMpGZmYkqF6yEkERDkPa9FuqCoiiMHDmSgw8+mLvvvpvt27czZ84cHNLsKKUceeSR9O3bl2nTptGnT5+qTfzWr2/SCfOe5OXlsWbNmoaL5wDk5+fLtKl9iEQiPPHEE1x66aV07tw52eGIMqFwGK/Hg8ViISMzU0qnC1FGEg0hKunXrx8HHXQQgwcPpk+fPixcuJAWLVokOyxRyZgxY1i2bBnPPPMMY8eOLf3l+vXQrVvpFMCmxmQqHZXcQ7KRm5vL8uXLGzio/bNt2zY5ed6Hl19+me3bt3P33XcnOxRRJhAM4vP5sNlsuN1uSTKEqEQmDwpRzemnn87SpUvZvn07PXv2ZN26dckOSVRy8MEHc/PNNzN37ly2bt1a+ku/vzTJUFUwm5vOj6qWPu69jOTk5ubi9XrxeDwN9ArtP5k6tXfBYJCnnnqKvn370qFDh2SH0+TpgD8QwOfz4XA4JMkQYjck0RBiN4488khWrFiBxWKhZ8+efPPNN8kOSVQyfPhwnE4njz76aNUbjMbSK/xN5acG64jy8vKA1C9xG4/H2bFjh0yd2osXXniBkpIS7rjjjmSH0uTpgM/nw+/343K5cDmdkmQIsRuSaAixB23atGH58uV06tSJvn378t577yU7JFHG5XJx55138sYbb/Dzzz8nO5yUVt5LI9UXhO/YsYNEIiE9NPbA6/Xy9NNPM2DAANq2bZvscJo0HfB4PASDQdxuNw67PdkhCZGyJNEQYi8yMjJ49dVXOf/88xkyZIhUeUkhV199Ne3bt+fBBx+UJn57kZmZicvlSvkRjfJpcJJo7N6zzz5LOBzm9ttvT3YoTVpC1ykpKSESiZCZmYldSjELsVeyGFyIfbBYLMyaNYucnBzGjx/Pli1bGDdunNRHTzKTycR9993Hddddx9dff81JyQ4oRSmKQtu2bVM+0di2bRsgzfp2p7i4mLlz53LdddfJ85NECV2nuLiYeDxOZmYmFrM52SEJkfIk0RCiBlRVZcKECeTk5HD//fezZcsWnnrqKczyRZNU5557LieffDJz587lRF2XOdJ7kJeXl/JTp/Lz87FYLGRmZiY7lJQza9YsEokEw4cPT3YoTZaWSFBcXEwikSA7K0sahgpRQ5JoCFELgwcP5uCDD2b48OH079+f+fPnk5GRkeywmqzyJn53nHsusXgcs8WS7JBSUm5uLh999FGyw9irbdu2cfDBB6NIk7MqCgoKeP7557npppto3rx5ssNpkuKaRnFxMeg62dnZGA2GZIckpG9S2pBEQ4hauvjii2nZsiWDBg2iV69evPLKK7Ru3TrZYTVZxx57LGeddRbRN9/EaLfLwrPdyM3N5e+//0bTNAwpepKUn58v6zN2Y+bMmZjNZm655ZZkh9IkxeJxiouLURWFrOzslP37aVKkb1JaJRvynSzEfujWrRsrVqwgFApxySWXSOWjJBs8eDC6rhMOh5MdSkrKy8sjFotVrINIRdIVfFdbt25lwYIFDB06VEZOkyAai1FcVITBYJAkI5VI36RkvwK1IiMaQuynDh06sGLFCgYOHEjv3r157rnnOO2005IdVpN00EEHETGb8YdCWJxODLJQv4ryErebNm1K2dG3bdu2cfzxxyc7jP9JgakZr06bxvEmEzedcgqsXdswB03DqRn1IRKNUlJSgslkIjMzE1Wm9KWe8r5JTUk0muwIak0SDSEOQMuWLXnrrbe4+eabueaaa5g2bRp9+/ZNdlhNktlshngcv99Phtud7HBSSnnfhU2bNnHKKackOZpd6brO1q1bU2fqVApMzUgkEtwUCHCbxYL5vPMa7sBpODWjroXCYbweD+ay4gSSYgix/yTREOIAORwOXnzxRcaOHcvtt9/O1q1bGTFihCxqbWCKomCz2fAGgzjsdow16JrdVFgsFlq1apWyJW5LSkqIRqOpM3Wq8tSMJL2PAoEAMVXF6XBAQ32WxONpOTWjLgWDQbw+HzabDbfbLUmGEAdIvomFqANGo5HHHnuM1q1b88gjj7BlyxYeeughOdltYDarlUAkgs/vJ0vKpFaRyiVuy9eOpMyIRrkkTc2Ix+MEIpHSE92GLqGdhlMz6oJOaXLn9/tx2O04XS5JMoSoA3IWJEQdURSFUaNGkZOTw9133822bduYM2cOdrs92aE1GQrgcrlKO/dGo9JQq5K8vDz++uuvZIexW9IVvCp/IIDBYMAmXacbREWSEYngdDpxOBySZAhRR2TFpBB1rF+/fixYsIAvv/ySPn36sGPHjmSH1KRYLBZMJhN+nw892cGkkNzcXDZv3pzsMHZr27ZtqKpKixYtkh1K0sXiccLhME6nU052G0AsFiMcChEOh3G73TglyRCiTkmiIUQ9OOOMM1iyZAn5+flceumlKXsluTEqH9UoP2ETpXJzcykoKCAUCiU7lF3k5+fTokULmWoI+P1+jAYDVqs12aE0eqFQiAkTJhCPx3G5XNhlBEmIOieJhhD1pEuXLrzzzjuYzWZ69uzJt99+m+yQmgyzyYTVasXv96PrMq4B/ytxm4qjGtKsr1Q0FiMSieCQ0Yx65/V66d+/P2t++AGbzSbTLIWoJ5JoCFGP2rRpw7JlyzjssMO44oor+OCDD5IdUpPhdDpJaBqBYDDZoaSEvLw8gJRcEL5t2zZJNCgbzTAaZTSjnhUUFNC7d29+//13Hn/88dKRtPKKW03lJx5P9ssgmggZpxainmVmZvLaa68xYsQIbrzxRiZPnsz111+f7LAaPaPBgM1uJxAIYLfZUJt4E7+WLVtiNptTssRtfn5+Svb3aEiRaJRoNCp9G+rZpk2buOqqqwiFQixdupTDzebSymKxWNOruGUylTZoFKIeSaIhRAOwWCzMnj2bnJwcxo0bx5YtW7j33nub/MlvfXM6HIRDIfyBAG6XK9nhJJWqqrRt2zZlE42mPKKhUzqaYTKZsFgsyQ6n0fr1118ZMGAAdrud5cuXVzSyZPXqptk7RLrAiwYgiYYQDURVVe6//35ycnJ44IEH2Lp1K08++WRpR2tRL1RVxeF04vf5sNvtGA2GZIeUVKnYSyMUCuHxeJp0ohGNRonFYmRlZcloRj359ttvufrqq2nbti2LFi2qWuFMTraFqDdyOVWIBjZkyBDmzp3Le++9x4ABA/B6vckOqVGz2+2oBgN+ny/ZoSRdKpa4LW/WlzJdwRuYDvh9Pswmk1x0qCeff/45V155JUcccQRvvfWWlFEWogFJoiFEElxyySUsXryYn3/+mV69elU0LBN1T6F0YXg4EiEaiyU7nKTKzc1l48aNKVWJKz8/H4CcnJwkR5IckUiEWDwunajryfLlyxk0aBA9evRg0aJFuN3uZIckRJMiiYYQSXLiiSeyYsUKAoEAF198Mb/++muyQ2p0+hQXM8Hnw2q1YjKZ8DXxJn55eXkEg0F27tyZ7FAqlCcaTXFEo3xthtlsxmwyJTucRuell17illtuoVevXjz33HNSzUuIJJBEQ4gkOvTQQ3nnnXdo2bIlvXr1YuXKlckOqVFSAJfTSSwWI9KEm/iV99JIpQXh27Ztw+12Y2uCzdLC4XBpszip/FOndF1n+vTp3HPPPQwePJgnn3wSkyRyQiSFJBpCJFnLli1566236NatGwMGDODNN99MdkiNktlsxmKx4PP703JUQ9d14gc45am8yk4qJRpbt25tktOmykczLBaLnATXoUQiwf33389jjz3G6NGjeeCBB6S6nxBJJH99QqQAp9PJiy++yBVXXMGIESOYOXNmSs2jT2earjPO6+XwggJ6RKM8FY8TCAQAyNm+nQ+qjXB0Kijg9VAIgM2aRs727SwPh7msqIhDCgq4sKiIv+Jx1sRiXFBUxKEFBVxdXMzORKLiPkZ6PFxfUsI0v58uO3ZwWEEBY7xeYpVe04SuMzMQ4KTCQg4pKODsnTt5p1Isq6JRcrZv59NIhPN37qTdjh18fYBrTNxuN5mZmSmVaGzbtq1JTpsKhUJomoZTRjPqTCwWY+TIkTz//PM8/PDDjBw5EkWRlS9CJJOUtxVNz/r1KVkz3QRMu+46jlFVXpo4kfiaNdx2662lXWvrQhOtmb44HKa/zcZ72dn8EItxp8fDwT4fQ+32Gt/H44EAE51OWhsM3OH1MszrxakoTHI6sSkKN3k8POb3M7XSQtOV0SgWReGtrCw2axqjvF6yVJWxZSeWM4NB3gqHecTl4hCDgf/EYtzq9dJMVTmlUvWhKX4/9zud5BoMZNbBldlUK3G7bds2OnXqlOwwGpQOBAKB0rVDdfX33cSFw2FuvvlmPvvsM2bNmkWvXr2SHZIQAkk0RFOzfj1061baBTYFKcAgYEAsRviFF4guXIjBaq2bq3ImU2ljqiaWbOQYDDzodKIoCh2MRn6OxVgYCHBtLZLNW+x2zihrpDbYbucWj4fFWVl0K0sIBthsFaMg5UyKwnS3G5uicLjRyN0OB5P8fkY7HMSBpwIBFmdl0bVs2kye0cjXsRgLQ6EqicZop5PT6rCJW15eXkqNaGzdupUzzzwz2WE0qFAwiKZpZGVlJTuURsHr9TJo0CB+/PFHXnrppSb3fhIilUmiIZoWv780yVBVSOEriSazGd1sxufzEQ2HcblcGA7kanY8Xvq4U3Akp74dbzJVSdROtFiYFwziCwZrfB+dK71Xmpe9DuW/S+g62UBhpalTAEcajdgqHfcEk4mArrM1kSCg64R0nX7FxVX2iQFdqr0vj6nj92nbtm35/vvv6/Q+91c8HmfHjh1Nqlmfruv4AwFsNluTbyBZF3bs2MGAAQP4+++/Wbx4MV27dk12SEKISlL3TEuI+mQ0ll7hT2Fmkwm32UxxcTFFPh9ZWVkHdmISjdZdcGlOURRURYFEYpeF4bsb66r8TlGq/c7r9RKMRGq1wDxQtlZjYWYmB1dLIM3VRq/sdTzHPC8vjy1bthCLxZK+CHnHjh0kEokmtUYjGAqhJxI4HI5kh5L2Nm/eTL9+/QiFQixZsqTJTcETIh3IYnAhUpjJaKRZdjaKolBUVNTkG87tj++rPWffxmIcYjSS4XaTqetsqXT7+nicUC0X4dtsNrTdJCw/x+OEK93Xt7EYDkUhR1U5zGDArChs0TTaGY1VfnLq+Sp3bm4uiUQiJZpElvfQaCojGgldJxAIYLPbZTTjAP3+++9ceumlACxbtkySDJE05f2axO5JoiFEijMYDGRnZ2M0GCguLiYciSQ7pLSyRdN4wOdjXTzO0nCY+cEgg202rFYrJ6kqLwQCrI3F+CEWY4zPh6mWIwhmsxlVVdGrTZ2K6Tp3eL38Nx7nk0iExwMBrrfbURUFp6pyi93O/X4/i0MhNsTjrI3FmB8MsrjaWo+6lkq9NLZt2wY0nUQjGAyi67qMZhyg7777jt69e9OsWTOWLl1a8Z4WQqQemTolRBpQFYWs7Gw8Hg8lJSW4XS7staia1JT1tVoJ6zoXFRWhUrqY+xqbDQV4MCOD24uLuayoiIMMBia5XPzo8dTq/hXAYjajh0JoiUTFWpoeZjPtDQZ6FxcT1XUus1q5q9IJ5miHg2aqysxAgE2JBG5F4SijkRH1fBLaunVrVFVl48aN/OMf/6jXY+1Lfn4+ZrOZzMzMpMbREBK6TjAQwG6zHdh6qybu//7v/7jhhhvo0qULCxYswF2p0psQByKm67W+0NRQYrpOak/23jNJNIRIEwqQkZGBqqp4fT60RAKn00lqfiymhrcqVfWZupsTklyrlflWK5qm0ax5cxTgt5YtK25vazCwtVWrKvt0N5t3+d01GRmcH40SDAardHm+y+nkrj30SVAUhcF2O4P3kDDu7jh1wWQykZOTkxIjGvn5+eTk5KRsr4NVsRhXFBXxW4sWuPeQHEzz+3k/EuHjZs32el+BQAAdsJclkps1jZMKC/lndjZHpvh6sVSxYsUKbr31Vk4//XTmzp3bJLvJi6oSus6cYJCXQyG2JBK0UFWutdm43eFgss/H+5EI+YkELVWVy61WRjkcFclE+d/uDXY7MwIB/tY0tuzjMzeo64z1enkvEsGpKLstk+5JJLjP5+OjSIQocIrJxGSXi0MqFfZ4NxzmsUCA9ZpGK1XlBpuNoZUuMp1YWEh/q5X1msYHkQgXWSw8maYXF+WyihBpRAHcLhcul4tAIIDH40nLLtepxOlyEdc0QrWoQlWdqijYbDZCoVBaNFpMlRK3+fn5KbUQvE+fPkyYMKFW+wy123ljH2VqtUSCYDCI3W6X0Yz9tHDhQoYOHUrPnj15/vnnJckQADzk9zMzEGCkw8EX2dnMcrtpUfY35lRVZrjdfNGsGZNcLl4JhXi22uf8Bk3j3XCY5zMy+Dg7e5/Hm+Tz8WUsxouZmbyamcmXsRhrq60DvN3r5Yd4nJcyM3knKwsduLqkpKJh64+xGDd7PPSyWPgsO5s7HQ4eDQR2KZE+JxjkCKORj7KzGZnG0y1lREOINOSw2zEYDHg8HoqLi8nMzCytoiRqzWQ0YrPZ8AcCWG22/X4e7XY7wWCQcLVO46mobdu2/Pbbb8kOo1F0BXeoKns7BYjpOqFAAAXSbm2GrutomlZ3TUP3M4aZM2cydepUbrjhBiZOnIgqyZoA/LrOc6EQD7lcXFmWeLYDTiy7vfLJeVuDgaF2O8siEYZV+n0MeCojg2Y1eE8FEgkWhcM87XbTo6zX0Qy3m+MLCyu2WR+P81EkwvKsLE4o2+aZjAy6FhbyQSRCT6uVucEgPcxmRpWNdrc3GvkjHmd2MEi/Sgn0qWZzlVGOVO3/tS/pm2ikaHfnetdEuzuLXVktFtSsLEpKSigqKiIrK0uulu4np9NJOBwmEAhUmfpUG0aDAYvFQiAYZHqzZik9pS0vL48PP/ww2WGQn5/Pcccdl+wwABg5ciRffvklX375Jc/NmQM+H9PL3gs/xuNM9vv5bzzOkUYjT7rddCg7+a4+dWqkx4NH1znWZOLFYBATsELX+dNq5eqiIv7QNDoZjdxey2kQH0UiPODzsTWR4ASTiSutVkZ6vVWmdX0djfKQ388P8TjZisKFViv3Op0VJZLfDIV4LhRiXTyOXVE41WxmostV0RtmVSzGFT4fL3/1FY/cdRe//fYbr776Kt27d6+Lp7jWEokEkyZNYu7cudx1112MGjUqZafZiYb3h6YR1fWKk/7qlofDPBcMslHTCOg6ccBV7f3TWlVrlGRA6ehHTNc5vtJUx0xVpUOlCnJ/aBpG4LhK22SVbfNHPF66TTzO+dWasHYzm3k2FELTdQxlMR7TSKZUpmeikeLdnetVE+3uLHbPbDKRnZ1d2mtj504ys7IwpXAjwlRlUFUcdjuBA5zeYrfbKS4uJhaLYU7hL4m8vDyKi4vx+Xy4XK6kxKDrekpNnZo4cSLr1q2jU6dO3H3xxdCzJ/8tex9M9fu53+mkmaoyxutllNfL8r1Ms1gZjeJSFF7LysLv9xOMRhkaiXCa2czTGRls0rRalcPcpGkM8XgYbLMxwGbjp3icB6vtvyEeZ0BJCWOcTqabzexMJLjX52Ocz8f0svVJcUqLEHQwGCjUdR7w+Rjp8fBytalfU559lvsfe4zc3NykLdSPx+PcddddLF68mClTpnD99dcnJQ6Ruqx7ue3bWIzhHg93OZ2cYTbjVhSWhsPMrTZ1qiZ9irREglAoRFEgQGI3pczrS133UEqW9DwjSZPuznWuCXd3FntmLCt/W1JcTHFRERmZmVj2cIVH7Jnd4SAYCuH3+8nYz0o2ZrMZo9FIMBjEnJFRxxHWnbZt2wKlDc+OOOKIpMTg8XiIRCIpU9rW7XZjNpux2Wy0bNYMVJU/y24b63RyStnf1K0OB9eWlBDRdSx7OBGwKwrT3G6URIKdkQjvWCwkYjGecLuxKAqHG43kaxpja5hsLAwG6WAwcF9ZUtjBaOS3eJwZgUDFNjODQS63WhlSNlJyCDDZ5eLy4mKmulxYFIWrKk3LyCu7/cKiIgKJBI5KyfXo66/ntNNOq+EzV/cikQhDhw7lk08+4ZlnnqF3795Ji0WkrvYGA1ZFYWU0yoBqa3ZWR6O0MRi4vdLUo7+rlSDfl3g8TqDSdNhDrVYsoRDfx2K0KRvF8CQS/KVpnFK2T0eDgTil/ZvKp04VJxKs0zQOKztf7Wg0srrahfLV0SjtDYaK0YzGJL3P0tOgu3Odk+7OYjcMqlq1/K3bjc26t+s9ojpVUXA6nXi9Xux2+36NDCmUjmp4vV40lytlp7Ll5eUBsHHjxqQlGuXN+lJlRGNvOld6L7Qqe00LEwla76HpXiejEZOi4AkESksJqypHGI1VEpMT9vDdFYvH0XW9yojYOk3bZRrFcdX+/Us8zi/xOG9XWiOkAwlKR0Q6Go38GIsxLRDg53gcTyJB+WnXlkSCwyq9V485/PA9Phf1zefzcd111/H999/z4osvctZZZyUtFpHaLIrCcLudST4fJqCbycROXee/8TjtjUa2aBrLwmGOMRr5JBrl/Rr0oNKBaDRKMBAgEo1iUFWcDge2sh5I/YFJfj9ZqkpzVWWq31+lqtIhRiPnWyzc5fPxqMuFU1GY4vdzkKpWTJcaardzYVER0/1+elmtfBOLMT8U4uEkjS7Xt/RONIQQFVRFITMzE6/Xi8fjQdM0HA5HSq8VSDU2m41gMIjP5yMrK2u/njur1Yrf59ul1G0qadasGTabLamVp9KpK3jlU/ry98Tero3aFaW0klkohNvlQtG0Gh1Hp3SkR1EUsrOza/X+C+g619psDN5NNabWBgNBXad/SQlnmM0843bTTFXZomn0r1QNpyL+JF2kKCws5Oqrr2bjxo28/vrrdOvWLSlxiPQxyuHAqCg8FgiwvayM7UCbjf42GzfZ7dzr8xHVdc62WBjlcDBtDzNCdCAUChEMBonH45hMJjIyMrBarVX+Du9zOgn4fAwqKcGpKNxst+Or9vfzpNvNfT4fA0tKiAInm0y8kplZUVb3KJOJuRkZPBYI8GQwSEtV5W6Ho8pC8MZEEg0hGhGF0ikgBoMBv99PIpHA5XJJslFDCuByOikuKSEaiWCptmCvJiqXunU6HCm5eFVRlKSXuN22bRuqqtKiRYukxVCdyWRCq2FSsC9+vx+DwYDNbqdjMMib4XCV6Vbf7maNYflJTrNqxQQ6GAx8Wm00e021/Y8yGvlvPE67PYzE/RqLUZxIMM7pJKdsJOaHFFrn+Pfff3PVVVfh9/tZsmQJnTt3TnZIIg2oisLtDkeVKVLlxrtcjK82SjCkUhGGO51ORtrt+AMBgsEgiUQCi8WC2+3GZDLt9nvToarMzMhgZqXfDat27AxV5al9TJ292Grl4r0k9F83b77X/dNJao7rCyH2mwI4HQ7cbjehYJCSkhISadDbIVWYLRbMZjM+v3+/F/3Z7HYSiQShFC51m5uby8aNG5N2/Pz8fFq0aIEphaa/tm3blu+//57N27ZRpOt7HbXYm0QiQTgcLk00gd5lV0Xv8nr5bzzOJ5EIs6stStUSCfx+/26n7V1rt/OnpjHZ5+OveJzl4fAuNfdvtdv5JhZjnNfLz7EY6+NxPoxEGOf1AqWjGiZFYX4wyMayEpzTK63xSKY//viDXr16oWkaS5culSRD1LtYPI7H66WwsJBAIIDVaqV58+ZkZWZi3kOSIfaPJBpCNFJ2m43MrCyi0Sher1eSjRpSAJfLRTweJ1TtZK6mykvdBoPBlG2omJubm/SpU6k2beqWW27BYDBw+vXX08XvZ0stF4+Wi8XjGAwGrGVTIRyqykuZmfwaj3NuURFT/X7GV5tW5/P5UMrWCVWXazDwbEYG70UinFVUxIJQqOIKbvkISWeTibezslinaVxWXMy5RUU86vfTqmz0ollZ87IVkQinFxUxMxDg/hSYE/79999z2WWXkZmZydKlS2nXrl2yQxKNlA5EolGKi4vZuXMn0UgEp8NBixYtcLtcbAcOLSjY48+WOhrtbGpk6pQQjZjFbCY7OxtvURHBYBDPli20PuqoZIeV8kxGIzarFb/fj9Vq3a8mfhWlbqNRzClYBax86lQikUhKA7RUKm1brn379qxYsQLWroUePcBspl+1E/8jTSa2tmpV8e87nU7urLTNo3Y7ReEwTqezylXRriZTRa+NclvLpk5EolHC4TAZGRl7fK+dZ7FwXqWpfDMCAQ5W1SoLzI81mXhtL13KL7NauazadI3Kj6W7ycRWl6u0X1MD+Ne//sX111/PEUccwcKFC8lI4UptIn3puk4oHN7n+otWqrrX7uCtUrS4R6qTZ02IRs5kNJKRkYEC3HrrrXz33XfJDiktOJ1OdF0nWG2KS01VLnWbinJzc4lGo+zYsSMpx9+2bVvKjWjUBb/fj9FoxFrDBdW6ruP1ejGbzXvd58VgkDWxGBvjcd4MhZgdCFR0Q05H7777Ltdccw0nnXQSr732miQZos5piQT+QIAdhYV4vd6KUvDZ2dnYqiUZAEZFoZ3RuMcfYwqut0sHMqIhRBNgVFVUu502bdpwxRVXMGfOHM4777xkh5XSDAYDdrudQCCAzWardanayqVu45qGcQ+lUJMlNzcXKC1x26rSVe2GkopTpw5UNBolGo2SmZlZ4znegUCAB+NxPkgkoKBgl9v7WK084nazXtOYEQhQrOu0VlVuttu5bTcLYNPBK6+8wpgxY7j00kuZMWNGSq3TEekvFo8TrNT/wmazYbfbU+4zuKmQREOIJkJVFB5//HFCc+Zwww03MGXKFAYNGpTssFKaw+EgdABN/MpL3YZCoZQrdVvetG/jxo2ceOKJDXrscDhMSUlJyk2dOhA64PP7MZlMNa5WFtc0AsEgd9nt3FWpGk5lzrIE90GXiwdTYE3FgXrmmWeYMmUK1113HZMnT07KtD3R+JT3vwgEAkTL+184ndhstv2a+irqjiQaQjQhFouFuXPn8sADD3DPPfewZcsWxo4dK1/2e6AqCg6HA5/Pt19N/FRFwWa3EwoGcTgcKfWFZ7fbadGiRVIWhG/btg1Ijx4aNRWNRonFYjXuv6IDXq8XVVXJc7lSsgxyXdJ1nSlTpjBr1izuuOMO7rzzzkb/mEX9q+n6C5E8kmgI0cQYDAYmTpxImzZtePDBB9m6dSvTp0+X6Qt7YLfbCQaD+P1+sjIza7+/zUYgECAcDmNPsTn1yao8tXXrVqDxJBo6pWszzCZTjRf+h8NhotFoaWLSyE+44/E4o0eP5rXXXmPSpEnceOONyQ5JpLlEIkE4GMQfjZJIJLDuo/+FSB5JNIRoghRF4eabb+bggw/mtttuo6CggOeeew73fkwPauzKy92WlJQQiUax1LKClMFgwFpW6tZms6XUl2CyEo3yEY3GMnUqEokQi8Vq3M07oev4fD6sVmut30/pJhKJMGzYMD766COefvppLr/88mSHJNLYr7/+yopHH+WWQICoomC122X9RYqT+RJCNGGXXnopr7/+Oj/++CO9e/cmPz8/2SGlJIvFgslkwufz7VdfDLvdTjweJ1qtu3Oy5eXlJaVp37Zt23C73dj3sC4hnVSMZpjNmGs4Kuj3+9F1HVcjWHOxN36/n2uvvZZPP/2UF154QZIMsV8SiQSffvop/fr14+yzz+bbb7/FYrGQlZWF2+WSJCPFSaIhRCMyze/nnJ07a7XPySefzPLly/F6vVx88cX8+uuv9RRd+qrcxC+8H92+TSla6jY3N5dt27YRiUQa9Lhbt25tNNOmwuEw8Xh8t432dicWixEMBnE6nbWuZJZOdu7cSd++ffnhhx947bXXOOecc5Idkkgz4XCYl19+mTPOOINrrrkGr9fLrFmzePnllzGbzSm15k3sWeP9lBOigcXSuPP2YYcdxjvvvEOzZs247LLLWLlyZbJDSjlmk6miipRey9e6vNRtJBIhnkLdZctL3P79998NetzG0kOjfDTDYrHUaDSjfAG4yWhsFKM5e7JlyxYuu+wytm7dyttvv81JJ52U7JBEGikoKODRRx+la9eujBkzho4dO7J06VLee+89LrvsMllPmGZkjYZo0hK6zpxgkJdDIbYkErRQVa612bjd4WCyz8f7kQj5iQQtVZXLrVZGORyYyq6iTPP7eT8S4Qa7nRmBAH9rGlv20Y/gnXCYJwIB1msaNkWhi9HIi5mZ2Mvuc1EoxJxgkE2aRltV5Ua7nesqnZDsLabXQyGmBQIA5GzfDsB0t5t+NVyA3KpVK5YsWcKQIUO4+uqrefLJJ+ndu3etn9PGzOl0snPnTgLBIM5a9jAo7zQeCgZTZspMXl4eUFritkOHDg123Pz8fA4//PAGO159CYVCaJpGZg2LBASDQWLxeI3XcqSjP//8k379+mE0Glm2bBnt2rVLdkgHbv168PuTHUXDczrhkEMa7HC//vor8+bN4+2338ZkMtG/f39uvPHGxvEeasIk0RBN2kN+P6+EQjzocnGiyURBIsGfZVecnarKDLebVgYDv8Xj3OX14lQUhlU6wdygabwbDvN8Rgb7miW6XdMY5vEw3uXiQosFfyLBV7FY6dVxReHtUIjH/H6muFx0MZn4KRbjLp8Pu6JUdADeW0y9rFZ+j8f5LBrl9bITH3ctp2Y4nU4WLFjA6NGjGT58OFu3bmXYsGGNvipOTRkNBmxlVaRq28RPURRsNltpqVunMyWG/Q866CBMJlODLwjPz8/njDPOaNBj1jWd0mZ7Vqu1RmWPtUQCv9+P3W6v8VqOdPPDDz8wYMAAWrZsyWuvvZaURpB1bv166NYNYrFkR9LwTCZYvbpek41EIsFnn33G3LlzWblyJQcffDBjxozhmmuukeIkjYQkGqLJ8icSPBcK8ZDLVXEi3w4ob102slJC0dZgYKjdzrJIpEqiEQOeysigWQ1OOAsSCeLARRYLbQwGMBjoXOmE4/FAgPtdLi6yWgHINRj4r6axMBSqiG9vMVkVBYeiYABaHsDiOJPJxBNPPEHr1q2ZMmUKW7ZsYdKkSRhkwR1QmoyFQyECgQDuWo5MlHcaT5VStwaDgdatWzdoohGPx9mxY0faT50KBYNomkZWVlaNtvf5fCiKUuO1HOlm5cqVXH/99XTq1ImFCxfWeJQn5fn9pUmGqkIt++iktXi89HHX00hOKBTizTffZN68eaxbt45jjz2W2bNnc9FFF8nUqEamCf3VCFHVH5pGVNfpsYfyksvDYZ4LBtmoaQR0nTjgqnYVurWq1ijJADjCaKS70chZO3dyhtnM6RYLl1gsZKgqQV1ng6Zxh9fLXV5vxT5xwF3pmDWJqS4oisJdd91FTk4OY8aMIT8/n1mzZmFLgZPjZFMVBYfTib+siV9tKp4YVBWr1ZpSpW5zc3MbtPJUYWEhmqaldWlbXdfxl41q1eT1j0SjhMNhMjIyUmIkq669//77DB06lO7du/P88883zvUnRmPpFf6mpB6q5G3fvp0XX3yRBQsWUFJSwoUXXsgTTzzBCSecICPnjZQkGqLJsu7ltm9jMYZ7PNzldHKG2YxbUVgaDjO3WtUgey0+GPVEgpmxGD8bjXxnMDA/GGSq38+72dmUn74/7nJxfLUvs/ITk5rGVJcGDBhAq1atuOmmm+jbty8vvfQSzZo1q7fjpYvyJn4+n6/WTfzsdjtFRUVE96MnR33Iy8vju+++a7DjlZdQzsnJabBj1rVgKISeSOCowTqd8gXgZrMZq3Vvnzrp6bXXXuOuu+7ikksuYebMmXI1WuzWL7/8wrx581iyZAlms7li/UX5OjHReEmiIZqs9kYjVkVhZTTKgGpX6ldHo7QxGLi90onE34nEAR3PaDDQrFkzjvF4OCoUYqjDwRnBIO+Hw9zscNBKVdmoaVy+h1GDmsRkUhQOLMpdnX322bz99ttce+219OzZk0WLFqXu4rx4vEEOowAumw2/z0csFKrRHP1yJsCqqoT9fiwHuii8Dh5vXl4eS5cuRdf1BrmiWJ5opOuIRkLXK9bo1GQ0IxAIkEgkSjuAN0B8DWn27NlMmjSJgQMHMmXKFJleKaoo738xb948Vq5cSU5ODmPHjuXqq6+W9RdNiCQaosmyKArD7XYm+XyYgG4mEzt1nf/G47Q3GtmiaSwLhznGaOSTaJT3D7DXwHexGCujUU5zu7GFQnzo87FD12lfdnJ3t9PJeJ8Pt6pyptlMRNf5MR6nJJHgZoejRjG1NRjYpGn8HItxsMGAU1Ew18HJ4zHHHMOKFSu4+uqr6dmzJwsWLOC444474PutM05n6bSGWKxehvt3xwJoBgNRnw+jw1Hjk0gFcJhMhMNhEuEw6oH2UjCZSh//fmrbti0+nw+Px9Mg8+q3bduG2Wyu8dqGVBMMBtF1HUcNnvO4phEIBBpd52Jd13nooYd45plnGDlyJHfffbdMexEVQqEQb7zxBs8++yzr1q3juOOOY86cOVx44YUy4tUESaIhmrRRDgdGReGxQIDtZSVjB9ps9LfZuMlu516fj6iuc7bFwiiHg2kHsDDOpSj8Jxrl2WAQn67T2mDgbl3naL8fv67T3+HApijMCgSY6PdjBzobjQwpm+98nsWyz5gutlh4z2ymT3ExXl2vVXnbfcnLy2PFihUMGjSIPn36MHfuXM4999w6ue8DdsghpdVRGrAEpQL8+cMPjBo1ivHjx3PWWWfVeF89GmVgWZfbW2+99cACOcASlJVL3DZEopGfn8/BBx+cliemCV0nWMOKY+VTplRVrXUp5FSmaRpjxoxh0aJFPPjggwwZMiTZIYkUUb7+4qWXXsLr9XLhhRcyffp0unbtmpZ/76JuSKLRiEybNo3333+fjz/+ONmhALBq1SquuOIKfvvtt5QdJlUVhdsdjirTkcqNd7kYX21qy5BKixzvdDq5sxZXkjsajSyqdhW3vOFXwO8nEg5zSUYGvfeyBmJfMZkVhWfr8WQxKyuLxYsXM3z4cK6//noefvhhrr322no7Xq00YL33csccdRR5n3zC2Fde4V9Dh2KxWGq0nxk4afBgnpo/n+umTUtqJaLypn0bN27kmGOOqffj5efnp+20qUAggA41WpsRCYeJRqOlU6YayUlWNBpl+PDhfPDBB8yYMYO+ffsmOySRAn7++WfmzZvH0qVLZf2F2IV0Bm8EdF0n3kBz00XdUgCX00l2s2boQNHOnfj9flK5x7jVamXevHlcd911jBkzhqlTp9a6U3ZjMn78ePLz85k/f36t9hs0aFDFFINkysjIwO12s3nz5gY5Xrp2BdcSCYLBIHa7fZ+jGQldx+vzYbVaU2LBf13w+/1cc801/POf/+T555+XJKOJSyQS/POf/6Rv376ce+65rFq1invuuYdvv/2WiRMnSpIhKtQ60fjmm28455xzaNeuHRdccAEffPABOTk5/PzzzxXbfPTRR3Tv3p127dpxxRVXsHjxYnJycvCWle0sLi7mlltu4bjjjqN9+/acddZZLF26tMpx/H4/w4cPp0OHDhx77LHMmzePPn36MGHChIptorrORJ+P43bsoENBARcXFbGqFvOzV0ej9Ckupn1BAZ0KCuhfXIynbHFtVNcZ7/Nx1I4dtCsooFdREWsqNexZFY2Ss307n0cinLtzJ4cUFNC3uJjCRIJPIxFO27mTwwoKGObxEKp0EtanuJhxXi/jvF4OLyjgyB07eNTvr3Ki9mYoxAVFRXQsKOCYHTsY5vFQWGnR76o1a8jJyeHTTz/l/PPPp127drz11ltMmzaNX375hZycHHJycnj99deB0uouCxcuZODAgbRv357TTjuNb7/9lg0bNtCnTx86dOhAz5492bBhQ5Xn58MPP+S8886jXbt2nHzyyUybNq1KQpOTk8OiRYu44YYbaN++Pd27d+ejjz4CYPPmzVxxxRUAdOrUiZycHEaOHFnj1yYdbdE0Di0o2OPPlrJGgLtjMhpp1qwZDqeTQCDAzp07iaVwgyiDwcCkSZOYMGECTz31FLfffntKx1ufOnTowLXXXsuMGTMoLi6u8X4HH3wwF198Mc8//zyJAyw0cCAURWnQErdbt25Ny0QjEAiUrq+pwWiGv+wzPVU6wB+ooqIirrzySn744QdeffVVzjvvvGSHJJIkFArx0ksvcdpppzFo0CCCwSBz5szhyy+/ZOjQoSk7e0EkT60SDZ/Px8CBA+ncuTMfffQRo0ePZvLkyVW22bRpE0OGDOHCCy/kk08+4dprr2Xq1KlVtolEIhx99NG8/PLLfPbZZ1xzzTXcdtttfP/99xXbPPDAA6xevZqXXnqJ1157ja+++oq1a9dWuZ97g0G+icWYk5HBJ9nZXGKxMKCkhPU1uLr/cyxG35ISDjMYeCcri2XZ2ZxnsVB+KjjJ7+fdcJgZbjcfZWdziMFA/+JiSqqdEEwLBHjI5WJFVhZbNI2bPR6eDQaZ5XazMDOTL6JR5lcrP7o4HMagKLyXnc0kp5O5wSCLwuGK2+PAaIeDj7OzmZ+ZyWZNY6THs8tjmDJlCuPGjeOLL77g9NNPZ+jQoRx++OGsWbOGNWvW0KtXr4ptp0+fzhVXXMHHH3/MoYceyrBhwxg9ejS33XYbH3zwAQDjxo2r2P6rr75ixIgRDB48mC+++IJHH32UxYsXM2PGjKqPf9o0Lr30Uj799FPOPvtshg8fTklJCTk5OTz33HNAaSOnNWvWMHHixH2+LumslarycXb2Hn9a7eMqqAI4HQ6ymzVDURR2FhXhq5aEphJFURg6dCizZ89m2bJlXH311RUXE5qaO+64g0QiwZNPPlmr/W688Ub++usvPv/883qJq6Zyc3MbpGmfruts27Yt7aZOaYkEoVAIu8Oxzz4YsXicYDCI0+msVef4VJWfn0/v3r3ZvHkzb731FqecckqyQxJJsH37dqZOnUrXrl0ZN24cRxxxBMuXL+edd97h0ksvxdiUmhmKWqnVp+CSJUtQFIXHH3+cww47jLPOOothw4ZV2WbhwoV06NCB++67jw4dOtCrVy/69etXZZuDDjqIW265hSOPPJK8vDxuuOEGzjzzTFasWAGUXg164403mDBhAj169KBTp05Mnz4drdIV4S2JBK9HIjybkcFJZjPtjEZucTg40WTitUon7XvyTDDIMUYjD7vdHGEycbjRyPV2O9llzdMWhEJMcLk4y2LhMKORx9xurIrColCoyv2McTrpZjbTxWRigM3Gl9EoU10uuphMnGQ2c4nFwr+rjbLkGAw86HTSwWjkcpuNG2w25lVKRq6y2TjLYiHPaKSrycRkl4tPo1EC1U44R48ezWmnnUa7du046KCDcDgcGAwGWrZsScuWLavUbL/qqqu49NJLad++PcOHD2fz5s1cfvnlnHHGGXTs2JHBgwezatWqiu2nTZvGrbfeypVXXkleXh6nnXYao0eP5uWXX64SQ79+/bjsssto164d99xzD4FAgO+//x6DwVCxsLR58+a0bNmy0V/pMCoK7YzGPf4YazhP22Q0kp2djcvpJBgMsrOoiGgKjxb06tWL1157jR9//JHevXuzbdu2ZIfU4Jo3b86tt97Kiy++uMvI4N6ccMIJHHXUUTz//PP1F1wN5OXlNUii4fV6CYfDaTeiEfD7URRln43oyheAG43GRtG0bt26dVx66aUEg0GWLVtGly5dkh2SaGB//PEHI0aM4MQTT6yYMrdq1SrmzZsnTfZEjdQqBV23bh1HHHFElQWP1Utcrlu3bpcFhdW30TSNp556ihUrVpCfn08sFiMajVZ0Hd60aROxWKzKfm63mw4dOlT8+9dEAg04defOKvcd1XWya3AV6ed4nJ57WLi5MR4nput0q1SGzaQoHGcy8YemEQgGCZeNmnSulMW3UFVsikJetd99X+0k8XiTqcof5wlmM3ODQTRdx6Ao/BiLMS0Q4Od4HE8iUdEXYUsiwWGV7qc2Czc7d+78v5hatABKpzSVa968OZFIBJ/Ph8vl4pdffmH16tVVRjA0TSMSiRAKhSpeq8r3a7fbcblcFBYW1jgusXvlUzQsVisej4eioiLsdjtOpzMlOwufcsopFaMaF198Ma+88kqV91dTMGTIEF588UUefvhh5s6dW6N9FEXhxhtvZOTIkaxbt67KZ1xDatu2LX///TeaptVrL4StW7cCpFWiEdc0QqEQTpdrn397oWCQWCxGdnZ22vfMWLt2LQMGDKBZs2a8+uqrafWaiQOjA7FolHgwyM0334yvXTvuvfde+vfvn1oXDJva2tQ0fbxJGeuaPXs2zz33HBMnTqRz587Y7XYmTJhAtBbrK4K6jgH4MDub6l+LNenWfCD9WeOxGMFQiISu4yspwWazYS1LWqpXiFagVg3UgrpO/5ISzjCbecbtppmqskXT6F9SQqzaiEZtrphVrl1dnuTs7nfl03QCgQB33303F1100S73VTnRrF4TW1GUpM43b2yMBgPZ2dkEg0H8fj+RSIQMtxtzCi4wPfzww3n33Xe5+uqrueyyy5g/fz7du3dPdlgNxmazMXbsWEaOHMlNN91E165da7TfZZddxuTJk5k/fz5Tpkyp5yh3Ly8vj3g8Tn5+Pm3atKm345SPdqXTSWsgEEBRVez7KBOtJRL4/X5sNhvmNO8VsGbNGq6cNImOHTvy8ssvp23PE1E7CV0nHA4TDAQgHsdmMJTOLLnlltSaGpWEvkkp4wD7JiVDrd45HTp04K233iIajVac6KxZs2aXbT799NMqv6u+zerVqzn//PPp06cPUFq9YN26dRx2WOn1+tzcXEwmE2vWrKF169ZA6XD0X3/9xcknnwxAF4MBDShMJDhpP066jjCZ+Fc0yl27uS3PaMSkKKyOxWhTdnUvpuusicUYYreT4XCQabWiFBejAD6vFx/gNxjQdZ2Eru/1ylf1EY5vYzEOMRoxKAp/xmIUJxKMczrJKTv2DzWcNmMymersJP+oo47izz//PKAO0OXvEW0vi6DFvimAw27HYrHg9XopKi5O2dGNVq1asWTJEgYPHkz//v2ZMWMGl112WbLDajB9+vRh3rx5PPjggyxbtqxG0wrMZjPXXnst8+bNY8yYMUm5Yli5l0Z9Jhr5+fkoilIxqprq4mVrM9wu1z5fS5/PB4qS9gvAY/E4Y8aM4fjTTuP555+v0eJ3kd7KK6qFQiESiQRWqxWHw4EpkeCMM86AVEoyICl9k1LGAfZNSoZavXt69+7N1KlTufvuu7n11lvZsmULs2fPBv53Rbz8C3Py5MkMGDCAn376qaL6UblDDjmEd955h2+++YaMjAzmzp1LYWFhRaLhdDrp27cvkyZNIjMzk+bNm/P444+jqmrFcdqrKpebzYzwernf6aSLycTORIKV0SidjUbO2Uc9+9vsds4sKuIer5eBNhsmReHf0Sg9rVayVZVBNhsTfT4yFYXWBgOzAgFCus6AsqtaqqKgKApZmZk4gXAkAoEAWiLBjoICzBYLVqt1t2VKt2gaD/h8XGuzsTYeZ34wyP1lGWprgwGTojA/GORam43fNY3pgUCNXp+2bduyadMmfv75Zw4++GCcTud+X/m+4447GDhwIK1bt+aSSy5BVVV++eUXfvvtN8aMGVOj+2jTpg2KovDPf/6Ts88+u+LDS+wfo8FAVlYWoWAQX9nohtvtTrnymS6Xi5dffpm77rqLYcOGsXXrVm655ZYmMZfXYDBw//33069fP9577z0uvvjiGu03cOBAZs6cyeLFixk8eHA9R7mr8r/V+i5xm5+fT4sWLdKmO3DE78esqtjKr57uQTQWI16WkKhpOr0BIBIKEQ2F6H7RRdyzYEFKjpyKuhOLxwkGAqXnL4DdZist32wwpP5oQZqdbDdltVoM7nK5WLBgAT/99BPnnHMOU6dO5Y477gD+N50mNzeXZ599lvfee4+zzjqLBQsWcPvtt1fZZuTIkRx11FH079+fPn360LJlSy644IIqx3rggQfo2rUrAwcOpF+/fnTr1o2OHTtWmbYz3eGgr9XKg34//9i5kxtKSlgTi9G6Bms02huNvJaZyc/xOBcWF3NJUREfRiIVmdc4p5OLrVZu83o5r6iI9ZrGq1lZZOzmvtWyYXWnw4HRYMDldqPrOh6PB7/fTzweJxQOkyibltTXaiWs61xUlugMttu5piyBaaaqzHC7WRGJcHpRETMDAe6v4RWyiy++mDPPPJM+ffrQpUsXlixZUqP9dueMM85gwYIFfPHFF1x44YVccsklzJs3r1ZXOw866CDuuusuHnroIY4++ugqVa3E/lEonTLXvFkzDAYDxcXFeLzeivdWqjCZTDz55JOMHDmSyZMnM378+CYzsvWPf/yDM888kylTptS45G+rVq245JJLmD9/flKmHprNZg466KB6L3GbNj00nE7iqkoiFsNlMqFEo6UnXbv50SMRYn4/doMBC+xxu1T/ifr9RINBVKuV8Q8/LElGI6VTWvmzqLiYnTt3Eo3FcDqdtGjRApfLVa9rtETTpOgHWDvz7bffZtSoUfz+++9VqhxVNmPGDBYsWMC3336738cJBoMcf/zx3H///fTv0gV69ACzuXS+WorSEgnC4XBph9hYDEVRGKzrdDGbmZKZWftpL+VXGFauhKOOqp+gG7u1a9PivVMTOqU1zf0+H4qilI5u7GkkL4nvnZdffpmxY8dy3nnnMWvWrD1+TjQmv/32G+eccw4PPvggN954Y432+fbbb+nZsycLFizgnHPOqecId9W7d28OPvhgZs2aVW/HuOaaazCZTLzwwgv1doy6cufll7N93TpeeOGFvY7ALFiwgIULFzJv3jwOScOrrLquM3/+fF555RUGDBjAjbffjtK+fbLDSh2N5Duj8vqLuKZhMplKC45YLLsvXCDnG6KO1Hri3RtvvEFeXh4HHXQQv/zyC5MnT6Znz55VTh5efPFFjj32WLKysli9ejWzZ8/m+uuvr9VxfvrpJ/7880+OPfZYfD4fTzzxBADnn38+bNlS27CTwqCqOOx2HHZ7RdKBz0ckHGbHjh1YLBasFgtmiyXl5tqL1KdQOtRdvnajuKQEq9WK2+1OqffTNddcQ6tWrRg6dCh9+/blpZdeIjs7O9lh1atOnTrRr18/nnjiCfr27VujdRfHH388xx57LM8//3xSEo28vDzWrVtXr8fIz8+vWGeXyr777jte/c9/eOaZZzAdf/wet9uwYQMT3niDm0aM4JBLL23ACOuGpmnce++9LFy8mAmTJjF46NBkhyTq2O7WX7gzMtK+YIFIH7VONHbs2MFjjz1GQUEBrVq1omfPnowdO7bKNuvXr6/oktu6dWtuvvlmbrvttloHN3v2bNatW4fJZOLoo49myZIlpScoNUg0ri4u5qs9TFsY4XAwooHXCpQnHeZIBLuq4jSZCIfDlHg8KIpSmnRYrVjM5iYxl13UHYOqkpmZSTgcxufzUVhYiNvtrqiElgrOPfdc3nrrLa699lp69uzJokWLKhYgN1ajR49m6dKlPPXUU4wfP36f25eXur3tttv4448/6NixYwNE+T+5ubm7FPKoa/n5+WnRrO+RRx7h8MMP59K9JA+6rnPvvffSsmVLRo0a1YDR1Y1oNMptt93Gu+++y/Tp03fpdyXSW8X6i3AYFKXq+gshGtABT51KihoMZW7TNMJ7eGiZqkpminRsjWtaxfSqWDxeNemoPqQpQ5kHrpEMg++Jlkjg83oJRyJYrdbSObeqmjLvnQ0bNnD11Vfj8/lYsGABxx57bNJiaQjTpk1j5syZrFy5skbrm6LRKN26deOiiy7i4YcfboAI/+fNN99kxIgR/Pnnn/XSbC4cDtO+fXtmzJhB37596/z+68qqVau44ooreO6553Zb3rvcihUruPnmm3nppZc499xzGzDCAxcIBLjxxhv5z3/+w5w5c3ZZIykqSaPvjPL1F8FgkGg0isFgwG63Y7PZZKq2SJrUONuuBwcZDHvs0JwqSQaUVhJyOhw0a9aM5s2b43A4iMfjlJSUsKOgAI/XSyQS2W31KiGqM6gqGZmZZGZkEI1G2blzJ6FwOGXeP+3atWPFihXk5eXRp08f/vnPfyY7pHo1dOhQMjMzmTp1ao22N5vNDBw4kMWLF+P1eus5uqrKR5jqq0N4OvTQ0HWdRx55hKOOOooLL7xwj9v5fD4mTJjABRdckHZJRnFxMf369eO7777jlVdekSSjEUjoOsFgkMLCQkpKStB1vaJip8NuT6mptKLpSZ0zblGRdDRv1ozmzZphdziIxWIUlyUdvkCAeDxe40o2omlSAKvVSvNmzTCbzXg8Hrw+X8o0UszOzmbx4sWcfvrpXH/99SxcuDDZIdUbh8PB3Xffzdtvv80PP/xQo32uvfZa4vE4r776aj1HV1Vubi5AvZW4zc/PB1I70fj8889ZvXo1Y8aM2esU1sceewyv18ukSZMaMLoDt23bNnr37s2GDRt48803OfXUU5MdkjgAWiKBz++nsLAQr8+HyWQiOzubZtnZWPe0yFuIBiaJRooyGo0VIx3NmjXDZrcTi8UIhUL07duXu+++m5UrVxJP45rton6pqkpmRgaZmZnE43ECwSAffPABqTBb0maz8eyzzzJo0CDGjBnDI488khJx1Yd+/fpx2GGHMWnSpBo9xpYtW3LppZfywgsvNGhJ4BYtWmCxWOqtxG35iEaqrtEoH8044YQTOPPMM/e43U8//cT8+fO58847KxrKpoP169dz6aWX4vf7WbZsGUcffXSyQxL7KRaP4/F4KNyxg1AwiM1mo0Xz5mTKIm+RgiTRSHEKYDIacTmdZGVmYrfbufjii/nXv/7FlVdeyfHHH88999zDqlWrmkyfAlE7VouFrMxMjEYjjz76KAMGDGBLClRuMxgMFT02ZsyYwciRIxvlaJ3RaOS+++5j1apVfPzxxzXa58Ybb2TTpk188skn9Rzd/6iqSm5ubr0lGvn5+bjd7pRt2vnBBx/w448/Mnbs2D2OZmiaxpgxYzjssMMYMmRIA0e4/3766Sd69eqF1Wpl+fLldOjQIdkhNRl9iouZ4PMd8P3olDYGrtL/wuWieYsWuJzOXRZ552zfzgfh8AEfV4gDJYlGGlEoPTkbMmQIX375Je+99x59+/bl448/5oorrqBr166MGzeOr7/+OmWmyYjUoCoKNquVhx9+mP/+978VDRmT/T5RFIVhw4Yxa9Ysli5dyjXXXIOvDr6UU81ZZ51Fjx49mDRpUo1GIY899li6du3Kc8891wDR/U9ubm69rdFI5YpTmqbx6KOP0qNHD7p3777H7V555RW+//57HnnkkbTpbv6f//yHPn360KZNG5YuXUpOTk6yQxK1IOsvRLqTRCNNKYrCsccey3333cfXX3/NihUr6NWrF++//z6XXXYZJ5xwAvfffz/ffvtto52SImrvpJNO4rPPPqN3796MHTuWK6+8kg0bNiQ7LC677DIWLVrEDz/8wGWXXVYxzaaxUBSFCRMmsG7dOhYtWlSjfW688UZWrlzJ77//Xs/R/U99Jhqp3BV8+fLl/P7774wZM2aP2+zYsYOHHnqIq666im7dujVgdPvvn//8J/379+eYY45h8eLFjb5/TWNSsf5ixw58sv5CpDFJNBoBRVHo2rUrDz74IN988w1Lly7loosuYtmyZfTs2ZMTTzyRSZMmsWbNGkk6BG63m0cffZTFixezefNmzj77bJ577rmkT7079dRTWbp0KSUlJVxyySUNeoLdELp06UKfPn14/PHH8fv9+9z+4osvplWrVsyfP78BoitVnmjUx+dEqo5oxONxHnvsMc4991y6du26x+0mTpyIwWCoUU+UVPDmm29yww03cPbZZ/Pyyy/jdDqTHVKTlQAm+3x0LijgmB07mFbp73+LpnFdSQmHFhRwWEEBQ4qL+bOk5H/rL+x23nM6uTAWo2NxMT0KC3kzFKpy/+vjcXoXFdGuoIDTd+7k/yKRBn6EQuyZJBqNjKqqnHjiiUyePJlvv/2Wt956i3POOYfFixdz0UUXccoppzBlyhR++uknSTqauB49evDpp5/Sv39/JkyYQO/eveu9M/S+dOrUiXfffZeMjAx69erFqlWrkhpPXRszZgw+n49Zs2btc1uTycTAgQN544038Hg8DRBdaYnbUChEYWFhnd93fn5+Sk7befPNN9mwYQOjR4/e4zb//ve/eeutt7jvvvvSYlTgueeeY8SIEVx55ZXMnTsXs9mc7JCatDdCIWyKwnvZ2Yx3OnkiEOD/IhESus51JSWUJBK86nAw12Dgz3CYkeFwxfqLlUYjDwQC3Gy381l2NtfabIzyevl3NAqUTq260ePBpCi8l5XFIy4Xk2twIUOIhiKJRiNmMBg45ZRTePjhh/n+++9ZvHgx//jHP1i0aBHnnXcePXr04JFHHuHXX3+VpKOJcjgcTJ48mSVLlrBz507OPvtsZs2aldRqZgcddBBLlizh6KOPpn///ixbtixpsdS11q1bM2TIEObMmVOj6WHXXHMNmqbVeLrVgaqvXhqaplFQUJByIxrRaJRp06bRs2dPjjzyyD1uM3bsWLp168aVV17ZwBHWjq7rPPbYY0yYMIFhw4bx+OOPSyfoFNDZaOROp5NDjEb62mwcYzLxr2iU/4tG+TUWY5Km0dbv5xhF4amMDL5TFP4wmVAVhdnBIFdarVxnt9PeaORmh4OLrFbmBIMA/Csa5c94nJluN0eYTJxsNnOPjF6JFCKJRhNhNBrp0aMHjz32GGvWrGHRokWcdNJJvPDCC5x99tmcfvrpTJs2jf/+97/JDlUkwUknncTHH3/MjTfeyEMPPUTPnj357bffkhaP2+3mlVde4dJLL+WWW25h9uzZjSYZvvXWW7Hb7TzyyCP73LZFixZcdtllDVbqtm3btkDdJxqFhYVompZyicYrr7xCfn4+d9111x63mT17Nhs3bmTq1KmoKdTstTpN07j33nuZPn0648ePZ/z48XvtBSIazhFGY5V/t1AUtkYifF9cTEtdp63ZTLPsbLKzsznKbsetKPwWjRIKh/kjHqdbtcID3Uwm/ii7GPSHppFjMNCqUkLZNU0KFYimIXU/NUW9MZlMnHHGGTzxxBOsXbuWBQsWcNxxxzFv3jzOOOMMzjrrLKZPn85ff/2V7FBFA7LZbNx3330sX76cUCjE+eefz/Tp05NWctZkMvHUU08xYsQIJk2axH333Zf0dSR1we12c+edd7J48WJ+/fXXfW5/44038vfff/PRRx/Ve2xOp5Ps7Ow6L3Fb3qwvlaZOhUIhZsyYQZ8+fejYseNut9m4cSNPPvkkN910E507d27gCGsuFosxfPhwFi5cyLRp0xg2bFiyQxKVGMsSvlg8TonHQyQcJhaPYzabMRqNZGZkYDKZKkrYaokEPp+vRmu5hEh1kmg0cSaTiXPOOYcZM2bw448/8uKLL9K5c2dmzZpFjx49OPfcc5k5c2ZKVCYSDeP444/no48+YtiwYTzxxBNceOGF/PTTT0mJRVEUxo4dy9SpU3nxxRe56aabCDeC2vDXXHMN7dq1Y+LEifvc9uijj+aEE07g+eefb4DI6qfyVHmikUojGi+++CJFRUWMGjVqt7frus64ceNo1qzZHrdJBcFgkEGDBvH+++8zb948+vfvn+yQRDXxeJyioiJ27txJLBbDZDJhs9k4yuEgP5Hg73icQCBAYWEh3xYX49N1jnY6ad68OR2NRlZXu9izOhbjsLJRko4GA1s1je2VLsJ81wj7EYn0JYmGqGCxWDjvvPN45plnWLt2Lc899xwdOnTgySefpHv37lxwwQXMmjWLzZs3JzvUAxePQyzWdH5quebCbDYzZswY3n//fQAuvPBCHnnkEaJlCxAb2sCBA3nhhRf47LPP6Nu3L0VFRUmJo66YTCbGjx/PF198weeff77P7QcPHsyqVatqNAJyoOoj0di2bVtFec5U4Pf7efrpp7nqqqto167dbrd57733+PTTT5kyZUrKNhksKSmhX79+rF69mldeeYWLLroo2SGJMuX9LyKRSMXFkfL+F0ajEQU4WVU5FLhpxw5W+/38ajAw0WjkVKuVk5xOFGCY3c7icJiXgkHWx+PMDQR4LxxmqN0OwD/MZtobjdzu9fJLLMZX0ShTA4GkPW4hqpNEQ+yW1WrloosuYs6cOaxdu5a5c+fStm1bHnvsMU466SQuueQS5s6dy9atW5Mdau04nfD/7J13dBTV34ef2ZrdbDqEKr0XUYqAEEAQASFBUYpIkdAEFaUXaSIgIIiiSG9Kx0ICiBTpglRFpIkIyEsLqZvN9p15/0h2fwlJIIFUmOecOZtMufdumd37ud+mVoMogt3+5GyimPy8sxkkWKtWLbZv386wYcP4+uuveemllzh16lQuvTn3p3Xr1nz//fdcvXqVsLCwXKtgnVe0bdvWk3r6QS5h7dq1o3jx4nli1ciN6uA3b96kRIkSBSZmYMmSJSQlJWVqqTCZTEyYMIHWrVvTpk2bPB5d1rhz5w6dOnXi33//ZdOmTTRt2jS/hyRD+voXCoUCnU5HYEr9C0iOp7HabMTGxvKFQkGAWk0/QaCfw0E5tZqFvr6e9tp6eTHFYGCh2UyL2Fi+tViY6+vL8ymZxBSCwHI/P6ySRLu4OIYbjYwpoMJY5slEkApjhOWZM9C0KWg0yZOnJwWHI3nSeOgQ1K6dL0NISkpi165dREREsGfPHhwOB/Xr16djx4506NCBYsWK5cu4ssWVK/Ak+r4aDFC+/ENffv78eYYNG8aZM2cYOHAgI0eOxMvLKwcHmDWuXr1K9+7dMZlMfPvtt9SpUyfPx5BTnDp1ig4dOjB37ly6du1633PnzZvH3LlzOXXqFAEBAbk2pjVr1jB69GiuXLmSY9Wv33vvPa5fv87mzZtzpL1HIT4+noYNG9K1a9dMXdc++ugjVq1axf79+z0B8gWJq1ev0q1bNxwOBxs2bKBSpUr5PaTHlyzONxwOB0lmMzarFUEQ0On16PV6lCkJBFyiiMViwWI24xJFNBoNer0ebUEtvlcA5hsyjwey0ChMFLAb32g0snPnTrZs2cK+fftwOp00bNiQsLAw2rdvT9GiRfN7iDI5jNPpZPHixcyaNYtSpUoxd+5cnnvuuTwfR0xMDL179+b8+fMsXryYVq1a5fkYcoq3336bo0eP8uuvv6JPcYfIiJiYGOrVq8fIkSN55513cm08hw4dokuXLhw+fDhTt6Ls0rlzZ4oUKcKCBQtypL1HYcaMGSxZsoTffvstw++oc+fO0aZNG8aMGZOrr/PDcu7cOd544w18fX1Zv349pUqVyu8hPd7cZ74hATabDXNSEnaHA6VSibdej5dOh8IdAO5wYLZYPO5TXl5e6PV61PdkoipwFLD5hkzhRXadknlofH19ef3111m1ahV//vknn332GTqdjokTJ/Lss8/SpUsXvv3220LvTy/zP1QqFYMHD2b37t0EBgby6quvMn78eJLy2Cc4KCiITZs2ERISwltvvcWaNWvytP+cZOzYscTGxrJ48eL7nhcUFMSrr77KihUrcrXOSW6kuHW7TuU30dHRLF26lL59+2YoMkRRZPTo0VSsWJEBAwbkwwjvz7Fjx+jUqRMlSpRg8+bNssjIJ0RJIslsJjo6mvj4eBAET/yFXq9HEAQsVisxsbHExMZit9sxGAwULVoUP1/fgi8yZGRyEFloyOQIfn5+dO3alTVr1nD69GlmzZqFIAiMHTuWOnXq8MYbb7Bu3brkL2WZQk+lSpXYvHkzkydPZu3atbRq1YpDhw7l6Rh0Oh3Lli2jR48ejBw5kk8//bRQ1tooW7Ys4eHhzJ8/n7t379733L59+3Lz5k127NiRa+MpWbIkSqUyx+I0JEni9u3bBUJofPXVVyiVykzTv65bt46TJ08yY8aMHHMbyyl++eUXunXrRu3atdm0aRNBQUH5PaQnjtTxF6bERNRqdXL9i4AAvLRaRFHElJTE3bt3SUhIQEglQLz1eo+VQ0bmSUIWGjI5TkBAAN27d2fDhg388ccfTJs2DafTyYgRI3j66afp0aMHGzduxGg05vdQZR4BpVJJ//792bNnD6VLl6ZLly6MGjWKxMTEPB3D9OnT+fDDD5k7dy5Dhw7Nt7ofj8L777+PSqVizpw59z2vVq1aNGzYMFeDwtVqNSVLlswxoWE0GrFYLPme2vb27dusXLmSgQMH4u/vn+54TEwMU6dOpUuXLjRq1CjvB3gffvjhB/r06UOLFi1YvXo1Pj4++T2kJwqXy4UxRWBYzGZ0ej1FihbF388PlVqN3eEgPiGB6Lt3SUpKwsvLiyJBQR4BIssLmScZWWjI5CpFihShV69ebNq0id9//52PPvrIk+2ldu3a9O7dm++//z5PJ6cyOUu5cuXYuHEjM2bM4Mcff6RFixbs3bs3z/oXBIF33nmHr776ih9//JGePXsWus+Tv78/H3zwAWvWrOHvv/++77l9+/blt99+4+zZs7k2nrJly+ZYGmt3DY38tmh8/vnn6PV6+vfvn+Hxjz/+GIAJEybk5bAeyPLly3n33Xd57bXXWLx4MdqUzEUyuYvL5eKnn35iyJAhmM1mnE4nPj4+FClaFB+DAYVCgcVqJTY2ltjYWBwOBwYfH4oWLYqvjw8q2T1KRgaQhYZMHhIcHEyfPn348ccfOXnyJBMmTCA2Npb33nuP2rVrEx4eTkRERJ77+8s8OgqFgl69erF3716qVKnCm2++ydChQ0lISMizMXTq1Ik1a9bw+++/06lTJ+7cuZNnfecEffr0oVSpUkybNu2+57Vt25YSJUqwdOnSXBtLTqa4vX37NpC/QuO///5j7dq1vPvuuxlaA3777Tc2btzI+PHjC4xLkiRJzJkzh/Hjx/P222/z2WefyZPXPMBkMrF06VKaNGlCv379UCgUeOl0BPj7o9frkSSJRJPJ4x6lUCgICAiQ3aNkZDJBFhoy+ULx4sXp168fW7Zs4fjx44wZM4Y7d+4waNAgateuzYABA9i6dSsWiyW/hyqTDUqXLs3atWv57LPP2L59O82bN8/VeIJ7adq0KZs3byYmJoYOHTo80DpQkNBoNIwbN45du3bx66+/ZnqeSqXyCPaYmJhcGUtOFu27desWgiAQHBycI+09DJ999hkBAQH06dMn3TGHw8GYMWOoV69egamqLYoi48ePZ86cOYwbN44JEyYUmBokjys3btxgypQp1KtXz/O4fft2Pv/8c1QqFU6n0+MeZTabk92jihQhwN8frUYju0fJyGRC4RYacnXnx4JSpUrx9ttvs23bNn777TdGjBjBtWvXGDBgALVr12bQoEFs374dm82W30OVyQKCINCtWzf27dtHnTp16NOnD4MGDcqz7GPVq1dn27Zt+Pr6EhYWxm+//ZYn/eYEoaGh1K1blylTpiCKYqbnvfnmmwiCkGvZtsqWLUt8fHyOxFHdunWLokWL5ltw9T///MN3333H+++/j06nS3d80aJFXL58mZkzZ6JQ5P9PosPh4N1332XlypXMmjWLd999VxYZucipU6d4++23adSoEWvXrqVXr14cPXqU+fPnU6VKFbZt24Y5KYmEhIRk9ylf3/+5RymV+T18GZkCT+Gso3HlCjRokDz5ftJQq+H48UcqvFZYuHLlClu2bCEiIoLz589jMBho06YNYWFhNG/eHE1KZVSZgoskSfz444+MHz8epVLJtGnTCA0NzZOJk9FopF+/fhw9epQvv/ySsLCwXO8zJzh27BivvPIKX331FZ06dcr0vBEjRrBnzx6OHj2a45N4dyHBnTt3UqtWrUdqa+TIkZw5c4aff/45h0aXPd5++21OnDjB4cOH031nXL9+nebNm9O7d28mTZqUL+NLjcVioX///hw8eJD58+fToUOH/B7SY4nL5eLnn39m0aJFnDhxgnLlytG/f3+6dOmCt7c3//33H6tWrWLt2rWUjIlhS3w8Kr0etV7/5Fgu5DoaMjlE4RQaIFd3flgK6ev233//sW/fPvbt28fVq1cxGAw0adKEFi1aULdu3QdPtB71dZN5JO7evcu4cePYtm0bL7/8MtOnT88TVxqHw8HQoUP54YcfmDRpEgMGDCgUq8N9+/blzz//5NChQ5kG/54/f55WrVqxcOHCHBdRMTEx1K5dm6VLl/Lyyy8/Uls9e/ZEqVSycuXKnBlcNjh37hwvvvgis2fPpnv37mmOSZLEW2+9xZkzZzhw4AAGgyHPx5cao9FIz549OXv2LCtWrCAkJCRfx/M4kpiYyPr161m6dCnXr1+nUaNGDBw4kBdffBGFQsGhQ4dYtmwZu3btwsfHh+7du9OvYUNKdukiFwiWkXlICq/QkMk+j4klyOVy4XQ6cTqdye4lgoBapUKlUqFUKjOeSD5BlqCCzNatWxk7dixOp5OPP/6Y1157Ldcn/pIkMWPGDL788kv69evHpEmTUBZwl4d///2XFi1aMGbMmExrPgC8/vrrOBwOIiIicrR/SZKoUqUKw4YNY9CgQY/UVqtWrWjYsCHTp0/PodFlnbfeeou///6b/fv3p1uM+PnnnwkPD88RMfWo3Llzh+7du3Pr1i3WrFnDs88+m6/jedz4v//7P5YtW8batWuxWCyEhYUxYMAAnn76aZKSkti0aRMrVqzg0qVLVKtWjfDwcDp16oRer79vZfDHGlloyOQQcgqLJwmTKfnLQ6GAQpy9RJmyaQCny4XdZsNstyNaLAgKBVqNBo1Gg1qtTjZzu2N5CqEl53GjQ4cOPP/880ycOJEhQ4YQGRnJzJkzczUjkbtwZMmSJfnwww+5desWX375JV5eXrnW56NSoUIFevbsyRdffEG3bt0IDAzM8Ly+fft6rB9PP/10jvUvCAJlypTJkRS3t2/fzpcaGqdOnWLnzp189dVX6URGUlIS48ePp1WrVrRr1y7Px5aaa9eu0a1bN2w2G5s3b6ZKlSr5Op7HiZMnT7JkyRK2bduGt7c3vXv3Jjw8nOLFi3PlyhUmTpzIhg0bSEpKol27dsyYMYNGjRoVCqunjExhIf8j32TyHpUqeWWmkG+CWo3aywtvPz8CihbFr0gRNHo9FqeT2MRE7sbHY7RacUChrBj9uBIYGMhXX33FypUrOXPmDC1atGDdunW5/h717t2b5cuX88svv9C1a1fi4uJytb9HZdiwYQDMnTs303Nat25N6dKlWb58eY73nxMpbm02G3FxcfmS2nbWrFlUqVKFjh07pjs2d+5cT4G+/JxUnj9/no4dO6JQKIiMjJRFRg7gdDrZunUroaGhhIaG8ueffzJlyhROnjzJmDFjOHfuHD169KBJkyZ8//339O7dm6NHj7JkyRIaN24siwwZmRxGFhoyjwUCoFap8DEYKFKkCEGBgeh0Omw2GwkJCSQlJfHVV19x/Pjx+2bzkck7XnrpJfbv38/LL7/M8OHD6datW44Vibtfn9999x2XL18mLCwsx1K45gZBQUG89957rFq1iitXrmR4jkql4q233uLHH38kOjo6R/vPCaGRX8X6jhw5woEDBxg5cmQ6N7nz58+zaNEihg4dStmyZfN0XKk5ceIEr776KsHBwURERFC6dOl8G8vjQGJiIosXL6ZJkyYMGDAAjUbDypUrOXToEK+99hrr1q0jJCSEHj16cPfuXebOncvJkycZN24cpUqVyu/hy8g8tshCQ+axQwDUarVHdPj5+aFSqdi/fz8dO3akQYMGfPTRR5w6dUq2dOQzvr6+zJ07l7Vr13L58mVeeOEFVqxYkatisG7dumzduhWXy+VZ8Syo9OvXj+Dg4PvGN3Tv3h2lUsnq1atztG93dfBHeS/cxfry0nVKkiRmzpxJrVq10sVeiKLImDFjKF++/CPHnjwKe/bsoUuXLtSoUYPvvvuOIkWK5NtYCjvXr1/no48+ol69ekydOpXnnnuOHTt28P3331O+fHnGjx/vqY3x9NNPExkZyc8//0zXrl0LtPukjMzjgiw0ZB5r3JYOLy8vNmzYwI8//kjbtm354Ycf6NChAw0bNmTq1Kn8+eefsujIR1q0aMHevXvp3LkzH374Ia+//jpXr17Ntf7KlSvHli1bKF26NJ06dWLPnj251tej4OXlxZgxY9i2bRvHjx/P8Bx/f39ef/11Vq1ahSMHEz2UKVMGh8PxSBXW3RaNvBQa+/fv59ixY4wePTqdG8zGjRs5fvw4n3zySb7V9di8eTNvvfUWzZo1Y+3atfj6+ubLOAo7J0+eZODAgTRu3Jj169fTp08fjh07xueff86tW7fo2rUrzZs3Z+vWrQwcOJDjx4+zYMEC6tevL7tHycjkIbLQkHliUCgUNGzYkGnTpnHq1Ck2bdpEy5YtWb9+PW3btqVJkyZ88sknnD17VhYd+YCPjw+ffPIJmzZt4tatW7Rs2ZLFixfjcrlypb+goCC+++47mjRpQu/evVm3bl2u9POodOrUiVq1ajFlypRMP5fh4eHcuXOHbdu25Vi/ZcqUAXgk97Jbt27h4+OTZ6lj3daM+vXr07JlyzTHYmNjmTJlCq+//jpNmjTJk/Hcy8qVK3nnnXd49dVXWbp0qbyink2cTidbtmzxxF+cOXOGqVOncvLkSQYPHsyPP/5IkyZNeOuttzCZTHz55ZecOHGCESNGUKxYsfwevozME4ksNGSeSJRKJU2aNGHGjBn88ccfbNiwgeeff55vv/2W1q1bExISwqxZs7hw4UJ+D/WJo0mTJvzyyy/06NGDjz76iFdeeYVLly7lSl86nY5ly5bRo0cPhg8fzuzZswucyFQoFEycOJGTJ0+ydevWDM+pVq0aTZs2ZdmyZTnW71NPPQXwSHEat2/fztP4jB07dnD69OkMrRlTp05FFEUmTpyYZ+NxI0kSn332GePGjaNfv37MnTsXVSHO/JfXuOMvnn/+eQYOHIhWq2XVqlUcOnSIhg0bMmnSJJ599llmzJjBc889x7Zt29i2bRuvvfaaXNhVRiafkYWGzBOPSqUiJCSE2bNnc/r0adasWUODBg1Yvnw5LVu2pEWLFsyZM4d//vknv4f6xKDX65kyZQqbN28mLi6OF198kS+//BKn05njfalUKqZPn864ceP47LPPGDZsWI66IOUETZs2pVWrVkyfPj3TsfXr14+TJ0/y+++/50ifOp2O4ODgR7Zo5JXblCiKzJo1i6ZNm6azWBw7doz169czbty4PI+HcIub2bNnM3r0aCZPnoxCIf/0ZoXr168zefJkT/xFo0aN2LlzJ+vXr8dut9O5c2datWrF7t27GTJkCCdOnGDevHlyHRIZmQKE/G0nI5MKtVrNCy+8wNy5c/nzzz9ZtWoVTz/9NIsWLaJZs2a0atWKL774ItMsQDI5S4MGDdi9ezf9+/dn5syZdOjQgfPnz+d4P4Ig8O677/Lll1/yww8/0Lt3b0wFrO7K+PHjuX79eqYVtlu1akWZMmVyNNVtmTJlHklo5KVFIzIykgsXLjB69Og0+x0OB2PGjKFu3br06NEjT8aSuu/333+f5cuXM2PGDN5//305PiALnDx5kgEDBtC4cWM2bNjgib+YPHkye/fupVGjRvTv3x+Xy8XChQs5duwYH3zwAUWLFs3vocvIyNyDLDRkZDJBo9HQunVr5s2bx5kzZ1i+fDlVq1blyy+/pEmTJrz00kvMnz//kVOAytwfLy8vxo8fz9atW7HZbLRt25Y5c+bkitXhtddeY82aNZw8eZJOnTo9UiB0TlO1alW6d+/O3LlzSUhISHdcqVTSp08fIiMjiYqKypE+y5Yt+0if75s3b+aJ0HA6nXz66ae8+OKL1KtXL82xJUuW8PfffzNjxow8tSRYrVb69etHZGQkCxYsoFevXnnWd2HE6XQSGRnpib84e/Ys06ZN4+TJk3To0IGZM2dSt25d5syZQ0hICDt37mTz5s2EhYXlW2C/jIzMg5GFhoxMFtBqtbRt25avv/6aM2fOsGTJEsqXL8+cOXNo3LgxL7/8MgsWLODGjRv5PdTHlmeeeYYdO3bw7rvv8sUXX9C2bdtcSU3btGlTNm/eTHR0NKGhobkWH/IwjBgxArvdzrx58zI8/sYbb6BSqfj2229zpL9HsWi4XC6ioqLyxHXq+++/58qVK4waNSrN/hs3bjBnzhzCw8OpVatWro/DjdFo5I033uDQoUOsWrWKsLCwPOu7sGE0Glm0aBGNGzfm7bffxsvLi1WrVrFnzx78/f3p1q0bbdq04dChQ4wcOZJTp04xd+7cPH0/ZWRkHh5ZaMjIZBOdTkf79u1ZtGgRf/31FwsXLqRkyZLMmjWLBg0aEBoaypIlSzypPWVyDo1Gw8iRI9m+fTsKhYL27dvzySefYLPZcrSf6tWrs23bNgwGA2FhYRw9ejRH239YgoODGTx4MEuXLs2wuKGvry9dunThm2++yRGLT5kyZbhz5w5WqzXb10ZHR+NyuXLdomG325kzZw4dOnRIN/kcP348vr6+jBw5MlfHkJq7d+/SqVMnLly4wMaNG2nRokWe9V2Y+O+//5g0aRL169dn+vTpPP/88+zatYv58+dz5swZGjVqxKBBg9BqtSxbtowjR47wzjvvEBgYmN9Dl5GRyQay0JCReQT0ej1hYWEsXbqUM2fOMH/+fIKCgpg6dSr16tXjlVdeYfny5QXKBedxoGbNmvz000+MGDGChQsX8tJLL3Hy5Mkc7aNEiRJs3ryZWrVq0bVrV7Zs2ZKj7T8sb7/9NoGBgXzyyScZHu/Tpw93797NkfG6U9w+TMX2vKoKvnbtWm7cuMGIESPS7N+1axc7duxgypQp+Pj45OoY3Fy/fp2OHTsSExPDjz/+mM6N60lHkiROnDhB//79ef7559m0aRPh4eEcO3aM3r17s2DBAho0aMD8+fN56aWX2LNnD5s2baJdu3Zyli4ZmUKKIBW0XI4yuceZM9C0KWg08CT5tDocYLfDoUNQu3aedGk0GtmxYweRkZHs378fl8tFo0aNCAsLo3379nIl4Bzk4sWLDBs2jD/++IMBAwYwatQodDpdjrVvt9sZOnQomzdvZtKkSQwYMCDH2n5Y1q1bx/Dhw/npp5945pln0h3v1q0biYmJj1xX4+bNm9SvX59vv/2WVq1aZevan3/+mfDwcP78889c+7xbrVYaN25MSEhIGncys9lMixYtqFSpEmvWrMmTAOwLFy7wxhtvoNPpWL9+vUekySTHX/z0008sXryYU6dOUaFCBQYMGEDHjh3ZtWsXy5cv548//qBs2bK89dZbdOvWDT8/v/wedjLy72ae/W5mmStXoIAl68gTDAYoXz6/R5Ft5CUCGZlcwNfXl86dO9O5c2fi4+P5+eefiYyMZPz48Xz44Yc8//zzdOzYkXbt2smuAI9I1apViYiIYMmSJcyaNYsdO3bw2Wef0ahRoxxpX6PR8OWXX1KqVCkmT57MjRs3mDRpUr6mKO3SpQtLlixhypQpfP/99+km0n379qV3796cPHnykVbVixUrhlqtfqg4jVu3bqFWq3P1871y5Uqio6MZNmxYmv1z584lKiqKjRs35onIOHnyJD169KBUqVKsW7dOzn6UgtFoZO3atSxbtowbN27QtGlTVq1aRc2aNVm9ejUhISFER0fTvHlzVq1aRcuWLVEqlfk97IzJhdTaBZqC+nyvXIEGDZKF0JOGWg3Hjxc6sSELDRmZXMYd0NitWzdiY2P56aefiIyMZPTo0YwZM4aQkBDCwsJo165dwVnFK2SoVCoGDRpEmzZtGDp0KJ06daJPnz6MGzcOb2/vR25foVAwbtw4SpYsyfjx47l16xZffvklWq02B0affZRKJRMmTODNN99k586dtGnTJs3xVq1aUa5cOZYtW/ZIQkOpVFK6dOmHFhrFixfPNUFmMpn46quv6NatG+XKlfPsv3jxIosWLWLo0KFp9ucW+/bto2/fvtSuXZtvvvkGX1/fXO+zoPPff/+xdOlS1q1bh91u55VXXqF///6YzWaWL19O37590Wq1dOnShT59+lCpUqX8HnLmGAzJEzz3Cv+ThFqd/PwLEiZT8nuhUMCT5E7ndCY/70JoyZFdp54kZBNwgTIB37171yM6fvvtN1QqFc2bNyc0NJQ2bdrIE5aHRBRFVqxYwfTp0wkKCvKkw8wpduzYwaBBg3j66adZuXIl/v7+OdZ2dpAkiW7dunHz5k327NmTLsXnkiVL+Pjjjzl+/DjFihV76H66d++Ol5dXtutzDBkyhP/++4/Nmzc/dN/34/PPP2fu3LkcOXKEkiVLAsmvyWuvvUZUVBR79uzJ9arQkZGRvPfeezRv3pxFixblqMteYcMdf7F48WK2b9+Or68vvXv3plu3bhw5coRly5Zx9uxZKlSoQHh4OJ07d86z2JlHRnbVKTjI85gCNY/JCrLQeJKQb9ACe4PeuXOHbdu2ERERwfHjxz2FAzt27Ejr1q0xFLRVpULAtWvXGDFiBL/++ivdu3dn4sSJOSbeTp48Se/evQkMDGTNmjU89dRTOdJudjl79iwvvfQS06ZN46233kpzzGg0Uq9ePQYMGPBIWZfGjBnD8ePH+eWXX7J1XefOnQkKCmLhwoUP3XdmJCQk0LBhQzp37szHH3/s2b9x40Y++OADNm7cSNOmTXO839R88803jB07lk6dOvHZZ589sbUc3PEXixYt4vfff6dixYppiu2tXbuW+Ph4WrVqRXh4OM2aNZMro8s8PPI8psDOYzJDvttlZAoAxYoVIzw8nIiICE6ePMmHH35ITEwM77zzDrVr1/YU/jKbzfk91EJD2bJl2bhxI7NmzSIyMpIWLVqwe/fuHGm7Xr16bNmyBafTSWhoKGfOnMmRdrNLzZo16dy5M3PmzCExMTHNsdSpbu2P4PJRpkwZrl+/TnbXpG7dupVrGacWLlyIw+Hgvffe8+yLi4tjypQpvPrqq7kqMiRJ4osvvmDMmDGEh4fzxRdfPJEiw2g0smDBAho1asTbb7+Nt7c3q1atYvr06ezbt48WLVrw7bff0rlzZ3799Ve++eYbWrRoIYsMGZknDPmOl5EpYJQoUYIBAwawZcsWjh07xqhRo7hx4wZvv/02tWrVYuDAgWzbtu2hahs8aQiCQI8ePdi7dy/VqlWjV69eDBkyhPj4+Eduu3z58kRGRlKyZEleffVV9u7d++gDfghGjx6NyWRi/vz56Y6Fh4cTExNDRETEQ7dftmxZTCYTcXFxWb5GkqRcExrR0dEsWbKE8PBwgoODPfunT5+O0+lk0qRJOd6nG1EU+eijj5g5cyYjR45kypQpT9zE+dq1a0yYMIF69eoxY8YMmjZtypYtWwgNDeWTTz6ha9euXL58menTp3Py5EkmT56cJ7EyMjIyBZMn6xtSRqaQUbp0aQYNGsT27ds5cuQIw4YN48qVK/Tv359atWoxePBgduzYkeMF6x43SpUqxZo1a5g7dy47d+6kefPmbN++/ZHbLVKkCN999x1NmjShV69erFu3LgdGmz1KlCjBwIEDWbRoUboikRUrVuSFF15g2bJl2bZIuHGnac1OQHhiYiIWiyVXqoLPnz8fpVLJ4MGDPftOnDjBmjVrGD16dBrxkZM4nU6GDRvGkiVLmD59OkOHDs2TjFYFAUmSOHbsGP369aNJkyb88MMP9O/fn++//56AgADefPNNxo4dS7ly5di4cSN79+6lV69eOZKIQUZGpnAjCw0ZmUJC2bJleffdd9m5cyeHDh3i3Xff5eLFi/Tp04enn36aIUOGsGvXrhypCP04IggCXbt2Zf/+/Tz77LP07duXt99+m5iYmEdqV6/Xs2zZMrp3787w4cOZM2fOQ0/qH5Z33nkHg8HAzJkz0x3r27cvf/7550MXNCxbtiyQPaHhFjzuIO2c4vbt26xYsYIBAwYQEBAAJAuA0aNHU6dOHXr16pWj/bmxWq3069ePH374gfnz56eLh3lccTgcRERE0L59e1555RX+/vtvpk+fzty5czl79iwdO3Zkw4YN9OjRgyNHjrB8+XKaNm36xAgwGRmZByMLDRmZQkiFChX44IMP+OWXX9i/fz8DBgzg9OnT9O7dm9q1azN06FD27t0ri44MKFasGCtWrODrr7/m4MGDNG/enIiIiEcSByqVihkzZjB27FjmzJnDsGHD8vS19/HxYfjw4WzatImzZ8+mOdaiRQvKly/P0qVLH6ptX19ffH19uXbtWpavcQuNnLZofPHFF+j1+jRFE5ctW8bFixeZOXNmrtRgMBqNvPnmmxw4cIBVq1bxyiuv5HgfBQ2j0cjXX39No0aNGDRoEAaDgcWLF9O7d2+WLFnCW2+9xY0bN5g9ezYnT55k/Pjx+ZYQQUZGpmAjCw2ZdHyQkECfVD7sr8XFMfGeQNPHgTlz5vDiiy/m9zAemcqVKzN8+HD27dvHnj17CA8P59ixY7z55pvUqVOHESNGcPDgQZwFtQBTPiAIAq+88gr79++nSZMmDBo0iPDwcO7cufNIbb733nvMmzeP77//nt69e2PKw5SYb775JhUqVGDKlClpRJNCoaBv375s27YtnWtVVilbtmy2LBq3b98GeKS0uvfy33//sXbtWt59911PWtSbN2/y6aef0rt3b55++ukc68tNdHQ0r7/+OmfPnmXjxo288MILOd5HQeLq1atMmDCBunXrMnPmTJo1a8aKFSuoUqUKw4YNY/LkydSoUYMff/yRXbt2eSqhy8jIyGSGLDRkHsgyPz9GPYa+tm+//TabNm3K72HkGIIgUK1aNUaNGsWhQ4fYtWsXPXv25Ndff6Vr1648++yzjBkzhl9//RWXy5Xfwy0QFClShEWLFrFkyRJOnjxJ8+bN2bhx4yNZN15//XXWrFnDiRMn6NSp0yOJl+ygVqsZP348Bw8eZN++fWmOde7cGZ1OxzfffPNQbZctWzZbFo2bN29StGjRHM3GNHfuXPz9/dO4LU2cOBGDwcDo0aNzrB83169fp2PHjkRFRfHjjz9Sv379HO+jICBJEkePHqVv375p4i9mz57NnTt36NOnDxEREfTt25djx46xaNEiGjZsKLtHycjIZAlZaMg8EH+FAsNjmFnF29vb4+f9uCEIAjVr1mTs2LEcPnyY7du307VrV/bs2UPnzp2pW7cuH374IUePHkUUxfwebr7Tvn179u/fz0svvcQHH3xAz549uXnz5kO3FxISwubNm7l79y6hoaFcunQpB0ebOS+99BKNGjXi448/TiMmfXx86Nq1K99+++1DJQ5wp7jNKrdv385Rt6nLly+zadMmhgwZgl6vB+CXX37hp59+4qOPPsrx4pZ///03HTt2RBRFIiIiqF69eo62XxBwOBxs3ryZl19+mVdffZVLly4xefJkBg8eTEREBO+//z5xcXF88cUXnDhxgtGjR+daumIZGZnHl8dv9vgYYDKZeOedd6hYsSLPPPMMixcv5rXXXmPixIkAfPfdd7Rt25bKlStTp04dBg8eTHR0tOf6hIQE3nnnHWrVqkX58uV5/vnn2bBhg+f4eaeTznFxlI+Kosbdu4w0Gkm6z2TzXtep56KjmZeUxFCjkcpRUdSPjmb1PfUdbrpcDExIoFpUFNWjongrPp7r2VhFX2ux0CwmhnJRUYRER7MyVfvXXS5K3rnDT1Yrr8fFUSEqilYxMZy8xyd+jcVCvbt3qRAbS7jFwqJNm6hWrZrn+L2uUx988AF9+vRhwYIFPPPMM9SoUYNx48al8bW32+1MmTKFZ599looVK9K+fXsOHz6c5eeVHwiCQJ06dRg/fjxHjx5l69atvPrqq2zfvp1XX32VevXqMXHiRE6ePPlEi46AgADmzZvHqlWrOHv2LC1atGD16tUPbd2oUaMG27Ztw9vbm7CwMI4dO5bDI06PIAhMnDiRCxcusHHjxjTHwsPDiY2NfahK3WXKlOH//u//sux+d/v27RydlM6ePZvixYvTs2dPACwWCx9++CEhISGEhYXlWD8Ap06d4pVXXiEgIIDNmzd7guEfF1LHXwwePBhfX18++eQTGjduzIwZM5g1axbPPvssW7Zs4aeffqJz585otdr8HraMjEwhRRYaBZDJkydz/PhxVq1axfr16zl69GiagmBOp5NRo0axe/duli9fzvXr1/nggw88x2fNmsXff//NmjVrOHDgADNnziQwMBAAsyTRPTERP0Fge2Agi/38OGi382E2YzAWms3UUanYFRjIWzodYxITuZwyCXFIEm/Ex+MtCPwYEEBkYCDegkD3uDgcWZi0/WCx8KnJxBhvbw4EBTHWYGBWUhIbLZY0581ISuJtvZ7dgYFUVCoZlJCAM6X943Y7o41G+un17Pbzo5lSyRdr1jyw78OHD3Pt2jW+++47vvjiCzZs2JBmwjZu3DhOnDjBwoUL+eWXX+jQoQPdu3fnypUr2Xn58g1BEKhbty6TJ0/mxIkTnowykZGRhIaG8txzzzFlyhR+//33PM+cVFBo3bo1+/fvJzQ0lFGjRtG1a9dsxSekpmTJkkRERFCzZk26dOnC1q1bc3i06XnmmWd45ZVXmDVrVpoCj+XLl6dVq1YsXbo02+9tmTJlcLlcWbby3Lx5M8eExvnz54mIiGDo0KFoNBogOSj89u3bzJgxI0ddeA4cOECXLl2oXLkyP/zwQ47GmOQ3V69eZfz48Z74i6ZNmzJ58mQkSWLs2LHs2LGDwYMHc+zYMebPn0+9evVk9ygZGZlHRhYaBQyTycSmTZuYOHEiTZs2pVq1asydOzeNG0S3bt1o2bIlZcuWpV69ekydOpU9e/aQlJQEwP/93/9Rq1Yt6tSpw1NPPUVISAitW7cG4EenEyswz8+PaioVTTUapvn48J3Vyt1sWBxaaTS8pddTTqXiHb2eQIWCX1OqD0darYiSxBwfH6qr1VRWqfjc15cbosjhLFQonp2UxCQfH1728qKMUsnLXl4M0Ov59h6hMUiv50WtlgoqFSMMBv7P5eJqynNYZrHQUqtlkLc3FZRK3tJoaPnccw/s28/Pj2nTplGpUiVat27Niy++yMGDBwG4ceMGGzZsYMmSJTRs2JBy5coxaNAgnnvuOdavX5/l166goFAoaNCgAVOnTuXkyZN8//33vPTSS3z33Xe0b9+exo0bM23aNM6cOfPEiQ5fX1/mzJnD+vXruXLlCi+88ALLly9/KIuPr68va9eupV27dgwcOJAlS5bkwojTMnbsWGJjY1m4cGGa/f369ePs2bPZtq5kN8VtTrpOzZo1i3LlytGlSxcALl26xIIFC3jvvfcoX758jvQBsHXrVnr27EmjRo1Yv349fn5+OdZ2fuGOvwgPD6dJkyZs3ryZXr16MXjwYI4cOcLkyZOxWCx8/fXXHD9+nOHDhz9W4kpGRib/UeX3AGTS8t9//+FwOHj22Wc9+3x9falYsaLn/z///JM5c+Zw9uxZEhISPJOfGzduUKVKFXr37k2/fv04c+YMLVq0oG3btp5AxksuFzWVSvSpVqoaqNWIwGWXi6JZTA9ZXfW/j44gCBRVKIhOGcc5p5OrLheV795Nc41NkvjX5aL5fdo1SxJXXS6GGY2MMBo9+52A7z2ra6nHUCwlhiRaFKkEXHY6aXePuf/ZatXY/fvv931eVatWTZMiMzg4mAsXLgDJK6sul4smTZqkucZut3ssRoUVpVJJ48aNady4MVOmTOG3334jMjKStWvXMn/+fMqVK0dYWBgdO3akWrVqT8xKZ7Nmzdi7dy/Tp09n/PjxREZGMnfu3GxPcDUaDfPnz6dUqVJMmjSJGzduMHHixFyrKv3UU0/Rr18/vv76a3r06OEpYtesWTMqVarEsmXLaNiwYZbbK1WqFIIgZElo2Gw2YmNjc8Si8fvvv7Njxw6+/PJL1Gq1Z/W9VKlSvPvuu4/cvps1a9YwatQoXnnlFT7//PMcDWLPDxwOB1u3bmXx4sWcPn2aypUr8/7773Pjxg1WrFiBKIq8+uqrhIeH50q2LhkZGRk3stAoZJjNZt544w1atGjB/PnzCQoK4saNG7zxxhueWIKWLVty/PhxfvnlFw4cOEDnzp3p06cPE199NcfGob5noikA7jXvJEniabWa+fcEaCYYjeiNRu6azahUKlQqFWqVCpVajVKpRABPrMhsHx/q3vNjr7inz9RH3UceNcJApUp7SwiC4BFyZrMZpVLJjh070uXrdweoPg6oVCqaNm1K06ZNmTZtGocPHyYiIoJVq1Yxb948KlasSMeOHQkNDaVq1ar5Pdxcx2AwMH36dEJDQxk+fDgtW7Zk9OjR9O/fP1t1GxQKBePHj6dUqVKMHz+eW7duMW/evFzzfx8yZAjr1q1j9uzZzJo1C0j+PIeHhzNhwgRu3LhBqVKlstSWRqOhRIkSWRIa7tS2OSE0Zs2aRZUqVTy1K77//nsOHz7MunXrcuR1kySJ+fPnM336dPr06cPHH3+ca+IvL0hISGDNmjUsXbqU27dv07RpU9555x2OHz/O559/TokSJRg6dChvvvkmQUFB+T1cGRmZJ4DC+436mFKmTBnUajV//PGHZ5/RaOTff/8F4J9//iEuLo4PP/yQhg0bUqlSpTSB4G6CgoLo0qULX331FVOmTGH16tUAVFYqOetyYU7lCnPc4UABVMyhYle11Wr+dToJUigop1J5tpp+fjzl74+XlxcAVouF+IQEoqOjiYqKIiY2Fk1SEkWByw4HZZTKNNeXycb4KqpUnL4ncPWPixcf6XnVqlULl8tFdHQ05cqVS7O5V4wfN9RqNc2bN+ezzz7jzz//5Ntvv6VevXosXbqUF154gRdeeIHPPvuMy5cv5/dQc53GjRvzyy+/0Lt3bz7++GPCwsL4+++/s91Onz59WLp0KTt37qRr167Ep6pZk5P4+fkxdOhQ1q5dy8VUn/3OnTuj1+tZtWpVttrLaorbnBIav/32G/v372fkyJEolUoSEhL46KOPCAsLo3nz+9lFs4YkSUydOpXp06czfPhwpk6dWmhFxtWrV/nwww+pV68eM2fOpGHDhrz11lv8888/zJ8/H4VCweLFizl69ChDhgyRRYaMjEyeUTi/VR9jDAYDnTt35uOPP+bXX3/l4sWLDB8+HIVCgSAIlCpVCrVazfLly7l27Ro7d+5k7ty5adr49NNP2bFjB1evXuXixYvs3r2bypUrA9BJpcILeD8hgQtOJ7/a7YxPTOR1L68su009iE5eXgQqFPRJSOCo3c5/LheH7XYmm83EqdX4GAwE+PtTtGhRgosWJSAgAIPBgEqlwuFw8LYk8VVSEp/fucPxu3c5GhvLqvh4vkpIwOlykZVogb46Hb/YbCxKSuKKy8W3djt7HjHrT4UKFejUqRNDhgzhp59+4r///uP333/nyy+/ZPfu3Y/UdmFArVbTqlUrPv/8c/78809WrVpFzZo1WbBgASEhIbz44ovMmzePq1ev5vdQcw2dTsfkyZOJjIwkMTGR1q1bM2/evGxXAW/Xrh3fffcdly5dIiwsLFupY7PDW2+9RZkyZZg2bZpnn7e3N2+88QarV6/GarVmua0yZcpkyaKRE1XBJUli5syZ1KpVi3bt2gHwySefYLPZ+Oijjx66XTdOp5Nhw4axYMECpk6dyvDhwwudO6AkSfz222/06dOHJk2aEBERQVhYGG3atOGnn35i/fr1tGzZkt27d/PDDz/QoUOHdBZbGRkZmdxGFhoFkMmTJ1OvXj169epF165dadCgAZUrV0ar1RIUFMQXX3zBli1baN68OV9++SWTJk1Kc71arWb69Om0bNmSV199FYVCwYIFCwDQCQJrfXyIlyTaxcbSPyHBExCeU+hSsk2VUijom5BAs5gYhhuN2CQJwz0/5gqFAq1Gg7dej5+vL0WCgni7eHFm+/qyTankNZeLnnY7G61WAiwWoqOjiY6ORhRFTElJmC0W7A4H4j3Byg00Gmb6+rLIbKZVQgJ7XS4GvPaax5rysMydO5fOnTvz0UcfERISQnh4OH/88UeWXVAeFzQaDa1bt+arr77izJkzLFu2jEqVKvHFF1/w/PPP06ZNG77++uuHztZU0KlXrx67du1i4MCBfPrpp7Rv355z585lu40tW7Zgt9sJDQ3lr7/+yvFxqtVqxo4dy+7duzl06JBn/1tvvUVCQgI//vhjltvKjtDw8fHBYDA81JghOfvT0aNHGTVqFAqFglOnTvHtt98yZsyYRw5WttlsDBgwgO+//56vvvqK8PDwR2ovr3E4HPzwww+0a9eOTp06cfnyZbp06ULZsmVZv349f/zxB6NHj+bUqVPMnj2bGjVq5PeQZWRknmAE6UlLJ1MIMZvN1K1bl0mTJvHGG288fENnzkDTpqDRQCEMdnSJIk6nE6fTicPhwOl04nI6PRYOlVKJSq1OE/uhUCgQHA6w2xnxyiv8k5DwUHUEZLKGxWJh9+7dbNmyhd27d2O1Wnn22WcJCwsjNDSUkiVL5vcQc5zTp08zbNgwLl26xJAhQ3j//fezFUx89+5devXqxT///MOSJUto0aJFjo5PkiTCwsKw2Wz8/PPPHveg3r1783//93/s3r07S6v533//Pe+99x6XLl3C29s70/MmTpzI/v372b9//0OPt3379iiVSiIjI3G5XLRr1w6FQsFPP/2UrbiYe0lMTKRPnz6cPHmSJUuWpKmjU9BJSEhg9erVLFu2jNu3b9OwYUNKlCjB4cOHiYqKomnTpvTt25cXX3zxkV4jGZkCTSGfxzw0KfMYDh2C2rXzezTZQrZoFED++usvNm/ezNWrVzlz5gzvvPMOAG3atMnnkeUvylTWD38/P4oEBRFcrBhBQUH4+fmh0WoRRRGz2UxcfDyzo6I4dOcOfyYksCgpiW8iI2nYsGG23EVksodOpyM0NJTFixdz5swZFixYQLFixfjkk0+oX78+YWFhLF26lDt37uT3UHOMOnXq8PPPP/P+++8zb9482rRpkybG6kEULVqU77//nsaNG9OzZ880xTVzAkEQmDRpEn/99Rc//PCDZ3+/fv04f/48R44cyVI7WU1x+6ipbXfu3OlZlRcEgZUrV3Lu3Dlmzpz5SBPomJgYOnfuzF9//cWGDRsKjci4cuUKH374oaf+RfXq1WnevDknT55k586dtGvXjn379rFx40batGkjiwwZGZkChWzRKID89ddfDB8+nMuXL6NWq3n66aeZPHky1atXf7SGC8hKQKWoqEyPrfH3p2FKUa6HRQJEUWRgfDxHHA5MkkQZSSLGx4d4gwGFQkHFihWpUaMGNWvW9DwGBwcXOj/twkJiYiI7d+4kMjKSffv24XQ6ee655wgLC6N9+/aPTTD9uXPn+OCDDzh37hyDBg1ixIgRWc6O5HQ6GTt2LGvWrGHEiBEMHTo0Rz+P/fv35/fff+fQoUN4eXkhSRItWrTwpLt9EFFRUTzzzDOsWLHivoseoaGhVKpUKV3sWFYQRZHWrVsTGBjIpk2buH37NiEhIXTu3Jnp06dnuz03N27coGvXriQmJrJu3boC707kjr9YtGgRu3btwt/fn7p163Ljxg0uXLhAuXLl6NOnD127dsX3nux+MjKPNQVkHpPnFGKLhiw0niQKyA169Z5sUKkprlTildOT/ZQb1Lp7N+dVKs6dO8e5c+c4e/Ys58+fJzGlKnpgYGA68VG5cuVCn1O/oGE0Gtm+fTtbtmzhwIEDiKJI48aNCQ0NpX379oU+I47D4WDhwoXMnj2bMmXKMHfuXE8dmwchSRLz5s1j5syZvPHGG8yYMSPHPn9XrlyhefPmjBo1ylOD4ptvvmHcuHEcOXKEp5566oFjq1ixImPGjGHAgAGZnle/fn06d+7M6NGjsz3GiIgIBg0aRGRkJPXr12fgwIH89ttvHDx48KEn1JcuXaJbt26oVCo2bNhAuXLlHqqdvMDhcBAZGemxCJYtW5Zy5crx559/EhcXxwsvvEB4eDgvvPBCoc2QJSPzSBSQeUyeIwsNmUKBfIOmu0ElSeL69ese4eEWIe40nmq1mkqVKlGjRo00IqRIkSL58UweO+Li4vj555+JiIjg119/BaBJkyaEhYXRrl07AgIC8nmED8+lS5cYNmwYp06dol+/fowePTrL9VY2btzIiBEjaNasGYsWLbpvTER2mDBhAhs2bODIkSMEBQV54r969OjB+PHjH3h9ixYtaNq0KVOnTs3wuMvlonz58kydOpVevXpla2xOp5MWLVpQvnx5vv32W/bu3cubb77JV199RadOnbLVlps//viDN998k2LFirFu3boCW/U6Pj4+TfxF7dq10Wg0nD59Gp1OR9euXenTpw8VKlTI76HKyOQvj+k8Zo7JxCqLhWhRZLmfH23vTVwjCw2ZQsFjeoM+kIe4QRMTE7l48WIa8XH+/HnMZjOQXDH8XvFRsWJFOX3kIxATE8NPP/1EZGQkR44cQaFQEBIS4hEdhdFFxOVysXTpUmbMmEHx4sWZM2cOzz//fJauPXDgAP369fNMvHPCvSw2NpbGjRvz+uuve1Lefvzxx6xZs4ZTp049UAi5xcM333yT4fE7d+7w7LPPsmrVKlq3bp2tsW3YsIGhQ4eyY8cOKleuzAsvvMBTTz3Fhg0bHsqF7NChQ/Tp04dq1aqxevVq/Pz8st1GbnPlyhWWLFnChg0bcDgc1KpVi7i4OK5du0alSpXo06cPnTt3fqQMXjIyjxWP4TzmktNJ85gYlvv5UVetxl+hQHPvd54sNGQKBY/hDZolcugGFUWRa9euecSH+/HGjRtAcsrXqlWresRH9erVqVmzJv7+/jn0RJ4coqKi2LZtG1u2bOHo0aOoVCpatGhBWFgYL730Ej45mI45L7hy5QrDhg3j6NGj9O7dmw8//DBLk8ezZ8/So0cPNBoNa9eupWLFio88lq+//poZM2awb98+KlSowPXr12ncuDEzZsygR48e9712/PjxHDp0iH379mV4/PTp07Rr144dO3ZQOxv3msPhoGnTpjz99NMsWbKETz/9lK+++oo9e/Y81HP+6aefGDRoEE2bNmXJkiVZtiTlBZIkceTIERYvXsyuXbswGAxUqFCBy5cvk5SUxIsvvkjfvn0JCQmRY8ZkZO6lAM1jHJKEOgfu0V02G73j47lxvzhRWWjIFAoK0A2ap+TyDZqQkJAm7uPcuXNcuHABu90OJFdIdls93CKkXLlycnaYLHL79m22bdtGREQEJ06cQKPR0LJlS8LCwmjdunWOuRXlNqIosmrVKqZNm0ZAQACzZ8/OUoXrGzdu8OabbxIVFcWqVato0KDBI43DZrN5JvXuIPDw8HCuXLnCnj177ju5XbJkCdOnT+fff//N8LwdO3bQp08f/vzzz2y5F65atYpx48axZ88e1Go1L7zwAu+88w6jRo3K9vNbt24dI0eOJDQ0lHnz5hWYGCt3/MWiRYs4c+YMxYsXx8/Pj0uXLuHj40P37t3p3bu3J7uXjIxMBtxnHrPVauWzlCK9OkGglkrFSn9/esbHU1OlYkqqBao+8fH4CQKfp1g6n4uOpodOxxWXi61WK34KBR/o9fRIWaS47nLRMDqahX5+rDSbOeV0MtPHh85eXnyelMRqi4UYSaKyUsmHBgMvpEoCct7hYKLJxAmHA50g0F6rZbLBgLdCwRyTiTlJSWmex82MXDwLsdCQ/TxkZB4RPz8/GjduTOPGjT37nE4nV65cSSM+NmzY4EnrqtPpqFatWhrxUb169UK3Up8XFC9enL59+9K3b19u3LjB1q1biYyMZPDgwWi1Wl588UXCwsJo1apVgVq5vheFQkGfPn148cUXGTFiBG+88QbdunVj8uTJ93ULK1WqFBEREYSHh9O5c2fmz59P+/btH3ocWq2WsWPH8u6773L06FEaNmxI37596dy5M7/++itNmzbN9NoyZcpgs9mIiorKMN7h5s2bqNVqAgMDszweq9XK559/TqdOnahSpQpdu3alZMmSDBkyJNvP7euvv2bq1Kn07t2bqVOnFggxHx8fz7fffsvy5cu5desW5cuXp1ixYty5c4eAgABmzpxJp06dCvRnV0amoHPH5WJwQgLjfXxop9ViEkWOOhxkZy19odnMKG9v3g8MZKvNxpjERBprNFRM5RI9zWRiksFALbUaLbDUYmGR2cwsX19qqVSss1h4KyGBfYGBlFepMEsS3ePjqadWsz0wkGhRZITRyIeJiXzu58fbej2llUqGGo388ZjGfspCQ0YmF1CpVFSuXJnKlSvTsWNHz/7Y2Ng04uP3339n48aNOBwOIHkil1p41KhRgzJlysgZZlIoVaoUAwcOZODAgVy/fp0tW7YQGRnJwIED0el0tG7dmrCwMFq2bPnIVeBzi6eeeor169ezbt06PvroI/bu3cusWbPuG9Pg5+fHunXreP/99xkwYABTpkyhb9++Dz2GV155hUWLFjFlyhS2bt3K888/T7Vq1Vi2bNkDhQbA9evXMxQa7hoa2fm8rlq1iujoaIYPH87mzZs5dOgQa9asydb7J0kS06ZN4+uvv2bo0KGMGDEi392O/v33X5YuXcqGDRuw2Ww89dRT6PV6rl+/Ttu2bQkPD6dx48b5Pk4ZmceBKFHECbys1VJaqQSlkurZtGa20mh4K0Xwv6NUsths5le7PY3Q6K/X83Kq76YFSUm84+1Nx5R94318OOxwsMRsZrqvLz9arVgliXl+fuhT7vVpPj70jo/nQ4OBokolfin7gwvAwkhuIAsNGZk8JDAwkKZNm6aZzDkcDv755580sR8rV64kJiYGAIPBQLVq1dK4X1WvXv2JXwF96qmnGDx4MIMHD+bq1ats2bKFiIgI+vXrh7e3N23atCEsLIwWLVqgecTaLDmNIAh0796dFi1aMGrUKHr37k2nTp34+OOPM820pdFomD9/PqVKlWLChAncuHGD8ePH/29Sf+UKmExZ6l8BzHjzTYYPH86+L7/khRdeYPhLL/HZZ59xa+dOSpQokeF1Zc1mqjkcxB04ABnVBzlzhoZ6fbJ7QxYwm838/OmnfNCqFUVu3WL16NEMfP55XihSJMttOL28GL1wIevWrWPKlCn069cvS9flBpIkcfjwYU/8hZeXF4GBgdy4cQOj0Ui/fv3o3bs3pUqVyrcxysg8jtRQqWiq0dAyJoYWGg3NtVo6aLX4ZWPRo7pKlVyHy+XCJYoEATdtNhJFEaPTiSRJ1EklOhJFkTuiSIN7BE0DtZpzKWn8Lzmd1FSrPSLDfVwELrtcFH1MxUVq5BiNJwk5RqNQ+TZGRUWli/34559/cLlcCIJAuXLl0tT8qFGjBiVLlnziV0gvX75MZGQkW7Zs4cKFC/j4+NC2bVvCwsJo1qxZgfHZdyNJEt9//z0TJkxArVbzySefPNA1avny5UyYMIHQ0FC++OILtDdvQoMGyZ/1bGC2WBBFEe8U0WpKSkKtVuN1nyKDJpMJtVqdYSFCs9mMIAjodLos9W+z2bDb7Xh7e2O323E4HHh7e2fZIiJJEia7nTYBAQz78ktef/31LF2X0zgcDiIiIjzxF/7+/kiSRGJiIrVq1aJv37507NixwFrZZGQKDfeZx0iSxAmHg/12O9ttNqJEkW2BgQw3GqmqVDLZ29sjIsJNJgKAT9RqXKLIizYbbwoCb6Zqr4so0lIQeEel4pYg0MbhYHdQEDVT+k0URarevcv3AQE0TrWYNSkxkXNOJ5sCApicmMhfTiffpVpAMooi1e7e5YeAABppNPxstRKekJBxbIabQjqPAdmiISNTYAkODiY4OJgWLVp49tlsNv7+++804mPRokUkJCQA4OvrmybjVY0aNahateoTNcGpWLEiQ4cOZejQoVy8eNFj6di0aRN+fn60a9eOjh078vzzzxcI0SEIAq+//johISGMHTuW/v3706FDB6ZPn55pQHV4eDglSpRg8ODBvPHGG6waMQIfhwMUCshGimWNSkVCXBwKSULv5YXC5SLJakVjMKDIRLA6lUokhQJtBlYim9mMRqVKngQ8AFGSMJlMaL28cKlUmJKS0Ht7o8jiZ1WUJJKMRnA6mfPRRzTOB5ERFxfHt99+y4oVK7hx4wbBwcGoVCrMZjMvv/wyffv2pX79+k+8+JeRyQ0kQBJFXKKImLJVd7moKkn0USpp6XCwMToavSRxTZKItlgAcEkS5yWJhgoFLoUChVKJIAhoNRr8dDoUCgUKhQJNQgIGrZaiBgNWlwshOjpN/z4KBcUUCo47HGmExnGHg2dTvocrq1RssFgwS5LHqnHc4UABVHwCrBkgCw0ZmUKFVquldu3aaVKHSpLErVu30tT82LdvHytWrECSJBQKBRUrVkxX9Tz4fqn0HhOqVq1K1apVGT58OBcuXCAyMpKIiAjWr19PQEAA7du3JywsjEaNGuV7DZRixYqxbNkytmzZwrhx42jevDlTp07llVdeyfB9ateuHZs2baJXr14MGTKEZaKIIpvWShWg9vbGZLXiZTDgZTCQaLVidbnQZ2aVUKtxiGL61UTALknJbmpZGIPZZMIpCPj7+hIfH4+gVqPLYjIEURSJi48Hlws/vT5NIoa84N9//2XJkiWsX78es9nscXdTKBS8//779OrVq8AWB5SRKeg4HA6io6O5c+cOUVFRaR7VFy4wLikJm9mMQxRxu+T8KUkcA5ooFBRRKDgDxALVtFr8gek2G78bDJRXqVhqtWKx2dBptQSmZJ1SOhxotVp09yx0POgXcrC3N7NNJsopldRUqVhvsXDW6WR+SoKPTl5ezDaZeD8hgeEGAzGiyPjERF738noi3KZAFhoyMoUeQRAoWbIkJUuWTBNQbLFYuHDhQhr3qz179pCYmAgkx4vcKz4qV65cIFb5cxpBEKhevTrVq1dn1KhRnD17loiICCIjI1m9ejVFihTxiI7nnnsu37IVCYJAWFgYTZo04cMPP+Sdd95h8+bNzJo1K8OJa/369dmyZQsTX30Vs9mMVq3O9vvn7e2NxWIhyWTCx8cHLy8vzGYzOp0uwx9ZpVLpSV6QGkmSkCQJZRbcnkRRJMlsRq/XY7NacTgcBAYGPvBHHcAlisTFxSGKIgF+fqhcrixc9eikjr/YsWMHSqUSlUqFIAhUrFiRadOmERoaWuDigWRkCgpms5moqKh04uHex9jY2DTXKRQKihYtSnBwMPU0GpRKJV4qFV4ajcf6UEYUWWk2s8HpJFEUKa1QMMXXlw56PQ5J4nJiIsPNZlTAAL2e53MoaqCvTodRFJmcmOhJb7vSz4/yKQtXOkFgrb8/E00m2sXGpklv+6Qgx2g8ScgxGoXOtzGnkSSJ69evp3G9OnfuHNeuXQNArVZTqVKldFXPs1MToTAhSRKnT58mMjKSyMhIbt68SbFixTyio379+vma8Wv79u2MHTsWq9XKRx99RJcuXTK0bsTu34+2VSuskoQhICBDt6b7YUpKIikpiaCgIERRJDY2loBM2jFbLBiNRooVK5ZGGDidTqJjYggMDETzgO+XRJMJs9lMUGAgsbGxeHl5Zanyu9PlIi4uDoCAgABUopjr97bD4WDz5s0sWrSI06dPo9frsVqtaLVaOnbsSHh4OM8++2yu9C0jU9CRJImEhIQ0AiIzEWG6J1mFRqOhWLFiBAcHex5T/+1+DAoK+t/ijzyPKXTzGFloPEnIN2ihu0HzisTERC5cuOARH+fPn+fcuXNYUnxag4OD04mPihUr5ru7UU4iiiK///67J5D89u3blChRgtDQUEJDQ6lbt26+uJrFx8czefJkNm7cSIsWLfj000/TZy06cwapSROMdjtWhwNfP790LgD3Q5IkoqOjUWs0+Pn5ERMTg1KpJCCDqvY2u524uDiKFCmCKpXlx72/aJEi97UIuUSR6OhovL29cTqd2O12ihQpkmlMiBuH00lcXBwKhYKAgIBky0ku3tvu+Ivly5fz33//YTAYMJvNPPXUU/Tq1YsePXpQtGjRHO1TRqag4HK5iI6O9giI+1kh3MVp3fj4+KQTDEWLFqVYsWJp9vn6+mb/O1WexxS6eYwsNJ4k5Bu00N2g+Ykoily7di1N2t1z585x48YNIHk1qmrVqmnqftSsWRP/DCanhQ1RFDlx4gQRERFs27aNqKgoSpUqRVhYGGFhYTz99NN5Ljp++eUXRo0ahdFoZMKECfTo0eN/1paUe1vSaDBaLFgsFgwGA97e3llyRwKwWK0kJCQQGBiIy+kkwWhMJyYgeQJyNzo6ncXDff29lo57MSYmYrVa8U2JzfDLgiiy2+3ExcejUqkICAj4nyjJhXv78uXLLFmyhHXr1nkybEmSRKNGjejbty/t2rV7LN0LZZ4M7Ha7RyDcTzxER0cjimKaa4OCgjwiIbUF4l5RkdWscw+FPI8pdPMYWWg8Scg3aKG7QQsiCQkJ6dLuXrhwwbOqVaJEiTQ1P2rWrEm5cuUKRJXmh8HlcnH06FEiIyPZtm0bMTExlC1bltDQUDp27EiNGjXyTHQYjUY+/vhj1qxZQ5MmTZg9ezZly5ZNc29LajVJSUmYTCZ0Ol3yqmEW2paAmJgYFIJAQEAAd6Ojk12a7gnQloCoO3fw8fVNEzBuSkrCbDYTfJ9VfpfLRXRMjCcuRKlUEhAQcN/xWW02EhISUKvV+Pv7p7V85NC9LUkSv/76K4sXL2b79u2esRoMBl577TX69u1LrVq1Hrp9GZncRErJ4JZaKGTmyuTOUOhGpVJ5rA0ZuS25/y5SpEjBENjyPKbQzWNkofEkId+ghe4GLSw4nU7+/fdfjwBxi5A7d+4AoNPpqFatWhrxUb16dXyymGWooOB0Ojly5IhHdMTHx1O+fHk6duxIaGgo1apVyxPRcfDgQUaMGEF0dDRjx44lvEEDFM2apbm3LVYrxoQENFotfn5+D3RNgv+5P/n7++NwOLCYzRQpWjTdtW4R4pMqoNFoNOJwOgkKDMy0/QSjEZvNhk6nS47RCApKZzFJjdtK4uXlhZ+fX3pB8oj3tt1uZ/PmzSxcuJA//vgDpVKJy+WifPny9OnTh+7duxN4n+cjI5ObiCmJD1ILhrt372ZohXC7ubrR6XQPFA/BwcHJFsJ8jEPLNvI8ptDNY2Sh8SQh36CF7gYt7MTExHD+/Pk07leXLl3yZC0qU6ZMGuFRo0YNypQpUyh++BwOB7/++isRERFs374do9FI5cqVPe5VlStXztX+k5KS+OSTT1i+fDmvV63K3BMnUHp5pbm3bXY78SkuR/7+/lnKCBUXF4fL5cI/IICY6Gh8fHzSVaGPTYmV8E9JDQkkp5uFDOM6IDmQ2x2bYU7JOOVzn8wrSWYziYmJ6HU6fDKzyjzkvR0bG8u3337LkiVL+O+//1CpVCgUClq0aEF4eDgvvfTSYxV/JFOwyCx9673i4e7duzhTKky78fPzSyMcMhIRwcHBGAyGxzN9uTyPKXTzGFloPEnIN2ihu0EfRxwOB//880+62I+YmBgADAYD1apVS+N+Vb169XST3YKEw+Fg//79REZG8vPPP2MymahWrZpHdFSoUCHX+j569ChfDhjAgjNnUHp7o/PxSTMpdzidxMfFQYpL1P0sCO7zY2Ji8PXxwe5w4HQ4CCpSJE2bCUYjznusFzExMag1mnSuVm7iExJw2O0oVSpcLhdFgoIynAhJJFcfT0pKwtvbO3nClOlgs3dv//PPPyxZsoRvv/0Wk8mEIAj4+vryxhtv0KdPH6pXr/7ANmRkMsNisWRobXBbIdz7YmNjST31UigUFClSJEPrw70B1VqtNh+fYQFAnscUunmMLDSeJOQbtNDdoE8KkiQRFRWVxvXq3Llz/PPPP7hcLgRBoFy5cmlqftSoUYOSJUsWuFU7m83Gvn37iIyMZOfOnSQlJVGzZk2P6ChbtmyO92k9fhxCQjDZ7QgaDb6+vqhTrci7XC7i4uMRRRF/f/8Hpp91uzj5+fkRFxeXLvA7KSUdbnBwsGdf1N276PV6DN7e6dpzixedTofFYiHA3z/DCZNEsguWxWLBx8cH7weJyyzc2+74i6+//prt27d7Pk+VK1emX79+vPHGG/ilsszIyKRGkiSMRmMa4ZCRiIiKivLUKHKj0WgeKB7SpW+VuT/yPKbQzWNkofEkId+ghe4GfdKx2Wz8/fffaSwfZ8+e9QQ0+vr6psl4VaNGDapWrYpXNlK75iZWq5U9e/YQGRnJrl27sFgs1KlTh9DQUMLCwihdunTOdJRybzsUChLMZlwuF97e3mmyTomSRHx8PA6HAz8/P7zuszLqTkGr1+ux22wo7kl1a7VaiU9IIDg4GIUgIAF37tzJNINUfHw8DqcTSZLQpAR134tEcqIBm9Wa9fS897m37XY7P/74I1999RWnT59GFEU0Gg0vvvgi/fv3p2XLlvLk7gnG5XIRExNz38xL7r8zSt+aWQD1I6dvlbk/8jym0M1jZCdUGRmZAotWq6V27drUTvXFKkkSt27dSlNwcN++faxYsQJJklAoFFSsWDFd1fPg4OA8/9H38vLi5Zdf5uWXX8ZsNrN7924iIyP59NNPmTp1KnXr1iUsLIzQ0FBKlCjxyP2pVSqCgoKSXY9MJs+kXa1SebJJJSQkEB8fj28GsRdulAoF3no9SWYzBoOBxMREnC6Xx+3KPUF3uVwoUlyhgAxjaxwOB1abDY1Gg8PhwCeDwnxpRJC//31F0IOIjY1l1apVfP3119y8eROAwMBAevbsSXh4OJUqVXrotmUy4MoVuKcQW37icDiIiYkhJjaW2JgYYmJiiI2NTfMYFxfnqS6fGn9/fwIDAwkKDKRKUBCB5csTFBSUvC/lMTAwMDl9q8EA5cvn07OUkSk8yBaNJwl5JaDQrQTIZB2z2cyFCxfSuV+5q9EGBgamER81atSgSpUq+ZKy0WQysWvXLiIjI9mzZw8Oh4MGDRrQsWNH2rdvT7FixbLXYAb3tsPpJCEhAZfTmWzdSIl1yGoMhJhSxE+j0WC329OkuhVFkai7d/FPEQV2h4PY2NgM627ExcfjdDpxuVwZukOJkkRcXBxOp5MAf3802alqnurevuTlxYIFC1i9ejWJiYmoVCqqVKnC4MGD6dKlS6HLcFYouHIFGjRIfh9yEQlAkpAkCVEUkVL+zuj/jKY0giAgKBQIgoBCEDz/e/5OtWUZtRqOH5fFRl4jz2MK3TxGFhpPEu4bVKGAJymjitMJolgob1CZR0OSJK5fv57G9ercuXNcu3YNSM4hX7ly5XRVz4sUKZJnYzQajezcuZPIyEj279+P0+mkUaNGhIaG0r59+6xVn87kx1fif/EUSqUSP19fj7ByZ3XKNHUsYLZYMBqN6HQ6rFYrRVNS3UpAVFRUclFAvT6dK5UbtwBRKpUIgkBQUFCaflwp6TtFUSQgICBNXElWkBwOXGYzIxs3ZunRozgcDry8vHjppZcYNGgQzZs3LxQZzAotj/ibIgGSKOKSJCRRRJQkRJcLMeVvyeXyHLt3quIWC0pBQFAqUQgCCoUCRcrfnmMKRZaLVmYZ+Tcl/5CFRqH7zMlC40kij1afCiTy6pNMKhITE7lw4UIa8XH+/HlPLvrg4OB04qNixYq5nvI0ISGBn3/+mYiICA4ePIgkSTz//POEhYXx8ssvZ17T4QE/vg6nM7nOhcORbMXw9kYQhPsXw+N/RfwEQcDpcGBIZZGIjolBk5JlKslsxmQyUSxVcLgEHkuFKIoEBgamCUJ3ulzExcUBZCkb1r3jslosWEwmVE4nL6jV3CpShPDwcPr160e5cuWy3JbMI3AfgSuKYvLmcuHK6O+U7d4piEKh8GxKt3DI4H9BEHJeQGSVQjzpK/TIQqPQfeZkofGkUcD8afMM2Z9W5gG4XC6uXbvG2bNn09T+uHHjBpCcQaZq1app6n7UrFkzw8DmnCA2Npbt27cTGRnJr7/+iiAING3alLCwMNq1a5e23yz8+EqAOSkJU1ISSoUCXz8/NGo1doeD+Ph4FAoFAQEB6Wpt2Gw24uLjUavViKJIkZRUt/Hx8UiSREBAAMbEROx2O0WCgv53XUrxPwHw0unwSxWb4XA6iUupxZFRn5khiiJJZjPmpCRcooga0CkU7J48mdbDhuGdQcYrmZwjdfrWqKgoHKdO8fL06dgVCpwk30NuAXEv9woFpVtQpP5fqcw/8ZAdCvGkr9AjC41C95l7fIWGPKGWkZHJARISEtJUOz937hwXLlzwZKIpUaJEmriPmjVrUq5cuRzNaBQdHc22bdvYsmULR44cQalU0rx5c0JDQ2nbti2+165l+cfX6XKRkJCAw+FITkdrMHgqECNJ+N/jwuS2TLhcLlwulyfVbWJiIjabjSJFihCfkJBstQgI8FwTGxuL0+lEEASKFCnisZbYHQ7i4+JQqlQE+PtnybXJ6XJhMpmwWCxIkoQgCGi1WgxaLWpJQiiEP74FhXvTt94vC9O96VtrSRI/REXhVCpBpcrU+qDIDfel/KQQT/oKPbLQKHSfucdTaMguQrLYkJHJRZxOJ//++2+62I87d+4AoNPpqFatWhrxUb169RwJSL5z5w4//fQTERERHDt2DLVaTY86dZi8ezdKvR5FFoKpJZKD500mEwqFAl9fX1QqFfEpgdv+/v5p6ma462AoFArUajUB/v6YzWYSTSaCg4OJjY1FpVJ5rBZWmy25SCDg5+/vSVX7IFete8dot9sxpVhLJJJXxfV6Pd56fbKQK8Q/vrlNRulbUwuJ1PtsNluaaw0Gw33rPrj/9vvvP4SQEHnSJ5N3yEKj0H3mHs+IYJMp+U15EoOeHY4n05IjI5OHuDMaValShVdeecWzPyYmJo3b1e+//87GjRtxpCx6lClTJo3wqFGjBmXKlMlWwHKxYsXo06cPffr04fbt22zZsoXTq1djtVqx22wotVq8vLzQaLWZTuQFwFuvR6vVYjQaiYuLQ6/T4e/vj9FoJD4uLk0tC7VKhc7LC6vNhs1mw+lyoVSpPFl/RFH0PAd3VisATcpYACxWK8aEBLT3CT53X2+xWDCZTDidTs/rbTAY0Hl5PfF1Cex2e6aVp1Pvj46O9qQddhMYGOgRCxUrVqRx48bpxENwcHCmaY/Tcf16LjxDGRmZx4nHexauUj1ZiheSFa+MjEy+EBQURNOmTWnatKlnn8Ph4J9//klTcHDlypXExMQAyavH1apVS+N+Vb169SxN9ooXL07//v2hUSPE559HCVgcDuITEjzuRV5eXmg1mgwn6CqlkoCAACwWS7IrlN2Or48PCoUiOTVuSuE/IWWc1pSVb7PZ7Bmfy+lEdLk8rmI2qxWHw4EgCPj6+CDwvwxXOp0uuYhZBs9FFEVMSUmYzWZEUfSM38fHB7Va/Xi53mSAyWS6b9E492N8fHya61QqFUWKFPGIhDp16mRogShatGi+pHKWkZF5snm8hYaMjIxMPqNWq6levTrVq1f37JMkiaioqDSxH0eOHGH16tW4XC4EQaBcuXJpCg7WqFGDkiVLZrqir1Ao0Gs06H18cLpc2KzW5LSz8fFpRYdWm2bSLgB6nQ6tRkOC0UhcfDw6nQ5vb29MJhOiy4WPry9KpRK9Xk9SSqyEO/Da4XR63JokIDHFmuHt7Y1SpSLxATU7HE6nJ97DXXDRfW5Wg8QLKlJKjZAHVZ6OiorCbDanudbLyyuNYKhcuXK6CtRFixYlMDBQTuErIyNTYJGFhoyMjEweIwiCZ7L4wgsvePbbbDYuXryYxv1q0aJFJCQkAODr65sm41WNGjWoWrUqXve0r1IqUXl74+3tjdPlwnqP6PBK5V7lnvgrU6wb1hTrhiAI6PV6LGYzLlHEz88Pb29vz/9WqxWlQuFxb1IqlVitVpwOB0qVCm9vbxITEzGbzekK9UkpzzUxMdHjVqZSqTB4e6PT6wu89cLpdBIdHZ1GOGTkynT37l3P83Pj6+vree9LlSpF3bp1M4yDMBgMT7ybmIyMTOFHFhoyMjIyBQStVsvTTz/N008/7dknSRK3bt1KE3S+b98+VqxY4bEAtC5enK8tFiSXC6UkoVKpPJl+VEolhpTaGalFhyU+HoUgoPXyShYdGg0CyYHsmpTYDbPZnJwCNyVVrb+/P94GA0aj0VME0C00FAoFRqMRBAFfX1+MRiNWqxU/X190Ol3yc+F/BQTd8QMFyT3KarVman1wi4m7d+8SExOTpv6DQqGgSJEiFC1alGLFilG1alWaNWuWxvrgdmHSarX5+AxlZGRk8hZZaMjIyMgUYARBoGTJkpQsWZLWrVt79pvNZi5cuMC5c+eI2bcP6fRpLBYLjpSigwqFApVKhVqlQqVWo1KpPFYDg7c3DqfT415lsVhQKBQe9yqNRoO/vz9Wq9WT0tTpdBIbG0uAvz9KpRKX04lKqfS4elmtVlwuF1qtFovZjN3hwN/fHy+tFpcokpiYmCY9rbdej8HHJ9fdoyRJIjExMZ3VIaOAaqPRmOZatVqdRiQ899xzGVofgoKCcr2Yo4yMjExhRP5mlJGRkSmE6PV66tatS926daFOHdi0CZ1GgyvFncnpdOJ0OLDabLhS/P8FQJlKfKg1GnR6PWKKK1Rq0eF2rwoKCiIxMdEjJGJjY9F7eydXGnc6kSQJpVLpcbdCkrA7nQT4+4MgEB0Tg91uxwF4KZUYDAb0OeAe5a4+jcvFiaNHuXL2bIauTA9K31q0aFFq1qyZYSpXPz8/2X1JRkZG5hGQhYaMjIzMY4LbVUqlVEIqFx1RktKID4fTiTUl+Brw1MfwSkkf63K5sNntmMxmvhUEfgBuSRKBksRrQH+TiXmCwC8uF1FAEVGkHfC2QoFLFNHr9XwSG8tuSaIbsFwQuA3cCA5+4HOQANHlwpWSNlcUxQz/F0URpSShBcaOHcsFtTpd+tZGjRqlC6DOVvpWGRkZGZlHQhYaMjIyMo85CkFAo1ajUashVbyEy+XC6XTicDhwOp0eqwUku2x9KQh8L4qMEASelSRilEqukiwEdMAUoAhwOeVvgyTR2+UiMTEREbghCPyq1bIqJduUM0Ug3CscXO79KVtqBJKFkLvKtFqlQqHVJlefFkVULhfr168noFkzOX2rjIyMTAFDFhoyMjIyTyCprR9eGVg/4ux21iQm8qFKRUdJQhJFSosidSQJCeifqq1SQE9ghyTRWxDQaDSoJQmHw8EkpxP/uDgkSSI6df+CkCwWUkSEO4BdmSIoUm+ZOi+lVMsNDg5+8momycjIyBQCZKEhIyMj8wQikRwoLaWICDHlb1EUkSSJyy4XDqCpWo02xZ0qtevSDmAt8H+ABXAChpQ2nQ4HLqCEIFBUpUKpVHo2Vcr/giDke5YpGRmZQkpKtrsnhkL8fGWhISMjI1MIcblcmEwmjEYjtn/+oZzTiQs88QuZCYjUx6RM2hYAURCSRYPTiUuhSL4uRWicBsYBbwONAR/gZ2A1yZYKSZJwAVqSK4WTElDtDqxWCILHFereTen+O+W4LEZkZGQ8GAzJ1ssUa+YThVqd/PwLGbLQkJGRkckH7HY7RqMRo9FIQkICiYmJJCQkePZltrnPdaedBajmcBBhsWADnIKQbC1IcU1yPyoUChSCgJBqX5rHVMckSUJrs+GVkMBhl4tXU+I23JwGSgD9Uu27lfIoSVJymlxA5XTipVbjcDqTYy8kKVl0pPSROiYjo/gMuL8oUYkiSpeLmKgo/O12NBpNTr5FMjIyBY3y5eH4cTCZ8nskeY/BkPz8Cxmy0JCRkZHJJpIkkZSUlG1xkPrce1OupsbX1xcfHx98fX3x8/PDx8eHp556Ch8fH/z8/PD19fVsfn5+FL19G+/wcLw1GoRU1b6ziihJ2O127BYLdrvdU4QvXKFgriiiFgTqSBJGpZK/XS7KKxTcFkV2ALUUCg6IIntJSZ+rVGK32bABkkKBj68vKqUSURRxuDNfpdrcma+USiVarTZNjIb7tU4dPO5MES2iKKJKyTrVrVs3LqjV+Pr6EhwcTNGiRSlatChFihTx/H/v37IokZEppBTCyfaTjCw0ZGRknjicTicmk+m+IiG1OLhXJCQmJnqyM92LSqVKIwLc4qBkyZJpBMK9x937vL29USqV2XtCZ86AQpG8ZQEJcDgc2O12bDYbTocDiZTJvkaDXq/H6XQywGxGASwAooAgp5M31Go6iSI9gBmAQxQJEQT6SxKLU56/SqsFsxmXKBIdHY1Op8NgMKDVaNCmmuBLgMvpTCNAbFYrrhTLhiAInkKDarXaU/9DIQjJaXBtNrDZmDFjBld9fLh7926a7eLFi9y9e5fY2Nh0r4Gfn59HkKTegoODPVW+3ZuczUpGRkbm4ZCFhoyMTKFCkiRsNls6QfAgC0Lq/UlJSZm2r9frPZN/92PRokWpWLFipuIg9eauRVGQcE/obXZ7suXCbkeSJBQKBRqNBp1Ol7zCL0kkmc0etyxvvZ5RXl68bTTicrnwMRgwWyy4JImhgsBwpRK1Wo2YUnejr5cXDocDtUrFuKAg3k1IwCWKWC0WrFYrOp0uWUilCCIBPEIiNWJKQHlqAWK1Wv9n/VAoUKnVaAQBjSRRtGhR6r78cqbVuR0OB7Gxsdy9e5eoqCju3r1LdHR0mr8vXrxIVFQUcXFx6a53fwZSi5DMLCWyKJGRkZH5H7LQkJGRyVNEUSQpKSlL7kWZHXc4HBm2LQhChiKgfPny6cRDRiLBx8fnsZkoukTRIyrsNhsuUUQQBNRqNd7e3mi1Ws/E3G63k5iYiM1mQ6lQYDAY0Ol02G024uLiUCoUBAYGYjabcTmdSCS/1l5eXslB5SmuTna7HV9fX4xGIyq1mqAiRUhMTMRisaAUBCwWCxaLBb1ej7e3N4pMBJkiJUWu5l7rh8uVRoDY7HZEl4u+ffvyr7c3VapUoWbNmlSvXp0aNWpQo0YNAgMDUavVnoJ9D8ItStwiJKPtwoUL3L17N0NR4u/vn6Hr1r2bLEpkZGSeBGShISMjky0cDofHgnA/96LMxENiYqJnZfpe1Gp1hjEIpUuXzlQcpN68vb09vv1PGpIk4bDbsVqtaeIsVCoVXikWC41a7bG2SIDVasWclITD6USlUuHn54eXlxcAiYmJmM1mvLy88PX1xWazYbFYUKvVntgKL602ucK404lOpyMpKQlBENDr9SQajcltplh5jEYjSBJqtRqz2YzFbEbv7Y1er89UcKQmTd0P906HA8lmY+7cuZyw2Th37hznzp1j8+bNnhiYYsWKUaNGDapXr+4RIRUrVsx0kp9dURITE+OxlERHR3vEiPv/CxcuEBUVRXx8fLrr3aLkQZaSoKAgWZTIyMgUSmShISPzBCFJElarNdsWhNT7LRZLpu17e3uncy8qUaIEVatWzVQcpD5Xm6pwnMz9cTqdnD59mgMHDnB1yxammkzJAdgqFRqNBm9vbzQajcdNyY0oSVgslmTrhMuFVqMhICAAjUbjqd4dHx+Py+XC19cXnU6Hy+nEaDSi1Wqx2WxoNBqcTidqjSbZqiBJeOv1JCUlkZSURFBQEE6nk/j4eIKCgtBqNAQFBXmsGxq1GqVKRVJSEmazGe8UwfEwDmeCIFCnTh3q1K7t2edyubh69Spnz57l/PnznDt3joiICL7++msgWUxUqVLFY/Vwb0FBQdnqW61WU7x4cYoXL/7Ac92iJCPXLfd27tw57t69m6koyUiE3LsVFlFy3eWiYXQ0uwIDqfmQ451jMrHdZmN3Nt83GRmZvEMWGjIyhQhRFD1BzFm1INy7uVe670WhUGQYg1CpUqX7igP3o8FgyNRHXubRkSSJK1eusH//fg4ePMjhw4cxGo34+PjQpXp1tFotOr0eZSZZp1yi6LEkSJKEl06HXq9Hneo9s1qtJBiNHlcptUqFKEnEJyQkF9kjOWDc5XKhTenHnQ7XHWvhsNtxuVz4+/kRExtLfHw8gQEBKAQhjXXDYbXi7e2Ny+XyWE+8vb3R6XSPXDtDqVRSsWJFKlasSFhYmGe/0Wjk3LlznD9/3iNCtmzZgtVqBSA4ODiN5aNGjRpUqlQpRybu2RUlbutIRpYStyiJiooiISEh3fUBAQEZWkrutZq43cryg5IKBX8UKULgE2qBlJF5UpBnBTIyeUjq2glZyW5075a6dsK9aLXadO5FgYGBlCtX7oGZjnx9fZNXlAtYEPOTTkxMDIcOHWL//v0cOHCAmzdvolKpqFevHgMHDqRZs2bUqVMH1fnzsHdvhlmnHE4nZrMZq8WCIAjo9Hr0en0aS4dEelcpd2anxJRAcD8/P+Lj4zF4e2NKSvK4WClShAYkB48nGI2YzWZ8fXzw9/cnNjYWo9GIr58fAqSxbphMJjQpFhWLxYLRaCQpKQmDwZAcVJ/Dr6evry+NGjWiUaNGnn0ul4tr16553K7OnTtHZGRkGutH5cqV07lfFS1aNIdH9z/UajUlSpSgRIkSDzzXLUoys5TcuXOHv/76i7t372YqSjILbE9tNQkKCsrRhQSlIBB8n+xq7qKPKvk7SUamUCMLDRmZLCJJEmazOdsWhNTn3q92grtuQuqtdOnSmVoQ7j1XrgtQ+LFYLBw7dowDBw5w4MABzp49C0DVqlV5+eWXadasGY0bN8bb2/u+7Ugki1pzUhI2ux2lUonBxwedTpcuHiIjVyn3GRaLBYvVip+fX3JAd8rEUEgJ1gYQFIrkSuOAl06HMUWw+BgMqFNiNOITElCp1Xjr9QDprBvx8fH4+Pjg7e3tsdi5BYf2IeqCZAelUkmFChWoUKECHTp08Ow3Go0etyv347Zt2zyug0WLFk0TdF6jRg0qV66c5xaC7IgSu92exjpy7+YWJVFRUckxNakQBMFjKXFvNUWRfnY7IiCkZDFzbwKw12bj86QkLjqdKASB+mo1UwwGyqlU6VynDtvtvB4Xx2p/f2aaTFxwuVjn78/zWfhe+9Zs5vOkJOIkidZaLZ/6+OCbIqT/cDj4xGTiL6cTpyRRU6XiIx8faqd6n/5xOhluNHLa6aSsUslUHx+6xsWx3M+Ptl5emXUrIyOTBWShIfPEkLp2wsMUWstq7YTUloPixYtnKg5S73+o2gkyhR6Xy8Vff/3FwYMHOXDgAMeOHcNut1OsWDFCQkIYOHAgISEhWQpMhv8FeCclJSXHUKQK8M5osp6Rq5Qbh9NJYmJisvVDqcRms+Hn50dSUlKayb8nuFwUUaaky7XZbFhtNnReXnh5eeGd0pZKpUpTRyO1dcNoNKLRaPDz9cWVIjji4+NRq9WeGhx5ia+vLw0bNqRhw4aefaIoprN+/PTTTyxcuBBI/h5Ibf1wC5CiRYsWCGuhRqOhZMmSlCxZ8oHnukVJZumAb9++jenKFXq6a7Hc8/wUCgW3gO6CQHW1Gqsg8IXTSZ+4OH7y9cWdN+7etBDTTCYmGQyUUSrxz4Jb1VWXiy02G9/4+5MoSQwzGhmbmMh8Pz8ATJJEFy8vpqnVSMBCs5ke8fH8GhSEQaHAJUn0iY+nlFLJTwEBmCSJj57EqtMyMrmELDQewAaLhUmJiVwIDs70nHsD0j5ISCBBkljh759Ho3x0rl+/TsOGDdm1axc1a9bM7+FkiM1my5JIyMwV6X61E3Q6XToRUKRIEU/thPvFJhTU2gkyBZNr1655hMWhQ4eIj49Hr9fz/PPPM378eEJCQqhSpUq2Pk8mkwmN3Y4pKQm7JKHVavH18UGdEuB9L5m5SrkRJYn4+HhUSiU+Pj7ExcV5iuY5nU4MBoPnXPd1oiShILkOid1mIykpCV3KarDBYMDpdJKQkEBgYCCqVKL6XutGdEwMPj4+BAQEYLfbMZlMxMXFodFoMBgMaPIx0FmhUFC+fHnKly9P+/btPfuNRiMXL15ME3y+fft2zGYzAEFBQekCzytXrlygrZBZEiVnzkDTphg0GkSlEtHlQhRFTxX3DimPoigiulxMdLloLoqciItDT/JnJiYmhiilkniShdw7gsCzdjuKlArxtlRWErelJDU2SWKery/FUz5T03x86BkfzySDgWClkqb3vMaf+vhQzWrliMNBa62WA3Y711wuvg8I8LhyjTEY6JpB6mIZGZnsU+CExoYNG5g0aRIXLlzI76Fkmbf1esJTXAIKKyVLluSPP/4gMDAwV9rPrHZCVjMdZbV2QurHrMYmPE61E2QKHvHx8Rw6dIgDBw5w8OBBrl27hlKp5JlnnqFPnz40a9aMunXrPtRn8MaNGyxdupTjy5ezzmZD7eWF7wOC8u/nKgXJIsRoNCKJIn5BQThSanH4+/tjs9nSuE1BsusU4InT0Gq1CAoFjpR6F2qVCoHkonexMTHJweGBgelcuO61blitVvxS4oxsNhsmk4nY2Fi0Wi2GFLesgoKvry8NGjSgQYMGnn2iKPLff/+lcb36+eefWbRoEZBs/ahUqVK64PPg4OBCt2ghkFxE8d4MZ1ecTmYlJfG7y0WsKCIKAgqFAouvL6UEAUV8PN4GA3pBQO1wILhcVCd5UUkURcQM0mCnFh2SWk0ppdIjMgDqqdWIwGWXi2ClkrsuFzOTkjhstxMjirgAiyRxI8U6fdnloqRSmSZe5JkC9NmSkSnsyHdTDuCtUHB/j+mCj1KpJPg+VpvUtRMeptCayWRCFMUM23bXTrhXBLhrJzwoNuFJrp0gU/Cw2+0cO3bMY7X4888/kSSJChUq0LJlS0+cha+v70P3cebMGRYuXEhkZCTe3t6MeOUVvJcvR+HlBfeZJN3PVcqNxWzGarXi7++PUqnEmJCAWq1Gq9UmT/Q1mjQiweM6lTIpVKQU8nOn0PVLeZ4KQcA/IICYmBiMCQn4+funW53OzLqh0+nQarVYrVZMJhMxMTF4eXlh8PIqsD9iCoWCcuXKUa5cOV5++WXPfpPJxIULF9K4X+3cudNjcQ0MDPRYPdwipHLlyoUy9XOv+HhKK5XM9vWlmEKBCLwQE4OYUjgSQUCn1WJQq/G22xFsNkr5+3viKyTwWElSW0rcW1Z432gkTpL42MeH0kolWqBDXBwZL1vJyMjkNDn+Hb13714+//xzLl68iEKhoH79+kyZMoVy5cpx+PBhXn/9dS5cuOD5kT179iytW7fm6NGjXL9+naFDhwJ4zLXDhw9n+PDhJCQkMGHCBHbu3Indbqdx48ZMnTqV8uXLA/+zhHz55Zd8NHo0NxMTaaXRMC8ggC1WK7OTkjCKIq/rdHxkMKBM+XFMEEUmJCay02bDDjRWq5nq40P5e36Af7ZamWIycVMUaaxWM8fXl5IpKyAPyuUtShLzzWZWWyxEiSIVlEqGenvTIYtBZhecTj5OTOSow4FeEGiu0fCRj48nLeBrcXFUV6nwEkXWWK1oXn+dXgMGMHz4cCB5AnD27FlGjBjBX3/9RbFixejZsydTpkwhPDyc8uXLc+3aNRYsWED79u1RqVRcuXKFEydOUKZMGW7cuIHdbvfEIaReLXUX8XKnu3zqqad45plnKF68OFWrVr1vFWb3ptVqC90KnoyMG1EUOX/+vCeA++jRo1itVoKCgggJCaF3796EhIRQqlSpR+pHkiT27t3LwoULOXToEE899RSTJ0+mW7dueP/7L6xced9rExMTMVssGbpKuXE4HCSaTOj1ery0Wmx2O3aHg4CAAERRxOFwoE/xfXeTOkbDjU6n81QB9/Hx8fSlUirx9/MjLj4eVVIShkyC2jOzbrhjPiwWC0kmEwkWC95qNcaoKDJfJilYGAwG6tevT/369T37RFHk+vXradLu7ty5k8WLFwP/S9d7r/tVsWLFCux3Z5woctnlYravLw1TLGDH7PZsteFOp5xZ/JpgMnHD5eKOy0WxlHNOORwogIop/x93OPjEx4dWKULtZop1xU1FpZKbLhd3XS6KplxzOpMU4DIyMtknx4WG2Wxm4MCB1KhRg6SkJGbNmkXfvn3ZtWvXA69t0KABU6ZM4dNPP+XgwYMAnuwq77//PleuXGHVqlX4+PgwdepU3nzzTfbv3+9xObBYLCxbtoyF48dj6tmTvjYb4fHx+CkUrPb355rLRb+EBJ5TqwlLmeS/bzRyxeVilb8/PoLAVJOJN+Pj2R8UhDrlC9wCfGE286WfH2pgbGIibyckEJlFN6MvzWa+t1qZ6eNDeaWS3xwO3jUaCVIoaPwAH12jKNI5Lo43vLyYqNdjliQ+SUqiX2wsawwGRFHE6XSywWajj1LJDwoFo7y8+PDDD1mxYkVyG0Yjt2/fTs48YzBw7do1xowZg8PhYOXKlQQHB6PVarHb7cTGxlK6dGlKly7Nn3/+iVqtZsiQIRQrVoxNmzahUChYsGABvr6+XLp0iXfeeYePP/6Yhg0bcu3aNUaOHEndunU9IkdG5nHk5s2bHDhwgP3793Po0CFiYmLQarU0atSIkSNH0rx5c6pVq5Yjlja73c6PP/7IwoULuXjxIs888wwLFy7k5ZdfzlK60Qe5Srlx18tQqVT4+Pggkbz6rlar0Wg0WMzm5PS096ysp47RcKPRaDz1NiwWiyfbFCnXGwwGTCZTctXyTFbq72fd0Ot0yWImMRGn2UyPHj1o1L8/7733Xq6mnc0tFAoFZcuWpWzZsrRt29azPykpyWP9cIuQ3bt3Y0oJVg4ICEiXdrdq1aoFwvrhJwgEKBSstlgoplDwf6LI9FwIstYKAkOMRiYZDCRKEuMTEwn18vK4QpVXKvnOaqWOWk2iJPFxYiJeqcRZM42Gskol7xuNTDAYMEkSM1PGWVBFnIxMYSLHhUbqADmAuXPnUqtWLf7+++8HXqtWq/Hx8UEQhDRuPFeuXGHnzp1ERkZ6VoHmz59PvXr1+PnnnwkNDQWSV+NmzJhBucREUKnoIEl8Z7fzZ5EieCsUVFGpaKJW86vdTpiXF1ecTnbabEQGBFA/ZcI/38+PetHR/GyzEZoiRhySxDQfH+qmCJovfH1pFhPD7w4Hzz7Ar9ouScxLSmKdnx91VSokUeQVpZLDajUrEhN5WqdDSvFFlSQJURTTPC50OqkkSYSbzWA24wd8KEm0kST+jI+nnEKBSxSpDAwWBDSSRIMqVTibmEjx4sUJCwvjxo0bLF++nK+//ppy5crh4+PDhQsXGDRoEMuXL6ddu3aeYPC5c+dSs2ZNDh8+zMGDB1m8eDFNmzYF4Omnn6Znz55UqVIFrVbLyJEjeffdd+nSpQsAZcuWZdSoUUydOlUWGjKPFUaj0XNP7N+/n3///RdBEHj66afp3r07ISEhNGjQIEcneAkJCXzzzTcsW7aMqKgoWrduzSeffELDhg2zPAGypFSBv5+rFCS7qCQkJCBJEv4pLk1Wmw2Hw0FgQIDnf41Wm94SIggI/M91CpJXor28vDAnJWGxWNJV/fb29vYEh6sCA+8rmDKzbiiVSvReXkiCQM+ePZm2cSNr1qyhf//+DBo0CL97LC+FEW9vb+rVq0e9evU8+yRJ8lg/3K5Xv/zyC8uWLUOSJE+63tTio2bNmhQvXjxPJ84KQWChnx/jExNpERtLxZS0sa/lcJB1OaWSl7Va3oyPJ16SaK3RMMPHx3P8M19fRiYm8lJsLCUVCsYaDExJJXiUgsAKf3+GG420jYujrFLJBIOB3vHx5L9ck5Ep/OS40Lhy5QqzZs3i999/JzY21uNHeePGDXQ63UO1eenSJVQqFc8++6xnX0BAABUrVuTSpUuefTqdjnLlyiVnwgCKCgJPKZV4p1pVLKpQEJ0ypksuFypIIxYCFAoqKpVcSmU6VZE2OKySSoWvIHDJ6UwnNCxWK1abDaskEXX3LpdEkSRRpGtsbJrzHEA1IMnlQlAoEFKC5ARBQKlQoFKpEASBf61WTjqdNEm5TgBQKFBIEiZ/f4pptWjj4qiqVOKv04HdzsSJE/l3zhwCAgIYMmQIS5cupWLFinTu3NnTf6lSpRAE4YE/PNWrV/f87U6xGR0dTalSpTh37hzHjx/niy++8Jzjcrk87lQP+37LyOQ3DoeDU6dOedyhfv/9d0RRpGzZsjRr1owxY8bQtGlT/HMhs9x///3HkiVLWLduHU6nk86dOzNw4EAqVaqU5Tay6irlxmw2Y7PZCPD3R6lQeKwZGo0GjUaDKIrY7fYM40oEUmpp3OMzr0sRGk6HA4fdnjaAnOQA6tjYWOLi4wkKCrrv+DK1bqR8T/bs2ZPQceP4+uuvWbx4MStXrmTw4MH07dv3gTVHChuCIFCmTBnKlClDmzZtPPuTkpK4ePFiGgGyb98+Tz0Mf3//dGl3q1at6im8mBuEaDTsv8el+GaqVM2p/35eo0nzf1YYbjAwPCUDWu9MErLUUqvZfo/3wb1uy5VUKiJSnXM8xcWrnJxyXEbmkclxodGrVy9Kly7N7NmzKVasGKIo8sILL+BwODxf+KlXvjLLJPQw3Ju1RSD9ExSArIWQPRyKlMm7UhDQ6/UgighJSawwGCiZ8qPoPkerUNy3MiqA3eHgJZWK8anSSbpJfe291VOFVNV6H4XUr6lblLjFY1JSEiNH3nGsrAAAyKhJREFUjkwT6OimIJjuZWSyiiRJ/P33357MUIcPH04OZPbzIyQkhK5duxISEkLZsmVzbQx//PEHCxcuZOvWrfj6+jJw4EDeeuutbLsCZdVVyo3d4cCUmIi3t7fnvrVarTidTk8WOnehyczua0EQ0mUIUqnVqFQqnC4XZoslXSpXhSB4KocnxMfjn2I5uR/3WjccKhU+Gg0KkifS48aNo1+/fsybN485c+awdOlShgwZQs+ePR/77yRvb2/q1q1L3bp1PfskSeLGjRse4XH+/Hn27t3L8uXLkVIK7FWoUCFN8HmNGjUoWbLkE+U2tN1qxVsQKK9ScdXlYkJiIg3UasrJ2adkZB6ZHL2L4uLiuHz5MrNnz/YUOTp27JjneFDKysadO3c8Zm135Vs3Go0mXVG0ypUr43Q6+f333z2uU+6+qlSp8tDjraxU4gR+dzg8rlPuALYqqb5gnCQHh7mtF5edToySROUMvoS0Wi1ajQarJGHw9qaOKKK1WIhWKmnxECtHtdVqtlmtlFYq04mJrFKxYkVu3rzJ3bt3PZOW06dPP1RbacZWuzb//PNPshVJRqaQcefOHU9mqIMHD3Lnzh3UajUNGzbkgw8+ICQkhFq1auVqIUVRFPnll19YsGABv/32G+XKlWPq1Kl07dr1oSyCVrudhKSkB7pKpe4/IT4edUqNCsBjzdBqtZ6aFVabLTnuIpOYk4wWNgSSK4WbTCasVisuH59016uUSvz8/IiLi8NkMuGTwYLKvaS2biSlVBDft2ULHWrV8rjdTp06lbfffpvPPvuMyZMns3DhQoYPH07nzp2zFNfyuCAIgifm7qWXXvLst1gsXLhwIU3weWrrh6+vb7rA86pVq+a5lbpFTAz/l0mR1Fk+PnTKofGYJImpJhM3RJFAQaCZRsOkVO5XMjIyD0+OfuP6+fkREBDA6tWrKVasGP/3f//H9OnTPcfLlStHyZIlmTNnDmPGjOHy5cueiqpunnrqKZKSkjh06BA1atRAp9NRvnx52rRpw4gRI5g1axYGg4Fp06ZRvHjxNKbj7FJepaKNVsuIxERm+fhgEASmmUwUVyhok2r1Sy0IjE9M5GMfH1TAh4mJ1FOrHxifAWBQKBik1zPJZEIEnksJSDvucGAQBLo84Iuyj07HGouFwQkJDPb2xl8QuOpyEWGzMdvHx5M96340a9aMsmXL8v777zNhwgRMJhMzZ84EHi3YbdiwYfTq1YtSpUrRoUMHFAoF586d48KFC4wePfqh25WRyQ1MJhO//fabR1hcvHgRgJo1a/Laa68REhLCc889lyeTKZvNxnfffcfChQu5fPky9erVY+nSpbRp0+ahhI3NZkOyWjE5HGhTik/ezxUJUsVlkPzd7T7bnUXO7RYmShJ2uz1Nkb57UWRiQfXy8vIELlsslgyzTGk1Gnx8fDyVw3VZXJDRajSo/fywJyYyd+5cvv3jD+bMmcNTTz0FQOnSpfnss88YPHgws2fPZvjw4Xz11VeMGjWK0NDQJzoltk6n49lnn03jjixJEjdv3kxj/Thw4AArV65EFEUEQaB8+fJp4j5qA8XggZaoh2W1vz/OTCzzRXLw/eus09FZdvWVkckVclRoKBQKFi5cyPjx42nRogUVK1Zk6tSpvPbaa0CyG86CBQsYM2YMLVu25JlnnmH06NEMGDDA00b9+vXp1asXAwcOJC4uzpPe9vPPP2fChAn06tULu91Oo0aNWLNmzSMXWvvc15cJiYn0io/HDjRSq1nj7+/JOAWgA97R63knIYFbokhDtZrPspEDf5S3N0EKBV8mJfGfKOIrCNRWqRiSBd/hYkolkQEBTDWZ6BYXhx0orVTyQoq7QFZQKpWsWLGC4cOH07ZtW8qWLcuECRPo3bv3I7kTtGjRgm+++YbPPvuM+fPno1arqVSpEt27d3/oNmVkcgqn08np06fZv38/Bw8e5OTJkzidTkqWLEmzZs344IMPaNKkCUWKFMmzMcXGxrJq1SpWrFhBTEwMbdu2Ze7cuWlSnWaXf//9l0/efZfPHA68DQa8DIYsTfySkpKw2e0EBAR4LA1Syn4vLy+PNcRmsyFJ0n19+TOK0YBki4Varf5f9ilv7wzHptfrcTqdGI3G5ArkWbQ6uGt2zJo1i7fnz6dly5ZMmDCBnj17ehZRKlWqxMKFC3nvvfeYOXMmgwYNYt68eYwZM4YXX3zxiXIRuh+CIFCqVClKlSpF69atPfstFgt///13mrofixYtIiEhgWoOBxEmE6JKhaDRoFKpPNuDhG5WKC3HSMjIFHoEKScc+QsaZ85A06ag0cCTVPHZ4QC7HQ4dgtq173vq8ePH6dixI4cPH5Zdn2QeCyRJ4t9///UEcB8+fJjExER8fHxo0qQJISEhNG/enPLly+f55PLq1assXryY9evXI0kS3bp1Y8CAAZ46QA/L5s2bGTlyJI0NBlZcuoRSp8vSd57NbicuLg6DwZDGymA2mzEmJlKkSBFUKZO8+IQEXC4XQfdJ5x2fkIAoigQGBKQ7ZrZY0gQkZ5bOVgJPApHAwMBM3bTSkOo7L7FcOT7++GNWr15N06ZN01g3UnPixAk++eQTjhw5Qr169Rg9erQns55M1pAkidu3b3N161bqDhmCDbCnpFp3o1IqPXE6bvGoUCpzzfqRp2Tjt1ZG5klHFhqPE/f58tu+fTve3t6UL1+eq1evMmHCBPz8/IiIiMinwcrIPDrR0dEcOnTIY7W4efMmKpWK+vXrExISQrNmzahTp06++eWfPHmSBQsWsH37dgIDA+nTpw+9e/f2xKs9LFarlYkTJ7J69Wo6derErJ490b/0Upa+81yiSExMDCqVioBUAdiiJBEdHY1Wq/VU85ZSsud5e3tnWlwPktP/OpzODMWIKEncjYpCUChQq9UE3CdTV2Zjy5QMvvMOHDjAsGHDPEVeU1s33EiSxMGDB5kxYwZ//PEHTZs2ZcyYMWkCqWWywD2/tZIk4XS5cDqdOBwOnE4nTqfTk0BEIQjJwiOVAMkp60eeIguNvOXKFciFGiwFHoMBHnExqiDw5ETFFVBGG418b7VmeOw1Ly9mZsNF636YTCamTp3KjRs3CAwMpFmzZkyaNClH2paRySssFgtHjx711LM4d+4cAFWrVqV9+/Y0a9aMRo0a5WtKU5fLxa5du1iwYAHHjx+nQoUKzJw5k9dffz1HUon++++/9O/fn3///Zc5c+bQrVs3hL/+ytK17rgMgbRxGZD82kqSlOa1s9ntD3SbgsxdpyB5cqnRanE6HNhsNpwul8daci9KhQJ/f3/i4uJITEzE9yECcps1a8bevXv5+OOPGTNmDFu3bk1n3RAEgWbNmhESEsKOHTuYOXMmHTp04KWXXmL06NFp0nrLZB1BEFCnWC/csTYSeArLOh0OHE4ndrsdi9mMe5VTmeJil1p8KB8X64fMo3HlCjRokCzunjTUajh+vNCLDVlo5DMjDQYGZZL/25CTwW6dO6epo5HnyCsSMg+By+XizJkznuxQx44dw+FwUKxYMZo1a8bbb79NSEiIp8ZLfmKxWNi0aROLFi3iypUrNGzYkJUrV/Liiy/mWOCx21WqWLFi/PTTT9meEJtMJhx2OwH3uCaJkkRSUhI6nS6NCLDZbMkTvwf4ymeU3jY1Op2OeJsNQRCwWCz3zS6lSSnc6o7X0D9EkK6Pjw+zZs2iQ4cODBs2LMPYDfe427ZtS+vWrYn4f/bOO76p8vvj78ymaTpZioqA7CEiw7LKlil7yd5lyd6CIlP2EKSML2UqewkUZSl7ioKIgAioqC3dzV7390fa/loo0JE2aXvfr1de4k3u85zbJDfPec7nnHPgAIsWLaJJkya0bduWCRMmZFnaJuJIFJdJpciUSjxSlDgWIDnikeSA6PX65OiHJDH68bQDkuuiHyJZQ6t1OBlSKeSjinFYrY7rzgPrpnz0rrknBaVSp1bPcEvEHQnR2cgADx8+THYszp49S1xcHF5eXtSuXZvp06cTFBRE6dKl3SaJNzIyMjnBOzY2llatWvHFF184VYbztFTq888/f2EVqLQwmc3odDq8vb2Ty9Ymodfrn4lmCIDJaHT0A3oJL+vb46FUIpVIkMlkjupTL0lYV3t6YrVakytRPW1veklPdAMcO+odOnTggw8+YMeOHSxZsoSgoCC6du3K2LFjKVq0aKbmF3k+EkiOfpBW9CNRfpVm9OMp+ZUY/cgHyOX5SwoPDnleHkB0NESyH3FHwtWWuDUxMTGcPXs22bn4888/kclkvPPOOwwYMIB69erx7rvvZrnCnLO5f/8+a9euZefOnUilUj788EMGDRrk9KZ+aUqlMuhk2ex24uLi8PDweMZxSIpmqD09U0U5zGYzdkHAIx1yL6lUiiAICKRd6lQikeChUmE2mbDb7RiNxpeWsfX29sZqtRKb2Dk8XcnhzxknKboxbtw4GjZsmBzdeDrSpFAo6NmzJ506dWLz5s2sWLGCXbt20bdvXz766KMcrVCWH8ly9OMpB0SMfoiIuJ58tOoTcTnijoQIDjnOlStXkvtZ3LhxA0EQeOutt2jcuDFBQUHUqlULHyflJzkTQRC4cuUKq1ev5rvvvqNgwYKMGTOG3r17J/edcCZJUqlXXnklU1IpcCzSYmNjkUgkz+RlgKOcLYD6qbwWk9GITCZLVyJ9kuMjCMJznSBPlQqDwYBCLkev17/U0UjKI4mOjiY2NpYAf/8sRbGCgoI4efIks2bNYsqUKRw+fPi5lalUKhWDBw+me/furF+/ntWrV7Nt2zYGDRrE0KFD3fKzmZdJT/TDarFgtlgcuUaJ5yV9fhVi9ENExGWIjoaIiEi2YrfbuX37dnJlqEuXLmE0GilQoAD16tWjb9++1KtXz63lKTabjbCwMEJCQvjxxx8pXbo0ixcvpn379lnqRfM8npZKzZ8/P30J7inKiyah1+nAYsHPxwfpU8/b7HbMOh0alQqZzQaJXZgFwGI0ovbwQJIOyaPUZkMuCAhmsyNymQYKiQSlVIoMsJjNWIxGFC/J/ZABfhoN8XFxaOPi0DzdhyON630RSdGNDz74gLFjx74wugGg0WgYPXo0ffv2ZdWqVaxZs4bQ0FCGDx/OgAED0iUrE8ke0hX9SJRfPTf6kSICIkY/RESyB9HREBERcTqPHz9O7mdx9uxZoqKiUKlUBAYGMnHiRIKCgihXrpzbd2fW6/Xs2LGDtWvX8ujRI2rXrs2WLVto2LBhttl+//59Bg8enDGplEbjiBYmld1MxGK1Yjca8fbwQGG3PxNhsxiNKAG1TJbqOZvNhtJuRyWRpCsqJ7PZ8AAwmeA5zoME8JLJsFgsyCQSbDodinTIshSAj4cHRqMRiyCgTLGodLxA4bj+DFCvXr10RzfA0f/j448/ZuDAgaxYsYJFixaxfv16Ro4cSa9evZ61ScRlpIp+pMCWMvphtTqiH0Zjcm5RcvQjhfMhRj9ERLKO6GiIiIhkmfj4eM6fP5/sXPzxxx9IJBKqVKlC9+7dCQoKokaNGrlmQRYREUFoaCgbN25Eq9XSunVrQkJCqFKlSrbOm2mpVIkSjqIDKfKBHj9+zNChQ3n33XcdpayfclbCw8Pp1asXvXv3pmfPnqmeW7NqFd9//z07dux4boQiJX89fEj//v1ZsWIFlSpVeu7r/nnwgAEDBlC/QQMuXbzIrl270hWpUQAb1qxh165dzJ8/n2rVqv3/k5ms7JbR6AZAkSJFmDNnDkOGDGHJkiV8+umnhISEMG7cODp16uSyfi0iLyc90Q+rxYLeYMCeKCdMin487YC49/aIiIh7Id4VRUREMozFYuHatWvJeRbXr1/HbrdTvHhx6tWrx5QpU6hTp0625C1kJ3fv3mXNmjXs3r0bhUJBjx49GDhw4HN3up1FpqVSKUmx2DaZTPQfP56EV19lbGgokjRyChZu2sQ/BQrQ/pNPIGW1KUFg47VrNO3QAWk6HSuPggX5TaEgokiRFzYwK1G5MtIqVfhdpeImsOu33+jbt2+65hi4fDk/REfTa8ECwsLCnJZ0nxTdmD17NlOmTEmuTFWsWLHnnvPGG2+wdOlShg8fzqJFixg7diwrV65k4sSJtG7d2u0jdSIO0hv9sFosGFNEP5RSKSqJhK83bKBQw4ZUrFiR4sWLI3uJFFBEJD8iOhoiIiIvRRAE7ty5k1wZ6sKFC+j1evz8/KhXrx5du3YlKCjohYszd0UQBC5cuEBISAjHjx+nSJEiTJw4kZ49e+Lr65vt82dKKvUSZsyYwZ07dzh06FCaicsPHjxgx44dTJs27RmH5saNGzx+/JiWLVume76kUrvadFRY69ixI/Pnz6dBgwZs3ryZPn36pOt6ZTIZq1evpkWLFvTr149vvvnGaY0Zvb29k5v2Pd1340VOQ6lSpQgJCWHEiBHMnz+fIUOGUKFCBSZNmkSTJk3cpgSzSMZ4WfTDbjKB2czRsDDO7dgBOAoIlCtXjvLly1OxYkXKly9PhQoVcuQeIiLizoiOhoiISJqEh4cnS6HOnDlDREQESqWSmjVrMnr0aIKCgqhYsWKu3cWzWq0cPnyY1atXc+PGDcqVK8eyZcto3759jpXSdUZVqbTG3LRpEwsWLHiujGnx4sUULFiQPn36PPPckSNH8PPzIzAwMN1zJi340+NotG/fntmzZ1OyZElOnDjBlStXqFmzZrrm8fX1JTQ0lNatWzN69GjWrFnj1OhBvXr1kvtupDe6AVCpUiW2bNnClStXmDdvHn369KFatWpMnjyZOnXqOM0+EdeRKvohk4Fczq5du4gqWpTbt2/z66+/cvv2bW7evMnu3buxJBZRKFq0KBUqVEh+lC9fnhIlSogyO5F8g/hJFxERARyLxAsXLiQ7Fnfv3gWgYsWKdOrUKTnPwjMTnZrdCa1Wy9dff826dev4+++/qVevHl999RX169fPsR3olFKpjh078vnnnztld/73339n/PjxdOjQgR49eqT5mt9++419+/Yxb948VE8lYwuCwOHDh2nWrFmGFkJSqRSNRkNCQsJLX/vKK69Qp04dbt26RfHixdm4cWO6HQ2AsmXLsnLlSvr168fy5csZM2ZMus9NDxqNJlPRDYAaNWqwZ88eTp8+zeeff07nzp2pV68ekydPpmrVqk61U8Q9KFCgAHXr1qVu3brJxywWC3/88Qe3b9/m1q1b3L59mx07dhAeHg6Ah4cHZcuWTeV8VKxYMddJTUVE0oPoaIiI5FOsVis//fRTctTixx9/xGq18tprrxEUFMSYMWOoW7cuBQoUcLWpTiE8PJz169ezZcsW9Ho9bdu2JTQ0lIoVK+aoHdkhlQIwGAwMHjyYokWLMn/+/OeOuXDhQl5//XU+/PDDZ567d+8ef/zxhyN5PINoNJp0RTTAIZ8aO3YsH330EatXryYyMjJDzfCaNWvG+PHjWbhwIRUrVuT999/PsL0vI7PRDYlEQv369QkKCuLo0aMsWLCAVq1a0axZMyZOnOiUqJWIe6NQKChbtixly5alXbt2ycejo6OTox+//vort27dYu/evZgTK7u98sorybKrpP+WLFlSjH6I5GrET6+ISD5BEAT++OOPZMfi/PnzJCQk4OPjQ506dZg5cyb169enePHieUpbfvv2bdasWcO+ffvw8PCgV69eDBgwwCV9O7JDKpXE1KlTefjwIWFhYc+Njty4cYOwsDCWLVuWpjzs8OHDaDQagoKCMjy/t7d3uiIaAC1btmTy5MmOCj5SKdu3b2fEiBEZmm/06NHcunWLESNGcPjwYUqXLp1hm19GWtGNadOm0bt375dGNyQSCS1atOD9999n//79LFq0iCZNmtCuXTsmTJhA8eLFnW6viHsTEBBAnTp1UsnprFYrDx48SHY+fv31V/bs2cPKlSsBUCqVaUY//P39XXUZIiIZQiIklVHIS9y8CXXrglKZvzpRJ9XQP3v2hZVfchzx/XDZ+xEZGZmcwH3mzBn++ecfFAoF1apVIygoiKCgIN5+++08t2MmCAJnz55l9erVfP/997z66qsMGjSI7t27u6Src3ZJpZLYsWMHY8aMYfny5XTu3Pm5r+vRowd//vknp06dSvM9b9q0KaVKlWL16tUZtqF169aULVuWxYsXp+v1Q4YM4d69e1SuXJkLFy5w/vz5DOf7aLVaPvjgA8xmM2FhYdn63mq1WmbPns3mzZupU6dOuqIbKbFYLGzfvp0lS5YQGRlJt27dGDNmjFs3qnwp4r092+7tsbGxyXkfSfKr3377DZPJBDhKLafM/ahQoQIlS5bMsfyyHEP8jLnXei4T5K3VhYhIPsdgMHDp0iVOnz7NDz/8wO3btwEoV64crVu3pl69egQGBjp1ketOWCwWDh48yOrVq/n111+pWLEiK1eu5IMPPnDZD3B2SaWSuH37NlOmTKF79+4vdDIuX77MqVOnCAkJSdPJePToEbdu3WLkyJGZsiMj0ilwyKf69OnDsGHD2LlzJ99//z2NGzfO8JyhoaG0aNGCoUOHsnnz5mwrTqDRaPj8888zFd0Ah5ymV69edO7cmU2bNrFixQp27dpFnz59+OijjzIkHRPJ+/j5+VG7dm1q166dfMxqtfLw4cPkyMft27fZt28fq1atAv5fspUy8lG+fPk8I38VyZ2IjoaISC7GZrNx8+bN5IjF5cuXsVgsFClShKCgIIYNG0bdunUpUqSIq03NVhISEti2bRvr1q3j33//pWHDhsyYMYM6deq4VAa2b98+Jk6cyKuvvup0qRQ4dtkHDx5MiRIlmD179nNfJwgC8+fPp0KFCrRu3TrN14SFhaFUKmnUqFGmbMmIdAqgQYMG+Pv78+uvv1KpUiU2b96cYUcDoHjx4qxZs4bu3bvz+eef8/HHH2d4jIxQt27d5L4bU6dOTe4qnt7ohkqlIjg4mO7du7Nu3TrWrFnDtm3bGDx4MEOGDHFJxE0kdyCXyylVqhSlSpWiTZs2ycfj4uJS5X78+uuvHDx4EKPRCDiiH0nldpOckFKlSuW96IeIWyI6GiIiuYyHDx8mOxZnz54lLi4OLy8vateuzSeffEK9evUoXbp0nsqzeB7//PMP//vf/9i6dStGo5EOHTowePBglyfcGo1Gpk+fzrZt27JFKgUO52HChAmEh4dz9OjRZypIpeTMmTNcuHCBTZs2PXf3/fDhwzRo0CDTdmo0Gv777790v16hUNCuXTv27dvHmDFjmDx5Mn/99VemmiMGBQUxbdo0Zs6cScWKFVMl4GYHWY1ugMMxGzt2LP369WPVqlWEhIQQGhrK8OHD6d+/P2q1OluvQSTv4OvrS2BgYKqS1DabLVX0I8n5+PLLLwHH96906dKpen5UqFBBjKyJOB3R0RARcXNiYmI4e/ZssnPx559/IpPJqFq1KgMGDCAoKIiqVavmq92pW7duERISwoEDB1Cr1fTp04cBAwa4ReQmu6VSSWzevJkDBw4QEhJCyZIln/u6pGjGu+++S5MmTdJ8TXh4ONeuXWP58uWZtsfb2ztD0imADh06EBoayquvvopGo2Hr1q1MmTIlU/MHBwdz69Ytxo4dS6lSpZ7bQ8SZZDW6AeDv78+0adMYNGgQK1asYOHChaxbt45Ro0bRs2dPlCmaxomIpBeZTMZbb73FW2+9xQcffJB8PD4+PlX04/bt2xw6dAiDwQBAoUKFnsn9EKMfIllBTAbPS7hr8pD4fmTo/TCZTFy5coUffviBM2fOcPPmTQRBoFSpUtSrV4/69esTGBiY7yQWgiDwww8/sHr1as6cOcPrr7/O4MGD6datW3Jnalezb98+JkyYQNGiRVmzZk22RVZu3LjBBx98QM+ePZkzZ84LX/vdd9/Rt29fduzYQb169dJ8zcaNG/nkk0+4ceNGpmv5L1y4kB07dnD16tV0nyMIAnXq1OG9997Dy8uLgwcPcu3atUwvaoxGI+3atSMqKoqwsLAc3Z09e/YsY8eOJTo6OsPRjZT89ddfLFmyhF27dvHqq68ybtw4OnXq5J4FG5Lu7VIpuKN92YXVCna7+/3WZhKbzcaff/6ZnHSe5IT89ddfgCP6UapUqWcaDxYuXDj7jRPXD7n+MyY6GnkJd/1giu/HC98Pu93Or7/+mlx29tKlS5hMJgoWLEi9evUICgqiXr16ubs6TRawWCzs27ePkJAQfvvtN95++22GDRtGy5Yt3WbxlRNSqSTi4+N5//338ff358CBAy/c8bbb7TRt2hR/f3927dr13MhKly5dksvMZpbVq1ezfPlyfvvttwydt3jxYtasWcOePXto1qwZq1evpm3btpm2459//qF58+aUKlWKHTt25OhObMrKVLVr12bJkiUZim6k5N69eyxatIhvvvmGkiVLMnHiRFq3bu3UTuhZ5sEDqFHDca/LbygUcOUKlCjhakuyjfj4eH777bdU0Y/bt2+j1+sBKFiwYKqk84oVK1KqVCnnRuHE9YN7recygeho5CXc9YMpvh/PvB9///03Z86c4YcffuDs2bNER0ejUqmoVatWctSibNmy7rWoyGHi4+PZsmUL69evJzw8nKZNmzJkyBACAwPdKv8kpVRq3rx5dO3aNdvsEwSBAQMGcP78eb777ruXLmIPHjzIkCFDOHDgADVq1EjzNTExMbz99tvMmTOH3r17Z9q2rVu3MmnSJP7+++8MXf+DBw+oU6cOISEhbNq0CUEQ2Lt3b6btAEeFrc6dO6cr4pMdpIxufPzxx/Tp0yfT3+VffvmF+fPnc+LECSpUqMDkyZNp3Lix+3wHHjyADErm8gQaTZ52Mp6H3W7nzz//TJX78euvv/Lnn38C/5+wnlb0I1OfWXH94F7ruUzgHtuBIiJ5nPj4eM6dO5fsXDx48ACpVEqVKlXo2bMnQUFBVK9eXdRj43DC1q1bx1dffYXZbKZTp04EBwdTpkwZV5v2DCmlUtlRVepp1q1bx9GjR9m4ceNLnQyr1cqCBQto1KjRc50McEir7HY7zZs3z5Jt3t7eCIKAXq/PUDSnRIkSvPvuu+zZs4c+ffowZMgQ7ty5Q9myZTNtS82aNZk9ezaTJk2iUqVKaXZBz05S5m58/PHHHD58ONPRjUqVKrFlyxYuX77M559/Tu/evalWrRpTpkxJVfrUZeTDxXZ+RiqVUrx4cYoXL07Lli2TjyckJHDnzp1k+dWtW7f49ttv0el0ABQoUCBV0nmFChUoU6aM+JuXDxAdDRGRbEDAsdCzm0yMHT6cA3/8gd1up3jx4gQFBTF16lTq1q2Lr6+vq011G27cuMHq1as5dOgQGo2GgQMH0q9fv5zRAWeQnJRKJXHt2jVmz57N0KFDef/991/6+j179vDHH38kV5l5HkeOHKFGjRpZ/jsn5clotdoM/y06duzIp59+yoIFCyhUqBCbN2/OciSiV69e/PLLL0yePJkyZcpQrVq1LI2XUdKqTJWV6EbNmjXZs2cPp0+fZt68eXTq1Il69eoxefJkqlatmg1XICKSfry9valevTrVq1dPPpYU/UiZ9/Hdd9+xdu1awJGwXqpUqWfkV5mOfoi4JXlbOiUmqLkH+SD0meRYmM3m5IfMbkclkbCkQwfeateOevXqZVqvnVex2+2cPHmSkJAQzp8/T7FixQgODqZr165uW94zJ6VSSURHR9O0aVNee+019uzZ89K8A4vFQt26dalcuTLr169/7uu0Wi0VK1bk448/ZvDgwVmy8fLly7Rr147Tp09TqlSpDJ0bFRXFO++8w6xZs/jvv//YsGED169fz7LzZrFY6Ny5Mw8fPuTo0aO88sorWRovszydu7F48WLefPPNTI8nCAJhYWEsWLCAu3fv0rx5cyZOnEi5cuWcaLWISPag1WpTRT+S8j+Sqtb5+/snOx61NBqafvYZUk9PJC5ePxQND2eDry/NX1BK/GXsMBj4NCGB3162sZOHpFN509EQE9TcK5ydRx0Nm93ucCpMJkxmM3a7HYlEgkKhwEOpRCmTIbfZkOSBG4WzMZlM7N27l5CQEO7du0fVqlUZOnQoLVq0yLbOzs5g7969TJw4MdurSqXEbrfTu3dvrl+/zvHjx3n11Vdfes6mTZuYOnUqJ0+efKEEKSmH49KlS5nqX5GS27dv07hxYw4fPpypHfbevXsTExPD6tWree+995g/fz49e/bMkk0AT548oXnz5hQpUoR9+/bh4eGR5TEzizNzN8BRLWjfvn0sWrSIv/76i/bt2zN+/HiKFy/uPKNFRHIAu93O33//nSrx/NatW3j+/jv7IyMxA4JCgVwuRyGXI0/8t1QqJadiHxE2G35SKcosbCzlR0cjb273lyjhWGyLCWp5ntFxccQJAqGJJTk7xsRQUS5nprf3S881CAIfxcVx2mxGKwj8VqgQTaKjGaRWMyiN3XS7IGAxmzElRiysVisACrkcT09PlEolSoXi/3e3LRaw2Zx2rXmBmJgYNm/ezIYNG4iMjOT9999n4cKF1KhRw61D5a6QSiWxatUqTp06xdatW9PlZBiNRpYtW0aHDh1emudw+PBh3n777Sw7GZBaOpUZOnbsyNChQ7FarTRp0oRNmzbRo0ePLH8uChUqxIYNG2jXrh2TJk1i6dKlLvusJeVuzJkzJzl3IyvRDZlMRqdOnWjbti1ff/01S5cu5eDBg3Tr1o0xY8ak6/MiIuJqLBYLCoWCYsWKUaxYsVT5YvpLl/Bo3Bi5VIpFELBaLOhMJuyJe+RSqRS5XJ7aAZHJkEgkWAQBhRO/64XdeBPMncmbjgbkq8V2XiEjTkISGXnt0+w0GLhksXAwIIAAiQRviYSwgADUiTcmAccNMClqYbFYEHD8uCuVSjReXiiVynxdGSq9PHr0iLVr17J9+3bsdjtdunRh8ODBvPXWW6427aWklEotWbIkR6RSSVy4cIH58+czatQoGjZsmK5zNm7cSGRkJOPGjXvh64xGIydOnGDkyJHOMBXvxO9iQkJCps5///338fLyYu/evfTp04eePXvy448/OiW3okqVKixevJgRI0ZQqVIlBg4cmOUxM4tGo2HevHm0atWKsWPH0rhx4yxHNxQKBb1796ZLly5s3LiRL774gp07d9K3b18++ugjChQo4OSrEBHJPB07dqRcuXLIZDL27NlD+fLlmT17NrNmzeLSpUuo1Wrq16/PZ599RoBaDTIZF4FlZjN3rFakEgnVFAqmeXjwmiBgtVp5YDTSzGplvkTCDkHgF2CGUkkXlSrZAUkZ/RAEgcqRkXzu7U3rRClUk6goIu12fipUCIDLZjOdY2P5rVAhPCWSVNKpv2w23ouMZL2vLxsMBn60WCghk7HAx4dqKZQbOwwGFup0RNvtNFAqqZmGqmOTXk+IXs8/djtvSKWM9vKiU6Lsf2ZICL8nJLB582bAURDk008/Zdu2bcm/CbVr12bEiBF07949e96wLCKukERyNT5SKT6Z/HF+ZLNRWiajnFzu2KmQSPAVBASDgZjYWCIiIoiOjkav0yGVSvH28aFgwYIULFgQXx8fVCqV6GS8hB9//JHg4GDq1KnD/v37GTZsGFeuXGH+/Pm5wsnYu3cvzZo1w2KxcOTIkWzr8p0WT548YejQodSqVeulTkMSWq2WlStX0q1bt5fKZ06fPo1er09VOSYrJEV4MutoeHp60qpVK/bs2UP9+vUpVqxY8o+rM+jQoQNDhgzhs88+4+zZs04bN7MkRTc6d+7Mxx9/TJcuXXj06FGWxlSpVAwZMoSLFy8ycuRIvvrqKwIDA1mwYAHx8fFOslxEJOvs3LkTpVLJwYMHmTp1Kp07d6ZSpUocPXqUr776iidPnhAcHJz8ej0QrFZzNCCAXX5+SCUShhkMqNVq/Hx9CQgIQCqVskomY6CXF2FqNbUlEvQ6HTGxsTx58oQnT54QHRNDfEICBqORmjIZ581mAOLsdu7ZbBgFgd8T1QoXLBbekcvxfME9/3OdjiFqNccDAnhLJmNoXBzWxGjLjxYL4+Lj6e/pyfGAAOoolSxLrMKVRJjRyCdaLcFqNacCAujl6cmY+HjOJUr/a1WpwuXLl7ElqiMuXLhAQEAA58+fB+C///7j4cOH7lGB7jmIqyQRt2B0XBwXzGbW6/UUDQ+naHg4D61WxsbH815kJCUiIqgbGcn6xEZBKc/rFxub4fk6xsQQotdz0WLh1f/+o01EBJGRkdSIiGB1fDyCIODl5cW7wDFvb8YClRISCIqJ4ZjJ5JRrzqvY7Xa+++472rdvT+vWrbl58yZz587l6tWrjBs3Lke7NWcWo9HIhAkTGDFiBC1atCAsLCxH8jGSsNlsDB8+HHBIp9Kbt7J+/XoSEhIYPXr0S1975MgRSpcuneHE7eehUChQqVSZlk6BY6fzwYMH3Lhxg169enHgwAFiYmKcYh/A1KlTqVOnDoMHD06u++9KkqIbO3fu5M8//6Rx48aEhoZit9uzNK63tzdjx47l0qVL9OnTJznvZdWqVcnN1kREXEmJEiWYNm0ab731FqdPn6ZSpUpMmTKFUqVKUalSJZYuXcq5c+f44++/AWilVNJSpaK4XE5FhYKlPj7ctlq5+5Q8ebCXFx18fKjo60u5gAAKFS5MoUKF8PfzQ61WI5VKMZvNxMfHU9ls5gedjsioKI7HxlJeIqGmTMZZkwkBOG82U+sl5XeHqtU08fCgpFzOeI2Gv202HibatF6vp6GHB8O8vCgplzNArabBUzliq/V6uqhU9FWrKSmXE+zlRUuVihCjEYD3KldGq9Xyyy+/IAgCFy9eZMiQIVy4cAGA8+fP88orr7h1XpboaIi4BTO9vammUNDD05OfChbkp4IFKSqTUVQqZa2vLz8UKMBYLy/mabUcTPwCZgZBEDCZzSyTyegAVLLbOQYslUpReXggl8nQeHsT4O+PJnGHdolWSxuVipMBATT28GB4XByxWVwI5EWMRiNbt24lKCiIvn37Yrfb2bBhA2fOnKF37954enq62sR08fvvv9OqVSt2797NkiVLWLFiRY7lYySxdOlSzp8/z5dffpnusrNxcXGEhITQu3dvXnvttRe+1mKx8O2339KqVStnmJuMt7d3piMa4JAAFClShD179tCtWzcEQchSt/KnkcvlhISE4OvrS79+/dxm0f10dKNz585Zjm6Ao3rPtGnTuHjxIu3bt2fBggXUqlWL0NBQzIk7uSIiruDtt99O/vevv/7K+fPnKVWqVPKjXr16ADx8/BiABzYbQ+PiCIyMpExEBDUjIwF4/JSjUeWpSqMSQCaV4uHhgcbLCz9fXwoWKECRwoVp7OvLA4kErULBZauVqjYbb1ssnIyP55/wcK6YTNR+SQGb8inmK5KocIhMXB/cs1qp+pQ91Z8a757VSo2njtVQKLiXeF0+Gg0VKlTg/Pnz/PbbbyiVSnr27MnNmzfR6XRcuHCBWrVqvdBGVyM6GiJugU9iJQdPiYTCMhmFZTKUEgnjNRqqKBQUk8no4OlJN5WKbzLgaCTlWWh1OqJjYoh48oSYmBhUJhNqmQy1QkH5woV5q2BBvL29kUilSJ8Kk3b19KRd4k7KFI0GnSBwPT9WNHsO0dHRLFmyhBo1ajBp0iTKly/PN998w4EDB2jevLlbV5F6mr1799K8eXOXSKWS+OGHH1i6dCkTJ07MUDh89erVWCwWPvroo5e+9sKFC8TFxTlNNpWERqPJUkRDJpPRvn17Dhw4gI+PD61bt2bLli1Z3uFPiZ+fH6GhoTx69IjRo0fjLoUXU0Y3/v77bxo1auSU6AZAkSJFmDt3LmfOnKFhw4ZMnz6dOnXqsGPHjuSiFiIiOUnK8uU6nY6mTZty/PjxVI/z588TmOiQ9E5IINZuZ5GPD4cDAjgcEACA5anvrzqd92uJREJllQp/mYxflEp+kslo6u/P+wEB/CyTcd/TE6tEQvWXRDRSughJMzt7G7J27dpcuHCBCxcuEBgYiJ+fH6VLl+by5cuioyEiklVC9XqaRUVR6ckTSkVEsNVo5HEGfnitVitR0dHodDokEgkajcaRZ1GoECoPD2QyGbKX5Fmk3LFQJyaNR4oRDR48eMCUKVOoVq0aK1eupE2bNpw7d45169bleHO0rGI0Ghk/frzLpFJJ/PfffwwfPpwGDRowYsSIdJ8XGRnJ+vXr6d+/f7oiIEeOHOGNN96gYsWKWTH3GbIa0QCHfCoqKorTp0/Tt29fHj58yJkzZ5xkoYNy5crxxRdfcOjQIVasWOHUsbNKUnSja9euTo1uABQrVoxly5Zx6tQpqlatypgxY2jYsCEHDx50qjMnIpIRKleuzJ07d3j99deTu44nPdSensQIAvftdkZ7eVFXqaS0XE6cEz6vEomE9xQKvjWZuGOz8Z5CQSWFAjOwRxCoolCk23FJi9JyOdefcuSvPbVJWVou58pTx65YLJRJsUFXq1YtLl++zJkzZ5I3n2rXrs3+/fv5448/3Do/A0RHQ8SNOWA0MlOr5UNPT7b7+XE8IICuKhXmDOxAyuVyAgICKFy4MP5+fnip1Y7Sdxmw4+nAqQTn71jkJq5evcqAAQOoW7cuhw8fZtSoUVy7do05c+a4tU70eSRJpfbu3esyqRQ4Im/BwcF4eHjwxRdfZKjQwMqVK5HJZMl5HS/CZrMRFhZGy5YtnR6tcYajUaFCBcqWLcuePXuoXr065cuXZ9OmTU6y8P9p0aIF48aNY8GCBRw7dszp42cFLy8v5s6dmy3RDYDSpUuzdu1avv32W958802GDBlCs2bNOHHihNtEeETyD/369SM2NpZhw4bx008/8fDhQ77//nvGjBmDzWbDF/CXSNhqMPDQauWs2cwMJ7UvqK1Ust9opJJcjleioiFQqWSv0fjS/IyXMUCt5pTJxGqdjgdWK6F6PaeeyvEcplaz02hkk17PA6uVNTodR4xGhqRoChgYGIhWq+X48eOpHI29e/dSpEgRSpYsmSU7sxvR0RBxGxSALcWP3GWzmRoKBX3VaiopFBSXy5OTrNKLBBy9LZxrar7DZrNx5MgRPvjgA9q0acPdu3dZuHAhV65cYfTo0fj7+7vaxEzhDlKpJObPn8/169dZs2YNAYmygPTw77//snHjRgYPHoxfYj+ZF/Hjjz/y5MkTp8umIOvSKXDsMnbs2JGjR4+i0+no3bs33333Hf/++6+TrPx/xowZQ7NmzRg+fDi///6708fPKtkZ3QDHTvLWrVvZt28fGo2GXr160bZt2+REUxGRnKBIkSIcPHgQm81Gt27daNSoEZ988gk+Pj5IExf/IRoNN6xWGkRH82lCAtMT+/ZklVoKBbbE/yZRW6nEBi/Nz3gZ1RQKFvn4sF6vp3F0NN+bzYx+ahOruUrFTI2GEL2eBtHRbDEYWOrjk2puX19fypUrR4ECBZKLdwQGBmK32wkMDMySjTlB3uwMLuJepLMz+IT4eG5Zrazx9cVLImGf0cgCrZY1vr4Uk8nYbTTyP72eN2QyjifWhc9Kw75PEhK4ZbWyJ8UiuWZkZKqGfSnrZidRLiKCz7y96fqy5OY80NnTYDCwY8cO1q5dy8OHDwkMDGTo0KE0btw4V5f2NRqNTJs2ja+++opOnToxb948l0Qxkvjuu+/o27cvn376aaqSjulh8uTJfPPNN1y8eDG5l8WLmDFjBvv37+fHH390+ns4cuRI/vrrL/bt25elcR4/fkyNGjVYvnw5LVq0oGrVqgQHBzN+/HgnWfr/aLVaWrdujdVq5ciRI/j4+Dh9Dmdw9uxZxo0bR2RkJNOmTctyV/GnEQSBH374gc8//5wbN24QFBTE5MmTeeedd5w2h4hIhknn+iHPkQfWD0nk3pWCSJ5jqFqNDKifmJPRILGc3ZC4OFpGRxNjt9MnjY7dIs7nyZMnLFy4kOrVqzNt2jTefvttDh8+zN69e2natGmudjJSSqWWLl3K8uXLXepk/PXXX4wcOZJmzZoxePDgDJ376NEjvvrqK4YPH54uJ0MQBI4cOUKLFi2y5T10hnQK4LXXXqNWrVrs2bMHjUZDx44d2bZtG5ZsKMKg0WiSmxwOHz48uV69u5FWdOPhw4dOG18ikdCgQQPCwsJYv349//77Ly1btmTAgAHcuXPHafOIiIjkL8SIhkj2I+5I5Jodid9//501a9awa9cu5HI53bt3Z+DAgRQrVszVpjmFvXv3MnHiRIoWLcratWspV66cS+0xm820a9eO6Ohovv32W3x9fTN0/qhRo/jhhx+4cOFCusoH37x5k2bNmrFjx47k8pHOZN68eRw4cICLFy9meayvvvqKiRMncu3aNaKjo2ncuDFr166ldevWTrD0Wb7//nt69uzJ8OHDmTJlSrbM4SzOnTvH2LFjiYyM5OOPP6Zv375OdxxtNht79+5l8eLF/PXXX3To0IFx48blyjwskVxMFtYPPWJiuPSczYmRXl6MdOEG00vJheuH55F7tyVFREScQlIToD59+hAUFMSxY8cYP348165dY+bMmXnCyUhZVaply5aEhYW53MkAmDlzJrdu3WLNmjUZdjLu3bvHnj17GDVqVLp7lBw5cgRfX99s0/VqNBqnRDQAWrdujUwmY//+/ZQvX57q1as7tVP40zRo0ICpU6fyxRdfcPDgwWybxxnUqVMnOboxbdo0p0c3wFFquHPnzpw5c4a5c+dy9uxZgoKCmDRpEv/9959T5xIRyQ4W+/hwPCAgzUfvXNLXKS8gOhoieZJLZjOlIiKe+xBxlP49ePAgrVq1okOHDjx69IilS5dy+fJlRowYkeGFr7viblKpJL755hs2bNjAZ599RpUqVTJ8/qJFi3j11Vfp0aNHus85fPgwzZs3R5FNkUVnJIMn4ePjw/vvv8+ePXsA6Nu3L2fPnuX+/ftOGT8thg4dSrt27Rg9ejS3bt3KtnmcQVJlql27dvH333/TuHFjNmzY4PQytQqFgj59+nD+/HmmTJnCN998Q2BgIJ999hlRUVFOnUtExJm8IpNRXC5P8+GXi+W/uQ3xLy2SJ6miUDx3J+N4Bir65EV0Oh3/+9//qFOnDkOGDEGj0bB161ZOnTpF165dUWaxpJ878XRVqa5du7qsqlRKHjx4wLhx42jTpg19+vTJ8Pm3bt3im2++YezYsel+v+7du8fvv/+eLdWmkvD29sZisTit63THjh355ZdfuHPnDq1btyYgICBboxoSiYTFixdTqlQp+vXrlysW0jkR3QDw9PRk6NChXLp0iY8++oht27YRGBjIwoULiY+Pd/p8IiIieQPR0RDJk6gkkufuZBRP0YAvPxEeHs68efOoVq0aM2bMoHr16nz77bfs3LmTRo0aucUC3Fm4q1QKHLYNGjSIwoULs2jRokz93RcsWEDx4sXp3Llzus85cuQIarWaoKCgDM+XXpIS0p0ln2rUqBG+vr7s3bsXpVJJt27d2LFjBwaDwSnjp4WnpyehoaEYDAaCg4OzJQHd2eRUdAMc7/G4ceOS5ZZffvkl7733HqtWrcrW90VERCR3IjoaIiJ5nDt37jBmzBhq1qzJhg0b+PDDD7l48SKrVq2ici5PMksLd5VKJTF9+nTu37/PunXr0GSiFvy1a9c4duwYEyZMQJ4Bp/nIkSM0adIEDw+PDM+ZXpKux1mOhlKppG3btuzduxe73U6vXr1ISEjI9hyK1157jfXr13P58mU+++yzbJ3LmSRFN7p168a0adPo1KlTtkQ3AAICApg2bRoXL16kXbt2zJ8/n1q1ahEaGpornDMREZGcQXQ0RETyIIIgcPbsWXr27EnDhg354YcfmDRpEteuXePTTz/ltddec7WJ2YK7SqWS2L17N9u2bWPevHmUL18+U2PMnz+fcuXK0bZt23Sf8+eff3Lz5s1slU3B/0c0nJWnAQ751OPHj7l8+TJvvvkmDRo0YOPGjU4b/3m89957zJo1iw0bNrBjx45sn89ZeHl5MWfOHHbt2sU///yTrdENcDRbmzdvHmfOnKF+/fpMnz6dOnXqsGPHDqxWa7bMKSIiknsQHQ0RkTyExWJh3759NGvWjC5duvDvv/+yYsUKLl26xLBhw9y2GVlWMRgMjBs3zi2lUkncvXuXSZMm0aVLF7p27ZqpMc6dO8fZs2eZMGFChsqZhoWFoVQqadSoUabmTS/Olk4BVK9enTfeeCM5KbxPnz78/PPP/Pzzz06b43n07t2bHj16JJfZzU3UqVOHEydO5Eh0A+DNN99k+fLlnDx5kipVqjBmzBgaNWrEN998k21OjoiIiPsjOhoiInkArVbL2rVrqV27NsOHD6dAgQJs376d48eP06lTp2yrMuQOJEml9u3b55ZSKQC9Xs+gQYMoVqwYc+fOzVSURRAE5s+fT5UqVWjevHmGzj1y5Aj169fPlFQrIySN78yIhkQioWPHjnzzzTeYTCYaN27Ma6+9xqZNm5w2x4vmnjt3LlWqVGHgwIGEh4dn+5zOJKejGwBlypRh3bp1HD16lDfeeIPg4GCaN2/OiRMnENt2iYjkP0RHQ0QkF/Pff/8xe/ZsqlevzuzZs6lduzbHjx/n66+/JigoyK1kQ9lBklTKarW6pVQKHA7C5MmTefz4MevWrUOdye72J0+e5OrVq0yaNClD1xgeHs7Vq1ezXTYF2RPRAOjQoQPx8fGcOHECmUxGz5492b9/P3FxcU6dJy0UCgXr169HIpEwYMAATCZTts/pbHI6ugHw9ttvs23bNvbu3YtaraZXr160a9fOKc0cRUREcg+ioyEikgu5ffs2o0aNombNmmzZsoWePXty6dIlli9fToUKFVxtXraTUirVqlUrt5RKJfH111+ze/duFi5cSKlSpTI1RlI0o2bNmtSvXz9D53777bdIpVKaNWuWqbkzgoeHB3K53OmORqlSpahSpUqyfOrDDz/EarWya9cup87zPAoXLsyGDRv45ZdfmDx5cq7cmU8ruvG///0v22VNgYGB7Nu3j23btmE0GunQoQMffvhhjkjfREREXI/oaIiI5BIEQeD06dN8+OGHNG7cmHPnzjFt2jSuXr3KtGnTePXVV11tYo6QG6RSSdy6dYupU6fSu3dv2rdvn+lxjhw5krzIzWjE5vDhw9SuXRs/P79Mz59eJBKJU5v2paRjx44cP36c2NhYChcuTMuWLdm0aVOOLfrfeecdFi5cyI4dOwgNDc2RObODlNGN6dOn07Fjx2yPbkgkEho2bMjRo0dZt24djx8/pkWLFgwYMIA7d+5k69wiIiKuRXQ0RETcHIvFwu7du2nSpAndunUjKiqKL7/8kgsXLjB48OBkuUp+YM+ePc9IpdyV+Ph4Bg0aRJkyZbJUItVms7FgwQKCgoIIDAzM0LmxsbGcP38+R2RTSXh7ezs9ogHQtm1b7HY7hw4dAhyJ2vfv3+fcuXNOn+t5dO7cmcGDB/Ppp5/m6LzOJim6sXv3bv79918aNWqUI9ENiURCq1atOHXqFMuWLeOXX36hUaNGjBw5kkePHmXr3CIiIq5BdDRERNyU+Pj45GZYI0eO5NVXX2XXrl18++23tGvXLk8neD9NklTqo48+cnupFDiiT+PHjycqKoq1a9dmqXfF/v37uXfvHpMmTcrwud999x12uz3DyeNZQaPRZIujUahQIYKCgpLlU7Vq1aJ06dLZ2ik8LaZNm0bt2rUZPHgwf/31V47O7Wxq167NiRMn+PDDD3MsugEgk8no0qULZ8+eZc6cOZw5c4Z69eoxefJk/vvvv2yfX0REJOcQHQ0RETfj8ePHfPbZZ1SvXp358+dTv359Tp48ydatW6lTp47bJTtnN7lJKpVEaGgohw4dYunSpRQvXjzT41gsFhYtWkSzZs2oWrVqhs8/cuQI1apVo0iRIpm2IaN4e3tni3QKHPKpS5cu8ddffyGRSOjTpw9Hjx7N0WpQcrmckJAQvL296devH3q9Psfmzg5cFd0AR6J93759OX/+PJMnT+bgwYPUqlWLmTNnEh0dne3zi4iIZD+ioyEi4ibcvHmT4cOHExgYyNdff02/fv24fPkyS5cudevd++wkSSpls9kICwtza6lUEtevX+ezzz5j0KBBWZYs7dy5kz///JOJEydm+FytVsv3339Pq1atsmRDRsku6RRA8+bNUavV7N27FyC5dPNXX32VLfM9D39/f0JDQ3n48CFjx47NlcnhT+Oq6AaAp6cnw4YN4+LFi4wYMYKtW7fy3nvvsWjRIuLj43PEBhERkexBdDRERFyIIAicPHmSLl260KxZM65evcqMGTO4du0aU6ZMydGdaHciLalU2bJlXW3WS4mNjSU4OJjKlSszbdq0LI1lMplYsmQJbdq0yVQX8ZMnT2I2m2nRokWW7Mgo2ZUMDqBWq2nRogV79uxBEAR8fHxo3749W7duzfEu1OXLl2f58uUcPHiQVatW5ejc2UXK6MZ///1Ho0aNWL9+fY413PPx8WHcuHFcvHiR3r17s2rVKgIDA/nyyy8xGAw5YoOIiIhzER0NEREXYDab2bFjB40aNaJnz55otVpCQkI4d+4cAwYMcHtpUHaSllQqs70nchK73c6oUaPQarWsWbMmyzk0W7ZsITw8nPHjx2fq/CNHjlCpUiWKFSuWJTsySnZGNMAhn/r999+5efMm4OgU/u+//3LixIlsm/N5tGrVitGjRzNv3jyOHz+e4/NnFymjG5988kmORjcAAgICmD59OhcuXKBt27Z8/vnn1KpVi9DQUCwWS47ZISIiknVER0NEJAeJi4vjiy++oGbNmowZM4ZixYqxd+9eDh8+TJs2bZDL5a420aXkRqlUEiEhIRw7dowVK1bw2muvZWksvV7PihUr6Ny5M2+99VaGzzeZTJw4cSJHq00lkd2ORt26dSlUqFByUnjlypWpWrVqjnQKT4vx48fTpEkThg8fzv37911iQ3agVquZM2cOe/bscUl0A+CVV15h3rx5nDlzhqCgIKZNm0bdunXZuXMnNpstx+wQERHJPKKjISKSA/z5559Mnz6datWqsXjxYpo2bcrp06fZtGkTgYGB+S7B+2lyq1QqicuXLzNv3jxGjBhBkyZNsjzehg0biIuLY+zYsZk6//Tp0+h0uhzPz4DslU6BIxm7Xbt27N+/P1ku1bt3b77//vsc3XVPQiqVsnLlSl555RX69u2b53IKatWq5dLoBsCbb77JihUrOHnyJJUrV2b06NE0bNiQQ4cO5ajjIyIiknFER0NEJBux2WzMnDmT2rVrs2fPHoKDg7ly5UqWukTnNXKrVCqJqKgogoODqV69eqaStp8mPj6eVatW0aNHD954441MjXHkyBFKlSpF6dKls2xPRsnuiAY45FNPnjzh7NmzgKPHhq+vL1u2bMnWeZ+Ht7c3oaGhPHnyhI8++ijPLX7dIboBULZsWdavX09YWBivv/46gwcPpkWLFpw8eTJPJOSLiORFREdDRMTJCDikK3Hx8ej1eu7evcvs2bO5evUqEyZMoFChQq420W3YvXt3rpVKgcORHDFiBFarlZCQEKdI39auXYvRaGTUqFGZOt9isfDtt9+6RDYFjkW3wWDI1uTsypUrU6pUqWT5lEqlolu3bmzfvh2TyZRt876IkiVLsnr1ao4fP87ChQtdYkN24w7RDYAqVarw1VdfsXfvXjw9PenZsyft27fn4sWLOW6LiIjIixEdDRERJyEAeoOBqKgoYmJjEQQBlacnmzZtom/fvrlqlz67MRgMjB07lpEjR+ZKqVQSK1as4PTp03z55ZdOqRAWHR3NmjVr6NevX6bHu3TpErGxsS5zNDQaDQA6nS7b5pBIJHTs2JEjR44kz9O7d29iYmL45ptvsm3el9GwYUOmTp3K8uXLkzuY5zXcJboBEBgYyL59+9i6dSt6vZ4OHTrQvXt3bty4keO2iIiIpI3oaIiIZBG7IKDV6Xjy5Anx8fHIZTICAgLw8/VFIZcjk8lcbaJbkSSV2r9/f66USiVx9uxZFi1axNixY6lXr55Txvzyyy8BGDFiRKbHOHLkCK+//jqVK1d2ik0ZxdvbGyDb5VPt27fHYDDw7bffAlCiRAnq1avnsqTwJIYNG0bbtm0ZNWoUt2/fdqkt2UlSdKN79+588skndOjQwSXRDYlEQqNGjTh69Chr167l77//pnnz5gwcOJC7d+/muD0iIiKpER0NEZFMYrXZiE9I4MmTJ+h0OlQqFQULFsTPzw9lFkub5lVyu1QqifDwcIYNG0a9evUYPXq008bcsGEDgwcPJiAgIFNj2O12wsLCaNmypcsKDCRFNLLb0ShWrBg1a9ZMlk+Bo9TttWvXuHXrVrbO/SIkEglLliyhZMmS9O3bN093uFar1cyePZs9e/YQHh7u0uiGVCqldevWnDx5kmXLlnHjxg0aNmzIyJEjefToUY7bI+JkrFawWPLPI4f7AmUnoqMhIpJBzBYLsXFxREZGYjQa8fLyolDBgvh4eyMXoxdpklIq1bp161wrlQKwWq0MGzYMmUzGypUrnRaxWrFiBUqlkuDg4EyP8eOPPxIeHu4y2RT8f0QjOytPJdGxY0d++OEHnjx5AsD7779PkSJFXB7V8PT0JDQ0FJ1Ox5AhQ3K8mWBO4y7RDXBUJevSpQtnz55l9uzZnD59mnr16jFlyhTCw8NdYpNIFtBoQKEAux3M5vzzsNsd1524cZObkQhiqQaR7ObmTahbF6RSyKV9IgQcTfaMBgMWqxWZTIbK0xOVUvn8nWOr1XGzOHsWXCRjcQfu3btHcHAwDx8+ZN68ebk2ipHE559/zqpVq9i9ezfvvfeeU8b8+++/qVOnDuPHj+ejjz7K9DgzZ85kz549/Pjjjy6T7P3333+8++67bN26lUaNGmXrXLGxsVSpUoXp06czcOBAABYvXszq1au5fv16stPjKi5cuEDXrl3p27cvM2fOdKktOcWFCxcYO3Ys4eHhTJ06lf79+yOVum5P02AwEBoaysqVKzEYDPTv35/hw4dnOmoo4gIePIAc2LhwOzQaKFHC1VZkGdHREMl+HjyAGjUc4cBchiAIWCwWzBYLgt2OTCZDqVQik8tJlzBFoYArV/LEzSIz7N69m0mTJvH666+zdu3aXBvFSOLkyZP07NmTadOmMWzYMKeNO3bsWI4fP86FCxcy3RVeEARq1apFgwYN+Pzzz51mW0bRarWUKVOGkJAQ2rRpk+3z9e/fn3///ZewsDDA4ejUqFGDmTNn0q9fv2yf/2WEhoby8ccfs2zZMrp06eJqc3IEvV7P3Llz2bBhAzVr1mTp0qWUcPE9MD4+nrVr17JmzRokEgnBwcEMHjzY5c6oiEheR3Q0RHKGXLYjERMTw8GDB9m/fz8JCQnUCwqic6dOVKhQIWMD5ZEdiYxiMBj4+OOP2b59O126dGHu3Lm5MuE7JY8fP6Zp06bUqFGD0NBQp+3S/vHHH9SvX59PPvmEQYMGZXqcX375hffff5/t27cTFBTkFNsyg91u54033mDBggX06NEj2+c7fPgwgwYN4vTp08m9aQYOHMj9+/c5efKky5thCoLA+PHj2bNnD/v27aNq1aoutScnSRndmDJlCgMGDHBpdAMcfW9WrVrFhg0bUKvVjBgxgn79+uHp6elSu0RE8iqioyEikoI//viDNWvWsHPnTqRSKR9++CGDBg3izTffdLVpuYZ79+4xePBgHj16lCekUuDoTdGhQwf+++8/jh07hp+fn9PGHjZsGBcvXuTChQt4eHhkepwFCxawYcMGbt68icLFxQjKli3LmDFjGDJkSLbPZTKZqFKlCv37909umHjmzBm6du3Kvn37nCZvywpms5mOHTvy999/c/ToUaeUQs4t6PV65s2bx//+97+sRzecuGEVGRnJlq1bOXL4MH5+fvTo0YNWrVq5/LuTJvl0w0okbyA6GiL5HkEQuHLlCqtXr+a7776jYMGC9O/fnz59+jh1QZkfyGtSqSRmzJhBaGgo+/bt491333XauLdv36ZJkyZ8/vnn9OrVK0tjNWjQgCpVqrB8+XInWZd5qlWrxocffsj48eNzZL7x48dz5swZLl68iEQiwW63U69ePapUqZJcMtjVhIeH07x5c15//XX27NmDUql0tUk5ysWLFxkzZkzmoxvZJMG12+2YTCasVisSiQQPDw/kcrnLI2GpyOcSXJHcjVh1SiTfYrPZOHToEB988AHt2rXjjz/+YNGiRVy+fJlRo0aJTkYGSFlV6oMPPsjVVaWeJiwsjLVr1/LJJ5841ckAWLhwIcWKFaNbt25ZGuf333/n7t27Lq02lRJvb+9sL2+bko4dO/LXX39x9epVwFHqtHfv3hw+fDi5IpWrKVKkCP/73/+4ceMGU6ZMIb/t8QUGBnLixAl69OjBp59+SocOHXjw4EH6B9BqHU6GVApKpdMeUpUKT19fVH5+2BUK4oxGYg0GTIDgxHky/ZBKHdedi6THIiIpER0NkXyHXq8nNDSUunXrMnjwYFQqFZs3b+bUqVN8+OGHWZKv5Efu3btHy5YtkxvwLVu2LNfnYyTx8OFDxowZQ6tWrejfv79Tx/7pp584evQoY8eOzbJc48iRI6jVaurXr+8k67KGRqPJUUfjvffeo2jRoql6anTt2hWpVMr27dtzzI6X8e6777JgwQK+/vprNm7c6Gpzchy1Ws2sWbPYu3cv4eHhNG7cmHXr1mGz2dI/iFzu2OF38kOhUuETEIBvgQIglxOj1RKdkIBJEBCyYb50P3JppUYRkSTET7BIviEiIoLQ0FA2btyIVquldevWhISEUKVKFVeblmvZtWsXkydP5o033nBdFCObCg2YzWbmffQR1T08WNq/P5JffnHq+F9PnUqzokXpULq0owR0Rkmh2z5y5AiNGzdGpVI51cbMotFocqSPRhJSqZQOHTqwZcsWZs2ahUKhwM/Pj3bt2rFly5bkvifuQNeuXbl16xaffvop5cqVo1atWq42KcdJim7MmzePTz/9lEOHDrFs2TKXV6YCUCgU+Pv7Y7ZY0CYkEBMTg1KpRKPRiI1YRUQygZijIZLnuXv3LmvWrGH37t0oFAp69OjBwIEDeeONN1xtWq7FbapKZWPpZKPRiMViQa1WO32RarXZMOj1qFSqzEczEnXbfysU1KxZk9WrV9O2bVun2plZBg0ahE6n46uvvsqxOe/cuUPDhg0JDQ2lWbNmAFy/fp1WrVqxefNmmjRpkmO2vAyr1Ur37t359ddfOXr0KK+//rqrTXIZFy9eZOzYsfz777/JuRtpft+S+jEplY7Pfg6Q1D9Jm5CAxWrFw8MDjUaDIiejDBaLo4FbPu/HJJJ7EaVTInkSQRA4f/48vXv3pkGDBpw8eZKJEydy7do1PvvsM9HJyAJJuQBuIZXKJt22EdBaLMg1GmSenk4dW1Aq0Vks2GQy5F5eWdZth4WFoVAosr05XkbIaekUOCpdVaxYMZV86p133qFy5cps3rw5R215GXK5nJCQELy8vOjfvz8Gg8HVJrmMwMBAjh8/Ts+ePZkxY0bGczeyEQngoVQSUKAAfn5+2Gw2oqKiiI2Ly/Pd3kVEnIXoaIjkKaxWKwcPHqRly5Z06tSJv//+m2XLlnH58mWGDx+Or6+vq03M1ezatYsWLVogCAJhYWHuU7rWibptq1RKnE6HwtMTlUbjdM21WRAwWK14+vggcYJu+8iRIzRo0MCtGo95e3vnqHQqiY4dO/Ldd98RHx8PgEQioU+fPpw4cYK//vorx+15EQEBAWzcuJH79+8zduzYfJccnpKUuRtPnjyhUaNGrF27NmO5G9mIBFB5eFCgQAF8fX2xWCxERkURFx+P1U1sFBFxV0RHQyRPoNVqWbduHbVr12bIkCH4+Pjw1VdfceLECbp06eKetdFzEUlVpUaNGkWbNm04cuRInqkqlRJBEIiNjUUmk+Ht45O+7u8ZGR/HZ1WhUKB0QtGBqKgoLl++7DbVppLI6apTSbRv3x6LxcLhw4eTj7Vr1w5vb2+2bt2a4/a8jPLly7Ns2TIOHDjA6tWrXW2Oy3Hn6AY4HA5PlYqCBQvi4+2NyWQiKiqK+IQEbHa7q80TEXFLREdDJFcTHh7OnDlzqF69OrNmzeK9997j2LFj7NixgwYNGrhXLfRcytNSqaVLl+aZqlJPE5+QgM1mw8/XF2k2fHZMJhMWiwWNRuMUJ+b8+fNIpVKaNm3qhNGcR04ngydRpEgR6tatm0o+pVar6dSpE1999RVmsznHbXoZH3zwASNHjmTOnDmcOnXK1ea4HHePboDD4VCr1RQsWBCNlxdGo5HIyEgStFrs+TgyJSKSFqKjIZIruX37NqNHj6ZmzZps2rSJHj16cPHiRb744gsqVqzoavPyDG4rlcoGDEYjBoMBHx8f5NmQ7JkUzVAqlXg4qVnbmTNnqFWrFgEBAU4Zz1kkRTRcIQfq2LEj58+f5/Hjx8nHevfuTVRUFEeOHMlxe9LDxIkTady4MUOHDuWPP/5wtTluQVJ0o1evXnz22WeMGTMGu5tFDaQSCV5eXhQsWBAvtRq9Xs+TJ0/Q6nSiwyEikojoaIjkGgRB4OzZs3Tv3p3GjRtz5swZpkyZwtWrV5k+fTpFixZ1tYl5BoPBwJgxY/K8VCoJi9VKfHw8np6eeGZTiViT0YjVakWj0ThlPEEQuH79utvJpsDhaAiCgF6vz/G5W7ZsiUqlYv/+/cnHypQpQ+3atdm0aVOO25MepFIpK1eupFChQvTr188lsjN3RK1WM3PmTPbu3Ut0dDQ6nQ690Yi7LeGlEgkajYZCBQui9vREp9MRGRmJTq/P17k3IiIgOhoiuQCLxcLevXt5//336dKlCxEREaxcuZKLFy8m52OIOI8kqdSBAwfyvFQKwC4IxMXFIZPJ8MmmhOqkaIaHh4fTavFbrVZsNhstWrRwynjOJMmZcoV8SqPR0Lx5c3bv3p1qkde7d28uXbrEb7/9luM2pQcfHx82btzIf//9x4gRI9xu996VvPfee6xfvx6FQoFepyMmOtotk7ClUine3t4ULFgQlYcH2oQEIiMj0RsMbucciYjkFKKjIeK2JCQkEBISQmBgICNGjKBw4cLs3LmTY8eO0aFDBzHBOxvIT1IpcDgA8fHxjrwMP79sy+kxGgxYbTanRTPA4WhUqFCBIkWKOG1MZ5FUActVO/MdO3bkzp073L59O/lYixYtKFSoEFu2bHGJTenhrbfe4ssvv+T48eMsXrzY1ea4FSqVCpVKha+vLza7naioKEfEwNWGpYFMKsXHx4cCBQuiVCqJj48nMjISgxtGY0REshvR0RBxO/755x9mzZpF9erVmTdvHkFBQZw4cYJt27ZRt25dMcE7G8hvUqkkDAYDRqMRXx8f5NnUOVoAtDqdozmfk3I/7IKA1WqlXr16ThnP2bja0QgKCqJAgQKpksIVCgXdu3dn165d6HQ6l9iVHho3bszkyZNZunSp2+aUuBKFXE7BAgXw9PQkISGBaDeNbgDIZTJ8fX0pUKAACrmcuLg4oqKiMJpMosMhkm8QHQ0Rt+HWrVt89NFHBAYGsm3bNvr06cPly5dZunQp5cuXd7V5eZYkqdTBgwdZtmxZnpdKJWGxWklISECtVqPKprwMAINej83J0QxLYid0d3U0XCmdAodT0bZtW/bu3ZuqWlHPnj3R6/Xs3bvXJXallxEjRtCmTRtGjhyZKioj4kAikeDj7U1AQAB2N49ugMM58vPzo0BAAFKplNjYWKKjozGZzW5rs4iIsxAdDRGXIggC33//PV27dqVp06ZcunSJTz75hKtXrzJ16lS3lIXkJZ6WSnXp0sXVJuUI9sR+GXK5PFsb3QmCgFanw9PT06kRE5PZjFQqddsCCK6OaIBDPhUeHs6FCxeSj7322ms0btyYTZs2uXWSrkQiYcmSJRQvXpx+/foRExPjapPcEqVCkWuiG+BwgAP8/Qnw9wcgJiaGmJgYzIkbByIieRHR0RBxCRaLhZ07d9K4cWO6d+9OXFwcISEhnD9/noEDBzp191fkWdKSSpUpU8bVZuUIAhAfF4dgt+Pr6+v0pnwp0RsMCHY7Xl5eThtTACxmc7aU4HUWSdfrSkfjnXfeoUSJEqnkUwB9+/bl119/5ccff3SRZelDrVYTGhpKQkICwcHBWK1WV5vkluS26AaAUqkkICAAfz8/BLud6OhoYmJjsYjvsUgexH1/qbKDBw/ARaF8l6LRQIkSrrYCcCTebtmyhfXr1xMeHk6TJk2YPXs2tWrVEnMvcoi7d+8SHBzMn3/+ybJly/JNFCMJvV6P0WTCz88v2/IywBE10WVDNMNsMiEIgls7GgqFApVK5TLpFDgWoB07diQkJIS5c+fi6ekJQP369XnzzTfZtGkT1apVc5l96eGNN95g7dq1dOvWjdmzZzNjxgxXm+S2JEU3ErRaEhISHLlXvr7Z+h3PChJwVKHz8MBkNKLV6YiKikKlUqHRaNzWbhGRjOK+v1TO5sEDqFED8mOIUqGAK1dc6mz8/fffrFu3Lrk7b6dOnQgODs43u+juwq5du5g8eTJvvPEGYWFh+e7vb7ZY0CYk4OXlhcrDI1vn0ifW0PdycnTOaDIhk8mQuvlCJKlpnyvp0KEDixYt4tixY7Rp0wZwlCDt1asXCxYs4LPPPsM/UcbirtSpU4fPPvuMadOmUbFiRTp37uxqk9yWpOiGSqVKTrzWaDSo1epsjVxmBQmOiloeKhVGoxGtVktkZCSenp54eXnlo0WaSF4l/3yGtVqHkyGVghvvBDodq9Vx3S7aWbxx4warV6/m0KFDaDQaBg4cSL9+/ShcuLBL7MmvGAwGpk6dyo4dO+jatStz5szJFwnfKbHb7cTFxSFXKLJdmmcXBPSJ0QyZ1HkKVQEwmUx4KZVuu3BKQqPRuDSiAVC8eHGqVavGnj17kh0NgK5duzJ//ny2b9/O0KFDXWhh+ujXrx83b95kwoQJlC5dmnfeecfVJrk1uS26AQ6HwzOxhK9Br0er02E0GvFSKlHLZKLOXSTXko9W3InI5Y4d/vyE2Zyj09ntdk6ePJmcc1GsWDFmzpxJ165d893i1h3I71IpcCzQ4+LjEQTB0S8jm+fT6XQI4NTcDACz2Yzdbkfp4QFunPQK7hHRAEdS+CeffEJUVBQFChQAoECBAnzwwQds2bKF4OBgpE50BrMDiUTC/PnzuXfvHv379+fo0aPiZs1LyI3RDXA4HGq1GpWnJwa9HqNOh2C389WaNXSZNcvtI3AiIk/j3ndXkVyF2Wzm66+/pmHDhvTu3RuDwcDatWs5d+4c/fr1E50MF5Bfq0o9jU6nw2Qy4evr69QIQ1rY7Hb0ej1qtdrpc5kSZVPuvDObhLs4GkmRjIMHD6Y63qdPHx4+fMjp06ddYVaGUSqV/O9//0MQBAYNGpRc4ljkxeS2ylRJSCUSvLy88PfzQ6lUcuDAAQIDA1myZIlbfK9ERNKL6GiIZJnY2FhWrFhBjRo1GD9+PCVLlmT//v0cOnSI1q1bI8sFi6K8hl6vz7dVpZ7GbLGg1WrReHnhoVRm+3w6nQ4Jzo9mCIDJaMTDw8Otd2STcAfpFEBAQAANGzZ8pndG9erVKV++PJs2bXKRZRmnSJEirF+/np9++ompU6e6dYledyI3VqZKQiqR4OHhwbZt2/jwww9ZsWIFgYGBhISEYDQaXW2eiMhLER0NkUzz6NEjpk2bRrVq1ViyZAnNmzfn9OnThIaGUrNmTbGKlIvIrw340sJmtxMbG4tSqXR6Uvbz5jMYDKi9vJA6+fNvsViw2e3Z2lzQmbhLRAMc8qlr167x8OHD5GMSiYQ+ffpw7Ngx/vnnH9cZl0GqVavG/Pnz2bZtG5s3b3a1ObmK3BrdAPD392fGjBlcuHCBVq1aMWfOHGrVqsXmzZvF6JaIW5P/cjREssyPP/7ImjVrOHz4ML6+vgwbNow+ffpQsGBBV5uW79m5cydTpkzJt1WlUiIAcXFxANneLyMJnVaLRCLJFsfOZDIhlUpRKBS5onqeOzka77//PhqNhj179jBu3Ljk4x06dGDWrFls27aNCRMmuNDCjNGtWzdu3rzJ9OnTKVu2LIGBga42KdeQ2dyNHQYDnyYk8FsWc2M6xsRQUS5nZiYbhb766qssWLCAYcOGsWjRIqZMmcKXX37J+PHjad++fc4qCMSWASLpQHQ0RNKF3W7n+PHjrF69mkuXLlG8eHHmzp1L586dk+vTZzviTe256PV6Pv7443xdVeppdDodZrOZAH//bM/LALDabBgMBjTe3k6PZgiA0WhElUtkU+A+0ilwlA9t1aoVe/bsYezYscnRVo1GQ6dOnfjqq68YPXq0w4nLJcyYMYM7d+4waNAgjh49ymuvveZqk3IVGa1M1ValonE2l8TOCMWLF2flypUMHz6chQsXMnLkSFauXMnEiRNp0aJF9isKxJYBorORTkRHQ+SFGI1Gdu/eTUhICH/88Qc1atRgw4YNNG3aNOd3TsSbWppP3717l8GDB/PXX3/l26pST2Mymx15GRoNyhzIywCHYyOVSlFng+NttVqx2Wx45BLZFLhXRAMc8qkdO3Zw/fp13n333eTjvXv3ZtOmTXz77be0bt3ahRZmDIVCwdq1a2nevDn9+/dn//79ObfpkwuwCAKKlyy2U0Y34l8S3VBJJKjcUA5cvnx5NmzYwPXr1/n8888ZOHAgVapUYdKkSdSvXz/7HA6xZYCrLck1iDkaImkSHR3NkiVLqFGjBpMmTaJ8+fJ88803HDhwgObNm+d8gnfKm5pSmX8eUukLb2o7d+6kRYsWAPm6qlRKbIn9Mjw8PJyekP08rFYrBoMBL40mW37YTSYTUokkx5wmZ5AU0XCXhOVatWrxyiuvsGfPnlTHy5cvT40aNXJVUngSAQEBhIaGcu/ePcaPH+82f+vMYrfb+fLLL6lduzZvvvkm1atXZ/ny5QDMXruWOjodJaOjCYyMZIFWiyXF9S7WamkSFcVXBgPvRUZSPCLihXMdM5koFxGBTRBQKhT85+PDO3Y7s+Ljk3M3xsXHMyJRfrnDYKBcijGT5tttMFAzMpKyEREMiYtDa7cnv0YvCIyMi6NURATvPHlCiE73jB1xdjsj4+IoFxFByYgIesTE8MBqBUAQBCrFxHAoxQZbkyZNUvVRuXz5Mm+++SblypVjx44d7Nq1C7lcTvfu3enYsSOXL1/OwDuQCZJaBuSXR35yqpyE6GiIpOLBgwdMmTKF6tWrs3LlStq0acO5c+dYt24d1apVc7V54k0tkaSqUqNHj873VaVSIgBxsbFIJBJ8fXye2ZXsGBPDJ1ncZf/daqV1dDTFIyJoEhUFwI2EBHoLAhUSEpKPOROj0YiHSpVrZFPgiGhYLBbMOdzH53nIZDI6dOjAgQMHnkme7dOnD+fOneP33393kXWZp2LFiixbtox9+/YREhLianOyxNy5c/niiy8YPXo0P/zwA19++SWFChUCQOPpyXKVih98fZnl7c02g4F1en2q8x/abBw2Gvmfry/HAwJeONd7CgVaQeCXxEX9RauVAjIZN+RyhMTKVOeMRmq9wLl/ZLNx1GRis58fm/38uGg2szKFTbMSErhgsbDRz4+v/fy4YLFw86nP3qj4eH62Wtnk58chf38EoEdsLBZBQCKRECiXcyExYT0uLo579+5hNBqTP6sXLlzgnXfeSY5m1alTh4MHD7J582bi4+Np164dPXv25ObNm+l4B0REnI/oaIgAcPXqVQYMGEDdunU5dOgQH330EVevXmXOnDkUL17c1eaJpECsKvV8tFotFosFX1/fbGvCtkinQy2RcLZAAXb5+2OxWllmNOItlycfcyZWmw2r1YqHG+nD04N3YrKru8mnoqOj+f7771Mdb926NQEBAbm2ilObNm0YMWIEc+bMeebacgtarZb169czffp0unTpQvHixalZsybdu3cHYHSvXlSXyXhDJqOphwdD1GoOmkypxrAAK3x9qaRQUP4l+TY+UikV5XLOJzrC581mBqvV3LLbUfn7k+DhwUObjfJ6/XMrU9mBZT4+lJPLeU+ppJNKxdnE8XR2O18ZjXyi0VBXqaS8QsFyHx+sKc5/YLXyncnEYm9v3lMqqaBQsMrXl//sdo4mXltthYLzifNfvHiRypUrU6tWLc6fP++w+/x5atWqlcouiURCkyZN+O677wgJCeHhw4c0a9aM4ODgXOlMi+RuREcjH2Oz2Thy5AgffPABbdq04e7duyxYsICrV68yZswYAl6yIySS84hSqedjMpnQ6XRovL1RvmSRkRUe2mzUVCh4XSbDXypFq9Xyt0RCoEqVfCwzWJ4jezEZjUgkkhzpAeJMNInlhN0lIRwcMqny5cs/I59SKpV8+OGH7Ny5E4PB4CLrssakSZNo0KABQ4YMSVXGN7dw7949zGYzdevWTfP5g6dO0Uavp0pMDKUiIpiv0/H4KQfgNamUAhn4/tVSKrlgsSAIApcsFlp6eFBaJuOK1cpNDw9ekckoJghERUVhSiMy94ZMhibFfIVlMiITpVOPbDYsgsC7Ke5FflIpb6WQHd+z2ZADVVO8xj/xNfcSIy215HLu2u1ExcZy4cIFatWqRe3atTl//jwWi4WrV69Su3btNK9PKpXSpk0bvv/+e5YsWcK1a9do0KABY8aM4a+//kr330lEJCuIjkY+xGAwsHHjRurVq8fAgQNRKBRs2rSJH374gR49euS6ndP8gMFgEKVSL8BmtxMXH4+Hh8dLozs2QeDj+HjKRkRQ8ckTFqTIIygaHs7Rp5pglYuIYEfi4rNoeDg3LBaW6HQUDQ9nflwc5Q0GbgNLE48tTlxY/2OzEZyovS4fEUHf2Fj+SrEwGh0XR7/YWJbrdFR98oS6z5FcGU0mR5M+N0xEfRHuGNEAR1Tj22+/fcauXr16kZCQwIEDB1xkWdaQyWR8+eWXFCxYkL59+7qVg5ceXtQf5tq1awyfO5fGMhmbvb05FhDAKLX6GedcncHvSG2lkstmM79arSiAUnI5tZVKzpvNXDCbqe3hQYECBVB7eqYZIX1a2CrBEeVwJuVkMvwkEi78/DMXLlygdu3a1K5dm4sXL/Lzzz9jtVqpXr36C8eQy+V069aN8+fPM3PmTE6cOEHdunX5+OOPCQ8Pd7LFIiKpER2NfMSTJ09YuHAh1atXZ9q0abz99tscPnyYvXv30rRp02yTmohkDZvdzrBhwzh48CDLly8XpVJPIQgCCQkJjryMdPTL2Gk0IpNIOBIQwCyNhjV6PV+ls8PuTwULUlYuZ4hazU8FC9LdZuOUQkE5hSL52JDEBdCHsbF4SSTs8/fnYEAAXhIJ3WNiUi2OzprN3Lda2e7vz2Y/v2fms9ntWCyWXOn8u6uj0a5dO8xmM0eOHEl1vFixYjRs2DBXJoUn4ePjw8aNG/n333/56KOPsNudvezNPkqWLIlKpeLs2bPPPHflyhVeL1KEUR4eVJHLKSGX87cTri0pT2OtXk9gYsSwtkLBBYuF82YztRUKJBIJ3t7eKDKYBPymTIZCIuHHFDkZcXY7f6TYbCgtk2EFrqd4TYzdzn2bjTKJ80kkEt6Tyfj2/Hnu3LlDzZo1KV++PCaTiS1btvD222+n+/dAqVTSv39/Ll68yIQJE9izZw+1atVi9uzZxMTEZOj6RETSi7iyzAfY7HYWL15MjRo1WLNmDR06dOD8+fOEhIRQtWpVV5sn8gKMZjN6nQ6JREJYWBidO3d2tUluh8lkwmaz4efnl67+FUVlMj7TaHhLLqeDpyf9PT1Z+1RS6fMoLJMhA7wkEnxtNhQWCyW8vZOPFZbJ8JJKOWg0YhcEFnt7U16hoLRczjIfHx7b7cmacHDswC728aGsXE7ZNBYyxiTZVC50NNxROgVQtGhRateu/Yx8ChxJ4T///DM//fRTzhvmJEqVKsXKlSv57rvvWLp0qavNSTceHh4MHz6cWbNmsWvXLh4+fMi1a9f4+uuvKVmyJI8jIjhgsfDQZuN/ej1hT+VnZAZfqZTycjl7jUZqJ8qXApVKblqt/GGzvTAR/GV4SaV8qFIxS6vlrNnMb1Yro+LjUy26SsjlNPPwYHxCgiOyYrEwIi6OV6RSmqX4zteWydh/8iSVKlXCy8sLqVRKYGAge/fufSY/Iz2o1WpGjBjB5cuXCQ4OZuPGjQQGBrJ06VK3+76K5H5ERyOPIgBmi4W4hAT0Oh0XLlxg3LhxXL16lVmzZlGsWDFXmyjyAuyCQFx8PNqEBBQKBV9++aUolUqD77//HovFgtrLK907ju8m7lImUV2p5IHVii0DpUEFHAtohUKRphPwq9XKQ5uN0k+eUCoiglIREZR/8gSTIPAwxY5mObn8hbX+TSYTSqXS6Q0Ac4IkR8PdIhrgkE+dO3eO//77L9XxRo0a8dprr+XqqAZA06ZNmThxIosXLyYsLMzV5qSbMWPGMGTIEBYuXEj9+vUZMmQIkZGRvP/++wzu1ImpJhNN4+K4YrEwxkmlq2spldhwyKjAkUdRWiajsFTKW1ksZTpdo+E9hYI+sbF0jYmhpkLB20/ljy3z8eFtuZzesbG0jolBALb5+aW6L9SSybDZ7amcitq1a2Oz2Z6bn5EefHx8mDRpEhcvXqRbt24sX76cwMBA1qxZgzGdUV4RkZchEXJ74e30cvMm1K3r6E2QjYmirkbAkTyq0+sdkgupFI1CAWfPokjRpCrXkU/eP3D0ZIiNi8Nms+Hr5YUK4OxZqFzZ1aa5FQ8ePOCjBg3Y9e+/qHx9kaTjc9ExJoZiMhlLfXySj31rMjEoNpYHhQtTLCKC9b6+tEihF38rIoK53t50TSwf2SQqiuYeHoxQKJBKpSgVCppERdHCw4NxiYvryfHx3LRaWZViniQCpFJ8pFJGx8URJwiEpiGZAkck8smTJ/j4+DzbBNBiAbPZrT8XgiDw5ptvMnPmTPr27etqc1IRHx/P22+/zaRJkxg6dGiq51asWMGyZcu4fv06vr6+LrIw6wiCQHBwMKdOneLQoUOULVvW1SZljXz0G5CKHPyu//PPPyxdupTt27dTsGBBxo4dS7du3VCk9fcW3w+3vfe6G2JEI49gFwR0ej2RkZHExsUhkUjw9/fHz88PhUKR9o1CxO0wGI1ERUcDUCAgAFUuqzSUUxiNRgYNGkRAQACqDPaXuP5UHftrFgsl5HJkEgkFpFLCU2i/H1itGNLYi5EAKg+P51a3qqxQ8IfVSgGplOJyeaqHTzpzoUyJ0hBVLpRNgUNbrtFo0KXRpMzV+Pj40KxZszTlUx9++CE2m42dO3e6wDLnIZFIWLp0KcWKFaNfv37Exsa62iQRN6do0aIsXLiQ06dPU7t2bSZPnkxQUBB79uzB9pwSvyIiL0N0NHI5NrudBK2WyCdPkmU2BQoUIMDfHw+lMlc1+MrPJEml4uLiUKlUFAgIQC52IH0u06dP5/79+8yYMSPD1Zge22zMSEjgvtXKfqORDXo9A5OaXSmVhBoM/GKx8LPFwqSEhBdKm55HB5WKAKmUfnFxXDKb+dNm47zZzLSEBP5N5w92smwqFxdp8Pb2dkvpFDjkU7/++iu3b99OdbxQoUK0aNGCzZs35/pO215eXoSGhhIbG8uQIUOwWq0vPymP8NhmS5YtpvV4ujSuyP9TokQJVq1axfHjxylXrhwfffQRTZo0ISwsLNd/J0Ryntz7C5bPsVqtxMXHExkZiV6vx9PTk4KFCuHn65vh6hgirsVqtRIdHY3RaMTX19fR0ToXavJzit27d7Nt2zbmzZtHyZIlM3x+Z5UKoyDQMjqaKfHxDFSr6ZnoaMzQaCgqldIuJoZhcXEMUavxfMl4aeGZWG3qNamUAXFxBEVFMS4+HpMgoEnHe2sXBMwmU66NZiSh0Wjc1tFo2LAhfn5+7N2795nn+vTpw/379zl37pwLLHMuxYoVY+3atZw7d465c+e62pwco4hUyvGAgOc+iuRiBz6nKF++PKGhoRw6dIjChQszYMAAWrVqxQ8//CA6HCLpRlyR5iIEwGI2o9PrMZlMyKRSNBoNnp6euTJZVMQhlYqPj0cmk1GgQAHkKZo5iTzL3bt3mTRpEl26dKFr167wyy8ZOn9Piq7dn6eRP1FEJuPrpzp7/1a4cKr/P16gwDPnpXWssEzG8hdo/Je94DmTyYQAeLygt0BuwNvb222r2CgUCtq2bcvevXuZMmVKqshRYGAgZcqUYfPmzc9tIJebqFu3Lp9++imffPIJlSpVokOHDq42KduRSyQUFzfdnMK7777Ljh07OHv2LPPnz+fDDz+ka4UKLExsOCgi8iJElz4XIOBYkEZHRxMdE+NIEvb1pWChQnip1aKTkQtJUyolOhkvRK/XM2jQIIoVK8bcuXPzdNTHZDSiUCiQ5fJdV3eWToFDPvXvv/9y8eLFVMclEgm9e/cmLCwszzQ0GzBgAF26dGHcuHHcuHHD1eaI5ELq1q3LwYMH2bRpEwkJCRj0euISErDkI0meSMbJ3b9ieZyUCd5xcXFIExO8CxQogGcGE2BF3AdRKpVxBEFg8uTJPH78mHXr1uXphoV2QcBkNr+wU3JuQaPRuG1EA6BatWoUK1YszaTwTp06oVQq2bZtmwsscz4SiYT58+dToUIF+vXrx5MnT1xtkkguRCKR0LRpU9auXYtKpcJusxEVFUVsXBxWMe9FJA1ERyOb+ctmo2h4OLcSK92cN5spGh5OfGJlmx0GA+UiIlKd83SCt1KppECBAviLCd65nlRVpRIdRpGX8/XXX7N7924WLlxIqVKlXG1OtmI2mxEEIVc26Xsad49oSCQSOnbsyKFDh5KrfCXh4+NDhw4d2Lp1a55Jovbw8OB///sfNpuNQYMGYXmqApuISHqRSqUoFAr8/Pzw8fHBYrE4NkXj48UKVSKpEB2NbKaoVMpPBQum2fX3aSxJCd5PnmDQ6/FUqylYqBC+Pj5igncuR5RKZZ5bt24xdepUevfuTfv27V1tTrZjNBqRy+V54vPh7o4GQIcOHUhISODYsWPPPNenTx/+++8/jh8/7gLLsodXXnmF9evXc/36daZPn+5qc0RyORJA7elJwYIF8fb2xmQyERkVRXxCArYUpcJF8i+io5GNWAQBmURCYZkM+XOkMQIOWUhMTAxRUVGYzWY03t4ULFQIb40m12u0RUSpVFZISEhg0KBBlClThs8++8zV5mQ7AjiqTeWRSJe7S6cA3nrrLd5555005VOVKlXi3XffzfWdwp+mevXqzJs3j82bN7NlyxZXmyOSB5AAXmo1BQsWxMvLC6PBQGRkJAlaLXaxQlW+RlzFZgCt3c7wuDjeiojgnSdPWKvX0zEmhk8Sd+xqRkayVKtlZFwcZSIimBAf/4x0KomkBO+ERK/fbrc7ErwLFuSsTEbz6GiKR0QQGBnJYq0Wq/hFzZWIUqnMIwgC48ePJyoqirVr1+YJKdHLMJvN2POIbApyR0QDHPkYJ0+eJCYm5pnnevfuzQ8//MDDhw9z3rBspHv37vTt25ePP/6Yy5cvu9ockTyCVCJB4+VFwUKFUKvV6PV6hwxcpxMdjnyK6GhkgBlaLVcsFjb5+bHd359LZjM3n3IgQvR6KsjlfBcQwGgvr2fGsAsCgiAQlZTgLZUik0oJSFyEXjabGRkXx0C1mh8CAljg7c1Oo5HlbthdV+T5iFKprLNx40a++eYbli5dSvHixV1tTo5gNBqRy2R5plmjt7c3BoPB7XMc2rRpg91u55tvvknzOV9fXzZv3uwCy7KXzz77jOrVqzNw4ED++ecfV5sjkoeQSiR4azQULFgQlacnOp2OyMhIdHo9oruRv8gbv2Y5gNZuZ5fRyCofH+oqlQAs9fGhamRkqtfVUSoZksLB+CsxKcomCCQkJBCr1yMIAgoPDwp4eaGxWJAkJCQneC/W6Rjh5UWXxAZib8rlTLTbma3VMk6jyf4LFckyVquV2Li45DLEYhQj41y/fp0ZM2YwaNAgWrZs6WpzcgQBR/8MT0/PPFPwQZN4z9LpdPi+oG9IpnnwAJwgzSoI9KxShauhofSuWjXVcypgZMOGHN24EVPLlu4RbdJooESJLA+jUChYt24dzZs3p3///uzfvz/PyPZE3AOZVIqPtzdeXl5otVq0CQnodTq8EnuA5ZV7ncjzER2NdPKnzYZFEKiqUCQf85FKeeupXeoqKZ5PSWxMDAaJBJVKhdRoxNfbG4VUCk9FRH61WrlisaSKYNgAkyBgEAQ8RW2/WyM24Ms6sbGxBAcHU7lyZaZNm+Zqc3IMi9mM3W7P9d3AU+Lt7Q04cm2c7mg8eAA1ajxzD80sMy0WjEYj9tq1UzXvAwi22+mp0yENCoLn3ONzFIUCrlxxirNRoEABQkNDadOmDRMmTGDFihViDpmI05FJpfj6+CQ7HPHx8eh0OjQaDSqxXH+eRnQ0nIz6OTdoHx8fCqpUqC0WMBqfe75OEJjg5UXLNBYbeWf5kfewJ0asDAYDnp6e+Hh7iz/WmcButzNq1Ci0Wi379u1D4Q6LuhzCaDIhk8mQ56FrTopoZEuehlbrcDKkUnCC1EyuUGA2mZAJAurEqHUSUsBuNqOzWvFLQxKbo1itjut2YpJ9pUqVWLJkCcOGDaNSpUoEBwc7bWyRvIMgCEycOJFDhw4RFxWFj1ZLFw8PZvr5pXsMuUyGn68vlkSHIy4uLtnh8PDwSNPh+Mtm473ISI4FBFDRTe6Po+fPJ06hIDQ01NWmuD2io5FOislkKCQSfrJYeC1xlzrebucPm43AdJyvkMvT1cG7slzO7zYbxfOIRju3scNg4NOEBH4rXDjd54hSKecREhLCsWPH2Lx5M6+99pqrzckxBBzdwD3y2M5eUkQjWytPyeVOiTJIALmnJ3qLBU8fn2feBw+NhtjYWCwSievLjZvNTh+yXbt2/PLLL8yaNYvy5csTFBTk9DlEcjenTp1i586d7Nmzh2JaLYOaNs30WAq5HH8/P8wWC1qtltjYWBQKBRqNBqXYLyxPISaDpxONVEpnlYpZWi3nzGbuWK2Mi49HCk79Qoz18mK3wcBirZY7Viv3rFYOGI3Md/MSkfmVrFSVGh0XR7/Y2GyyLPdx+fJl5s2bx4gRI2jSpImrzclRLBYLNrs9z+njU0qncgMqlQqr1Zpm8rqHhwcyqRSDXu8Cy3KGKVOmUL9+fYKDg/NclS2RrPPo0SMKFy5M9erVKVyggFN2qpUKBQH+/vj7+wMQExNDTEwM5kRJpMXFlapcPX9eQHQ0MsAMjYZqCgW9Y2PpGhNDDaWS0nI5Hk6UyDTw8GCznx8/mM20iI6mdXQ0a/V6Xhe1/m7Fi6pK5fSNKS90942KiiI4OJjq1aszceJEV5uT45hMpuROu3mJJOmUu/fSSEKpVCKVSjEYDM88JwE81WoMRmOeLdMpk8n48ssv8ff3p1+/frnmfRPJfkaPHs3HH3/M48ePKVq0KDW7d3/mNXF2OyPj4igXEUHJiAh6xMTw4Cmn/bDRSIOoKN6MiKBmZCQhifmoHkolAQEBfCCR8KXFwrDISEr99x/j4uKSz/3dZuODxNL/DaOiuPBUZO+C2UzL6GjeTGxBMCchIVVrgFMmE22joykXEUGFJ0/oHRvLwxT2JbUjOGg00iFxnr1GIzZBYEZCguO8mBhmm0wIefQekB2IjkYG0EilrPL15X7hwvxUqBA9PT25b7VSInGBeblgQQap1anOeUMm458iRZJ1hbWVSv4pUgSfxGTDrp6ez8h0Gnh4cDAggD8KF+ZO4cIcDgigR2IVqvzAqVOnaNu2LeXKlaNChQr07t2bh48fJz//r83G0Lg4ykdE8FZEBM2jo/kxxWL7mMlEi8SbRMUnT+ifImpgFgRmJiRQ9ckT3oqIoFV0NOdfIkP41mTi/aio5L4mC+PjiYiKSm7AV95gYLPBQJ/YWN6KiGCFTodNEBgbH897kZGUiIigbmQk61PshC7WatlpNPKtyUTR8HCKhocn23HbYqFzTAwlEm9qE4xGdCnOHT16NP369WP58uVUrVqVunXrZvVP7lJsNhsjRozAarUSEhKSZ0q7phcBR1nb5+mTczNqtRqJRJJrIhoSwFOlwmg0plmC09PTEwQBYxqOSF7B19eXjRs38vfffzNq1CjsYndnEWDmzJlMmDCBV199lZ9++omwL7985jWj4uP52Wplk58fh/z9EYAesbHJm283LBaC4+Jo6+HBqYAAxnl5sUCnY0fi90kCSKRStgBV1Gp2ymT0MhqJi4932JCQwBC1mmMBAVRTKOgTG0tM4ufzP5uNnrGxvCOXcyIggM+9vfn6qdYAekEgWK3maEAAu/z8kAAD4uKe2TiYo9UyUK3mdIECNFAqWaPXs8NgYImPDwd8fIgRBMLOnXP63zivkr9+0bPILxYLv9tsvCOXkyAILEn8ADfLQ1Vi3AG9Xk9wcDAVKlRAp9OxYMECBnz6KccSK291iInhFamUTX5+FJZKuWG1Ji8KjptM9I+NZZSXFyt8fLAIAidSOBJTExK4a7US4utLEamUMJOJ7rGxnAoIoEQaC9xLiX1NZnl7855CwV2DgUlaLQaZjKlJVaXi4lis0/GxRsNMjQa5RIIdKCqVstbXF3+plKtmMxMSEigsldJGpWKIWs09m40Eu52lPj4A+Eul6AWB7rGxVFMoCAsIINJsZrxWy8dffMGy995Ltuvs2bN4e3uzffv27HwrcoQVK1Zw+vRptm/fTpEiRVxtTo5jtVqx2Wx5TjYFIJVK0Wg0ucbRABw1//V6zGYzHk8lhcukUjxUKvQGA55qdZ5zDJMoU6YMq1atom/fvixbtoyxY8e62iQRF+Pj44NGo0Emk1G4cGEID0/1/AOrle9MJg76+1M98XuzyteXapGRHDWZ+EClYo1eT12lkjGJkc6Scjn3rFZW6/V0TbGZWkep5CNfX0djY4OBu1otdrudD+VymisUyGQyPvf25pTJxNcGA8O8vNhoMFBUKmVOYiGWUnI5/9ntzNFqGePlhVQiodVT99ilPj5UevKEuzYb5VL8/g9Sq2mZ4rXr9Ho+8vJyHLNYmO/hwfdPbSqLPB/R0cggq3U67ttsKIC3FQr2+fsTIBUDQ86kVatWqf5/6dKlVCpXjrt2O1dNJqLsdsICAvBL/LunTJxfrtPRVqVifIqeIxUSo0mPbTZ2GAxcLViQIolRqKFyOafMZrYbjUxJo09JUl+TTioVCQkJVDYYGO3hwVKbjU9SyNnaq1SpbpRAKhuKeXpyzWLhG6ORNioVXlIpKsAkkVA4xTg7DQaMgsAKX19HBTNBYI6HB32OHePjJ08oVKgQ4NgpXrx4ca6X2pw9e5ZFixYxduxY6tWr52pzXILJZEIqkaB8alGbV9BoNLlKgiOXy5HL5RgNhmccDXBENWJiYrBYLChz+ffvRbz//vtMmDCBhQsXUrFiRZo1a+Zqk0TcmHs2G3JI1QLAP7EFwL1EedI9q/WZjdkaSiXrDAZsgoAsUYae1CZAAqg9PQlQKpE8eUIFm43IqCg8PT3RaDRUUSi4l9ir7J7VSnWlMlW1x5oKBTpB4F+7nddkMh5YrSzQ6bhusRBtt5MUq3v8lKNRJcW/4+12wu123k1xXXKJhCply4qNB9OJ6GhkgEoKBd8WKOBqM/I8Dx48YMGCBVy/fp3o6Ojk0P1jQeCW3U4luTzZyXiaW1YrPZ8jM7tttWID6kRFpTpuFoTnOotJfU2WJIZupVIpdqv1mb4mVdKIhoTq9Ww3GHhst2MUBCxAxZfIgu5ZrVRUKFKVSa4hk2EXBO7fv5/saJQrVy7XOxnh4eEMGzaMevXqMXr0aFeb4zLyqmwqCW9v71wV0ZDgSArX6XR4C8Iz1QKVSiVymQy9Xo8yO5oQuhGjRo3i1q1bjBgxgsOHD1OmTBlXmySSD3i6TYAEkEgk+Pn64mW1YnxBi4AX0Ts2ltdlMhb5+FBEKsUONIyKeiav8nltCkQyh+hoiLgdvXv35vXXX2fRokUUKVIEu91Ow3r1sACql9wAXvS8XhCQAd8GBPB0av3zbiw6QWC8lxcNbDY8VCrkKRySlPsyT59/wGhkplbLpxoN1RUKNBIJX+r1qXJJsoI6l4dtrVYrw4YNQyaTsXLlSmT5tNiB1WbDarUmJ03nRXKbowGOPA2tVuvo1P6U3CIpKVyr1WKz25Hl4Yi2VCpl+fLlfPDBB/Tt25ewsLDs6fAukuspLZNhBa5bLMnSqRi7nfs2G2USN9hKy+Vceeo38IrZTEmZLDma8SJ+slqp7eWFl5cXNkHghtVK/8SNxdJyOYeNRgRBSI5qXLZY0EgkvCqVJtuyyMeH9xLtu5yOMtE+UilFpFJ+tFgITDzPKgjcuHuXyjVrpu+Pk8/Ju3dIkVxJTEwM9+/fZ/To0dStW5fSpUsTl6LqRHmZjFtWK7HPSVCsIJdz5jk3j0pyOTYg0m6nuFye6lH4OQvdynI59202Kvr4UEqpTHXOi/qiXDabqaFQ0FetppJCQXG5nIeJId4kFIm5HCkpLZdzy2JBn2KH5YrNhlQi4a233nrufLmNRYsWcfnyZUJCQihYsKCrzXEZJpMJiUSSpkQnr5DbpFPgqL6kVCqfu3Pqmbi4Sas6VV7Dy8uLDRs2EBMTw9ChQ7E9dR8TEQEoIZfTzMOD8QkJXDab+dViYURcHK9IpclyqSFqNWfNZpZqtfxhtbLTYGCDwcCQdG6chRoMhBmN3LdamZqQQJzdTrfE72JfT0/+sduZlpDA71Yr35pMLNLpCFarkUok+Eok+EulbDUYeGi1ctZsZkY670sD1WpW6vUcNRr53WZjislEfIokc5EXIzoaIm6Fr68v/v7+bN26lYcPH3L27FlmzJiR/Hx7pZJCUin9YmO5YjbzyGrlsNHItcRdkrFeXuw3Glmk1XLPauW2xcKqxBtCSbmcDioVI+PjOWI08qfNxnWLhS90Oo6bTGnak9m+JiXlcn62WPjeZOIPq5UFWi0/P7WT84ZMxq9WK/etVqLtdiyCQAeVCpVEwqi4OH6zWjlnsTDNZKJT06bJsqnczsmTJ1mxYgVTpkzhvRQJ7vmRZNlUHg7V58aIBjjkUyaTCVsamxpSiQSVSoXBYMgXOu3ixYuzdu1aTp8+zbx581xtjoibsszHh7flcnrHxtI6JgYB2ObnhyLx/lZZoWCNry8HTCYaRkezUKdjgpfXM/mNz+NjjYaVej2No6O5bLGw0c8vWfb8ikzGVj8/rlutNI6OZlJ8PB+qVIzy8gIc39kQX19uWK00iI7m04QEpqczkhysVtNJpWJUfDyt4+PRSCS0qFMn43+gfIoonRJxK6RSKSEhIUybNo0GDRrw1ltvMXv2bDq2bQs4ogDb/f35LCGBnrGxWIEycjlzExuD1VYqWevry1Kdji/0erwlEgJT5DIs9fFhuU7HZ1ot/9ntBEgkvKtQ0OQ5O8pJfU2W6HSs0utRAKXkcrq/5MbYy9OTX6xWhiRGY9qrVPRRqzmZwqHp4enJebOZ5tHR6ASB3f7+1FYq+crPj0+0WlpER+MJtJLJmPHRR5n/o7oR//zzDyNGjKBp06YMGTLE1ea4FJvdjsViyfUyuJeh0Wj4999/XW1GhlElFoAwGo14pfEeqT09MRgMmE0mPPJB5cF69eoxffp0PvvsMypWrEj79u1dbZJIDjNo0CAGDRqU/P971GpI8dvpK5Wy4iXSulYq1TPVn1JyOY0Id1KbAIB2Lzi3llLJkYCA5z5fT6nkh6fybP9JUekw5TwpkUskzPT2Zqa3N1gsYDbD5MlQufJz5xL5f0RHQ8TtqFevHj/88EOqY/+cOAGJ/SJel8lY5+f33PNbqlSpStOlRCGRMF6jSVURKiVdPT2f2V1p4OFBgxcsJNK6MSklEpb6+CSXrk1iaop5C0ilbE/shpqS8goFu5KOJ93UUix0li1b9lxb3BmLxUJwcDBeXl4sX74caR7WtqcHk9HokE3l8UVqbo1oSBPfm+c5GnKFAoVCgd5gyPPvYRKDBw/m1q1bjB07llKlSlHZFQutNLq252ny2/WK5DlER0NERCRHmDt3Ljdu3GDfvn34vcBRzC8YTSZHJ+o8LJuC3OtogCOqERsbi9Vmc/TMSYEER65GfHx8ms/nRSQSCQsWLODevXv069ePo0eP5lyOlUYDCsX/b77kJxQKx/WLiORCREdDREQk2wkLC2PNmjXMmjWLd99919XmuBy73Y7ZbMbnqYhXXiQ3JoMn4eHhgVQiwWAw4J3GQk+lUqFNSHju83kRlUrFhg0baN68OYMHD2bHjh05U2q7RAm4cgVy6WcpS2g0jusXEcmFiI6GiIhItvLw4UPGjBlDq1at6N+/v6vNcQuMibk6qnwgufH29kab2Nk3t8nlknpqGI1GNBrNM71OpBIJqsRcjbSez6u8+uqrrF+/nk6dOvHpp58yd+7cnJlYXGyLiOQ6REdDREQk2zCZTAQHBxMQEMDixYuzr7pSLtMxWwwGPOVypDYbZKZcaC66Xo1GgyAI6PX6XNkvROXpid5geG4ncLVajV6vx2g0PtNzIy9To0YN5syZw8SJE6lUqRLdu3d3tUkiIiJuiOhoiIiIZBszZszgzp07HDp0KHtkQrlQty0IAlKz2ZFAnBWbc4luO8m50Gq1udLRUCgUyGQyjAZDmo6GPLHnhkGvz1eOBkDPnj355ZdfmDJlCmXKlKF69equNklERMTNEB0NERGRbGH//v1s2rSJBQsWUKlSpeyZJBfqtk8cP87cuXPZvn07hQsXzvxAuUS37Z1Yejq35mkkyacMBgPeif//NGq1mtjYWCxWKwp5/vpZnTVrFr/99hsDBgzg6NGjvPrqq642SURExI3IX3dEERGRHOH3339n/PjxdOjQgR49emTvZLlgsZ2S7cuX41mzJoUbN3a1KTlCkqORWytPAXiqVOh0OkwmU5p5NR4eHkilUgwGA4rE680vKBQK1q1bR4sWLRgwYAD79u3LN+V+RUREXk7uyswTERFxewwGA4MHD6Zo0aLMnz8/T3e9zigGg4GTJ0/SsmVLV5uSY6SUTuVW5HI5CoUCo9GY5vMS/r+Bn13ID73CU1OoUCE2bNjA7du3mThxIkI+/BuIiIikjehoiIiIOJWpU6fy6NEj1q1bh5eXl6vNcSu+//57jEZjvnI08kJEAxzyKZPJ9FxHwlOtRhCE5zojeZ23336bJUuWsGvXLtavX+9qc0RERNwE0dEQERFxGjt27GDHjh3Mnz+fsmXLutoct+PIkSOUL1+eErlM7pUVkpzNvOBovMiRkEmleHh4oNfrya/7+e3bt2fo0KHMnDmTs2fPutocERERN0B0NERERJzC7du3mTJlCt27d6dTp06uNsftsFgsHDt2LF9FM8Ch4VepVLlaOgWw22Qi6CURi7USCR3MZiwWCwCj4+LoFxvrlPnPm80UDQ8n3m53ynjZxdSpU6lbty6DBw/m0aNHrjZHRETExYiOhoiISJbRarUMHjyYEiVKMHv2bFeb45acPXuW+Pj4fOdogEM+ldGIxo4dOyhXrlw2WZQ5JBIJZrMZ23N6nwz39uZ/cjkGvT6HLXMfZDIZISEh+Pn50b9/f3Q6natNEhERcSGioyEiIpIlBEFgwoQJhIeHs27dOlT5rJdAejly5AjFixd3u8VzTqDRaHJ9RAMcjoZEIsHwnKiGRiqlqJcXRpMJu5tHHpJIir44E19fX0JDQ/nzzz8ZPXq0mBwuIpKPER0NERGRLLFlyxYOHDjAokWLKFmypKvNcUtsNhtHjx6lZcuW+aYK16lTp2jbti3lypXj559/Zvv27Tx8+BCA8+fPU7RoUeLj45Nff+vWLYoWLcpff/3F+fPnGTNmDPHx8RQtWpSiRYuyePFiAOLi4hg5ciTl2rShZEICPRISeJCiU/oOg4FyEREcM5moGxlJyYgIBsXGYhAEdhoM1IyMpFxEBNMSErClWADH2e2MjIujXEQEJSMi6BETk2rcJM7I5TRKSKB4RAQfxsTwT4roxmKtlrYGA0CazohdEPhCp+O9yEhKRETQOCqKQxlMHr9hs9Fcp6NkixZ88MEH3L9/P9XzmzZtolatWrz55pvUrVuX3bt3p3q+aNGibNq0iT59+vDWW2+xYsUKFi9eTJMmTdi9ezc1a9akbNmyDBkyJEvOYdmyZfniiy84fPgwy5cvz/Q4IiIiuRvR0RAREck0N2/eZPr06fTr1482bdq42hy35fLly0RFReUr2ZReryc4OJijR49Su3ZtLBYLAwYMSNdOf40aNZg5cybe3t789NNP/PTTTwwZMgSAUaNG8fPPP7Np9mwOqdUIQI/YWCwpnAYD8D+9nhBfX77y8+O8xUL/2FhOms1s9fPjC19fthgMHDaZks8ZFR/Pz1Yrm/z8OOTv/9xx19ntzAL2+PgQLwgMiYt7xn7Vc5LCv9Dr2WU0Mt/bm+8DAhisVjMiPp4LGegQ/7lez6ceHhxdvRq5XM6YMWOSnwsLC+OTTz4hODiYU6dO0atXL8aMGcO5c+dSjbF48WJatmzJyZMn6datGwCPHj3i6NGjbN68mc2bN3Px4kVWrlyZbrvSonnz5owbN44FCxZw7NixLI0lIiKSOxEb9omIiGSK+Ph4Bg8eTIUKFfj0009dbY5bc+TIEV555RXeeecdV5uSY7Rq1Sr536+99hr+/v6cOXOGu3fvvvRchUKBt7c3EokkVff0Bw8e8N1333Hw4EGqe3iATMYqDw+qxcVx1GTig0TZnkUQ+Nzbm+KJXbpbe3iw22jkRsGCeEmllJHLqaNQcM5spo1KxQOrle9MJg76+1NdqQRgla8v1SIjnxl3rq8vb8TG4mmxsNzHh6CoKK5bLFRVKJLt9FSrMURHY5fJIDGCZRYEVuh07PT3p1ria9+Uy7lssbDFYKBW4rwvY7JaTS1BgOLFGTFiBL169cJkMuHh4cHq1avp0qULffv2BSA4OJgff/yRkJAQ6tSpkzxG+/bt6dq1a6px7XY7y5YtS+570qlTJ6dUjhozZgy3bt1i+PDhHD58mNKlS2d5TBERkdyDGNEQERHJMIIgMGbMGGJiYlizZg3KdC6S8iOCIHDkyBFatmyJVJp/brkPHjxg6NChBAYGsn//fg4fPgzA48ePMz3mvXv3kMvlVK1aNfmYv1TKWzIZ91LInDwlkmQnA6CQVMobMhleKf7+haRSIhOjK/dsNuSQyllIa1w5UFUuR6VSYbFaKSWX4yORpHoNOBwluVyeKv/hgc2GQRDoGhNDqYiI5Mduo5FHz0kuT4vyMlnyv4sUKQJAZGRk8t+nRo0aqV5fo0YN7t27l+pYlSpVnhn3jTfeSHYyAAoXLpw8blaQSqWsWLGC1157jb59+6aSy4mIiOR98l9EIw3NbZ4mv12vSI6wfv16wsLC2LhxI8WKFXO1OW7Nzz//zL///puvZFMAvXv35vXXX2fRokVs27aNn376iUePHmGxWJJ7a6RMEnZmUrLiqf+X8OyPnQTIbLq2t0aTHKlICwmgVquxxcYmX6Mu8b9b/Px49SmHU5mBvJ2U15aU75PRxHO1Wv3MMbk89V9IIpE4LaFdo9EQGhpK8+bNGTZsGJs2bUKWwmESERHJu+Sf7TWNBhQKsNvBbM4/D7vdcd0pdqpERLLCtWvXmDVrFkOHDuX99993tTluz+HDhwkICKBmzZquNiXHiImJ4f79+4wePZq6detSrFixVOVtCxQoAEB4eHjysVu3bqUaQ6lUPlNGtnTp0litVq5fv/7/c9nt3LfZKCPP/L5ZaZkMK3A9hbOT1rhW4Ger1VF9CrhvtRIvCJROY26VSoVEIsGauNlTRiZDKZHw2GajuFye6lHUSYvu0qVLc+XKlVTHrly5QpkyZZwyflYoXrw4a9as4fvvv2f+/PmuNkdERCSHyD8RjRIl4MoVyAMlFjOMRuO4fhGRLBIdHc3gwYN55513mDx5sqvNcXsEQeDw4cM0a9bsmR3jvIyvry/+/v5s3bqVIkWKEBERwaNHj/D29gYci86kSlKTJ0/m/v37hISEpBrjjTfeQKfTcfbsWSpUqICnpyclSpSgWbNmjB8/ngWDB6Ox2ZhjMvGKVEozD49M21tCLqeZhwfjExJY4O2NRiJhjlb7zLgKiYRpCQnM8vZGDnyckEA1hSKV5CoJqUSCXC7HZrUi4Ch9O1St5lOtFjtQU6EgQRC4YrGgkUjo4umZafuTGDZsGMHBwVSqVImgoCC+++47jhw5wo4dO7I8tjOoX78+H3/8MbNmzaJixYq0bdvW1SaJiIhkM/nnlw/ExbaISBaw2+2MHDkSo9HImjVrUKSxuBJJzZ07d3j48GG+a2IolUoJCQlh2rRpNGjQALVanSyXAkcOw+rVq5k8eTKNGjXinXfeYdKkSQwePDj5NdWrV6d3794EBwcTExPDuHH/196dh0dZn/sffz/PM8/skw0Ei6LgvlGKirUVrbiBeCrV/kShmrhUcalWRYs9rkXt0YriUi3UWg3ggmurBaytCy1Hr1O0paLVujRunCMBss6+/v5ISJOwBZjkmeXzui6ui5DJzD0BJs9nvvf9/c5gxowZ3H333Vx//fXUXnstyWiUw91uHq2qwt7BbYPvrqjg+vZ2altaSAKH2/ZG9+sDLvH7uaS1lf/LZvm6bXNXRcVm79PlcpFLpUgkEng9Hn4UCDDINLkvEuGzbJYKw2CUy8Vl3b43O2LixInMmjWLuXPncsMNNzB8+HDmzJnDN7/5zbzcfz5ceOGFvPvuu1xxxRXsueeeHHTQQU6XJCL9yMjpJB0pBqtWwbhx4HZ3tIKVi1SqowVu+XIYNcrRUu677z5uu+02Fi5cyPjx4x2tpVjcddddzJs3j1WrVpX1wPwzzzzDpZdeSkNDA54dWHnooUheE5qamsAwqKmuzs8dFtBrwvaKx+NMnjyZpqYmXnzxxa5WOikiRfL/L+9K4P/fQCufGQ0R2W5vvPEGt99+Oz/84Q8VMrbBkiVLOO6448o6ZABdLVPd5zTKhc/vJ5lMkt6GnaVKndfr5eGHHyaRSHDBBRf0y+nkIlIYFDREZIvWrl3LRRddxDe+8Q1mzJjhdDlF45NPPuEf//hH2e02tSkbtk3dkZOmi5XX68U0TaLR6BZvN7Otrce2t91/zSzBLWGHDRvGgw8+yIoVK7jpppucLkdE+kl5zWiIyDbJZDJccsklANx///3aknIbLFmyBK/XqxUgyntFwwB8Ph+xaJRgMIi5mVmSq4NBLtrEtrPQMUheir7+9a9zyy23cM011zBq1KiuU8qliJTbFvrl9nzzQEFDiku5/Sd3+PnOmTOH119/nSeffLLHCc2ydUuXLmX8+PGbPLOg3JRz0ADw+3xEIhHi8Tj+zewuNdg0GVyigWJLamtreeedd5g5cyZ77703hxxyiNMlSV9sODJgw8xCOdGRAdtEQUOKg17UBvxhly1bxpw5c5g5c2ZB7VpTDL788kveeust7rvvPqdLKQjl3DoFYFkWHo+HWCy22aBRzm699Vb++c9/ct555/Hiiy+y8847O12SbI2ODHC6iqKhoCHFQS9qA/qQX375JZdccglHH300P/jBDwb0sUvB0qVLcblcHH/88U6XUhDKfUUDOlY1mltaSKVS2hq6F9u2efDBB5k4cSLnnXcezz77bP52J5P+o4tt6QMFDSkeelEbEKlUigsvvBCPx8N9992HWYbtHDtqyZIljBs3jootnLFQTtxuN7Ztl3XQcHs8WJZFNBajUkFjI0OGDOHhhx/mO9/5Dtdccw133XUXxg6ejSIiztMVhIj0cPvtt/PXv/6VefPmUVNT43Q5RaepqYk33niDk046yelSCoZhGASDwbJtnYJ/D4XH43GyOr5qk0aPHs3s2bNZtGgRv/71r50uR0TyQEFDRLr84Q9/4IEHHuDaa6/l0EMPdbqcovTSSy+Ry+WYMGGC06UUlFAoVNYrGtARNABisZjDlRSu7373u0yfPp2bbrqJ5cuXO12OiOwgBQ0RAeDzzz/nsssuY8KECVxwwQVOl1O0Fi9ezNe//nUGDx7sdCkFpdxXNAAs0+wYCo9G0ZrG5l177bUcccQRTJ8+nc8++8zpckRkByhoiAipVIrp06dTWVnJ3Xffrd7o7dTW1saf/vQnHdK3CVrR6OD3+0lnMiTLbfe8beByuZg7dy4VFRWcc845Wz3sUEQKl4KGiDBr1izeffdd5s2bR2VlpdPlFK1XXnmFVCqloLEJChodbNvG5XKpfWorqqqqePjhh/n000+54ooryGmuRaQoKWiIlLkXXniBhx56iJ/85CeMHj3a6XKK2pIlS/ja177GsGHDnC6l4Kh1qoNBx1a3iXicTDbrdDkFbb/99uPee+/lhRde0Jk0IkVKQUOkjDU0NDBjxgxOPvlk6urqnC6nqMXjcV5++WWtZmxGMBjUikYnr88HhqFVjT6YNGkSV155Jbfffjt//OMfnS5HRLaRgoZImYrH45x//vkMGTKE2bNnay5jB7322mvEYjEFjc1Q69S/mYaB1+vVUHgfXXnllZxwwglccsklfPTRR06XIyLbQAf2iZSp66+/no8//pglS5YQDAadLqfoLVmyhH333Zc99tjD6VIKUjAYJBKJOF1GwfD7/cRiMRKJBF6dgr2xhgbobLUzgZ9Pn84PVq3i5ilTuP/++0v3NSsY1OG0UlIUNETK0NNPP82jjz7KXXfdxf777+90OUUvlUrx0ksv8f3vf9/pUgqWVjR6sl0ubNsmGo0qaPTW0ABjx0Iq1fVHAeChbJZINIr5rW+R83pLcxXWtmHFCoUNKRkKGiJl5oMPPmDmzJlMmTKF008/3elySsLrr79OW1ub2qa2IBQKEYvFSKfTuFz60QMdqxqtra2kMxlcluV0OYUjHO4IGaYJ3f6tmIDHsmhrayOXyRDw+52rsT+k0x3PW5smSAnRq71IGYlGo5x//vnsvvvu/Nd//VdpviPogMWLF7P77rtrdWgLNrS6RCIRbaHcyev10t7eTiwWI1SqrUA7wuXqeIe/G7dt4wPa29uxPB68Xq8ztfUXna8iJUZBQ2RTuvUHl4pcLsc9t91G4F//Yu7cufg2NVSp/uBtlslkePHFFznttNMU3LYgFAoBHReIChodDMDn8xGLxQgGAvr300d+v59UKkVrWxuWy4WtFTKRgqX/nSK9baI/uBSkUil+EI9zldeLPW3apm+k/uBt9uabb7Ju3Tq1TW3FhhUNzWn05PP5iEQixBMJfKX27nw/MYDKigqamppoaWlhUE0NpqlNNEUKkYKGSG+b6Q8uZqlMhvZwGI/Xix0IbPpG6g/eLkuWLGHo0KGMGTPG6VIK2oYVDR3a15PLsvC43USjUQWNbWAYBlXV1axfv56W1laqq6vRepBI4SmNqyiR/rCJ/uBilM3laGltxXS5CFRVbfnG6g/eJrlcjiVLljBp0iS9o7oV3VunpCef309LSwupdFptQNvAMk2qqqpobmqivb2dis5/YyJSOPSTUaSE5aBjh5ZslsqqKr3jl2dvv/02q1ev5sQTT3S6lIK3oXVKKxob83g8WJZFNBp1upSi47ZtQhUVRKNRYvG40+WISC8KGiIlLBaNEo/Hqais1PaZ/WDJkiVUV1dz+OGHO11KwfP7/RiGoRWNTdgwFB6Px8nmdFb4tvL5fPh8Ptra2kiW2GydSLHTGq1IiUqlUrSHw/j9fh0I1g9yuRyLFy9mwoQJOheiD0zTJBgM5j9opNP5vT+H+FwuEtksiXB4y7MaJfJ888kAKioqSKfTtLa0UDNoEJZaGUUKgn46ipSgDXMZLperqzde8uvDDz/kX//6FzfddJPTpRSNYDCYv9apYLBjhiqVKonZIgvwu1xkYzFyprnlNkfb7nj+0sUAqqqqaFq/npaWFmpqatQqKlIAFDRESkwOaG1tJZfLUaW5jH6zePFigsEgRx11lNOlFI1QKJS/FY2RIzu2Yi6hmY/3/vpXrrrqKubMmcPo0aM3f0Odd7NJG4bDm5qbaWtro6KiQq9/Ig5T0BApMdFolEQiQXVVldoH+tGSJUs47rjjcLvdTpdSNPIaNKDkLrbHHHQQqV/9il8sX87cM890upyiZNs2FRUVtLa2Yrtc+P1+p0sSKWu6ChEpIclUinB7O4FAAI/mMvrNp59+yrvvvqtD+rZRXlunSpBhGNTW1rJ06VLWrl3rdDlFy+f14vf7aW9vJ1kCbXUixUxBQ6REZLNZWltasN3urq1EpX8sXboUj8fD+PHjnS6lqOR9RaMETZkyBcuyePzxx50upaiFQiFst5uW1lYymYzT5YiULQUNkRLQNZcBVFZWqi+5ny1evJijjz6awOZOWZdN0orG1lVWVjJ58mQWLFigC+QdsGE43DAMWlpayGnbYBFHKGiIlIBIJEIimaSyslJzGf1szZo1vPXWW2qb2g5a0eiburo6Vq9ezSuvvOJ0KUXNNAyqqqpIZzK0trWhqCEy8DQMLlLkEskk4XCYYDCIpxQGkxsaCnonoTd+8xsOzGaZuMsusGpV/u64DHYSUtDom6997Wt89atfpb6+nuOPP97pcoqa7XJRWVlJS0sLtsulVUiRAaagIVLEMtksra2tuN3u0vgB2tAAY8d2nI1QoE6IRjkB8J94Yn7v2LY7tmst4bCh1qm+O/vss5kxYwafffYZu+22m9PlFDWvx0MgEKA9HMZl26XxhoxIkVCPhUiR2jCXYVBCcxnhcEfIME1wuwvuV9a2iWUyGD5ffu/bNDued4lfhIdCIcLhMNls1ulSCt7kyZMJhUIsWLDA6VJKQjAYxOPx0NrSQlqzLyIDRisaIkUqHA6TSiaprqkpvbkMl6vjHf4Ck4jHSRkGbr+/IxzkUxlswxkKhcjlckSjUe2MthU+n48pU6bw+OOPc/XVV+u8lh204Q2ZpqamrpPDTaMk3p4RKWgldnUiUh4SySSRSIRgKIS7AC/IS1U8HsftdpdesBsgG8KF2qf6pra2lqamJhYvXux0KSVhw3B4NpPp2qVPRPqXflqKFJkNcxkej0en3g6gbC5HMpnUQYg7IBQKAWggvI/22msvjjjiCOrr650upWS4LIvKqioSiQQRBV6RfqegIVJEckBLSwuGYZTOXEaRSCYS5HI5vAoa223DioaCRt/V1tbyl7/8hffee8/pUkqGx+0mFAwSjkSIJxJOlyNS0hQ0RIpIuL2ddDpNZWWl+osHWDyRwLZtLMtyupSitWFFQ61TfTdx4kSGDBnC/PnznS6lIOVyOa5ua2P/xkaGrVnDu33csc4fCOD1emltbSWVTm/19neGwxy3fv2OlitSdhQ0RIpEPJEg0jlEq7mMgZXL5UgkEmqb2kFqndp2tm3zve99j6effloBbRNeTSZ5Mh5nQVUVKwcPZl9X3/a4MYCKigosy6KlpYWsTg4X6RcKGiJFIJ3J0Nbaitfr1VyGAxLJZEfblNfrdClFbcNZL7pg3jbf+973iMViPPvss06XUnA+zWQYYpoc6nYzxLJwbcNKr2kYVFdVkcvlaGlp6ffh8JTCjJQhBQ2RAtd1XoZpUlFRobkMByQSCVwuFy61Te0Ql8uFz+fTisY2GjZsGMcffzz19fXkdLHa5fLWVq5tb2d1JsOwNWs4bN06srkc90UifH3dOkY2NnLs+vX8Lh7v+prXk0mGrVnD8mSSiU1N7L1+PecYBh8mEj3+Xf48EuGra9eyd2MjV7a1Ed/E9/2xWIyj1q9nRGMjR65bxyPRaNfnPu+s6fl4nFObmhjR2Miz3eoQKRcKGiKbkMvluDoS6er73a+xkRscujhq75zLqKqq0lyGA3J0BA0NgedHMBhU0NgOdXV1vPfee7z11ltOl1IwZoVCXB0I8BXTZOXgwSytqeG+aJSn4nFuD4V4raaGC/x+ftDWxhu9zqm5LRzmxmCQF2tqcJsmsyyLaDRKLB7n+Xic2ZEIP+78/FDTpD4W6/H1z8Zi3BEOc00gwJ8GDeLHwSA/i0R4stftbg2H+b7fz58GDeJonYUiZUgH9olswquZDE8mEjxTXc1uloVpGOxo08ywNWv4dWUlE7eh/SYWjxONRqmoqMDuY++x5FcymSSbzeJR21ReBINBtU5th6OOOooRI0bwyCOPcOihhzpdTkGoME2CpollGAyxLJK5HPdGIjxZXc0hnXNsu7tc/CWVYkEsxje6XehfEwx2ffyDQICzWlowvV7a2tqYZ5pM83qZ6vMBMDMY5M/JZI9VjdmRCDeGQkzqfF3YzbL4IJNhQSzGlM6vAzjf7++6jUg50oqGyCZ8ms326Psd3PkDbXP6o/c2ncnQ1taGz+vF1+0HV39L9XHXlnKRSCSwLAuXgl5ehEIhBY3tYJomZ555Ji+88ALrtfvRJjVkMsRyOU5vbmavxsauX0/H43yayfS47f7d/j8P7XxtTwaDuFwuPsxkGNNrw41Dun0czeX4JJPhyra2Ho9zdySy0eOM1uuGlDn9DxDp5fLbb+fJRAIMg2Fr1rCrZTHcsjjQ5WJW5645h61bx1Svl4ZMhhcTCSZ5PNxRUcFN7e0sTiRoyeXYyTSp9fm4NBDgsHXrADi3tRVaW9nVsvjL4MGbrWHDcOKDwKJkknhjIyd7vdQYBq8mk/xx0KCu2z4WizE3GuWzTIbhpsl5fj9ndxsYv6W9naWJBP/XGZ5O9Xq5IhDA7mzDujMcZmkiwbluN/dEo3xx4omsXrOmH76zxSdHx2ngPp9PszF5EgqF1Dq1nc444wx+9rOfsWjRIi6++GKnyyk4kc43fBZUVfGVXm8MuXu1nXaPERs+kwNqqqsx167d8uNkswDMDoU4uFcg6d3e6le7q5Q5BQ2RXmZdcgm7P/00C9Nplg4ahGUYXNDautHt5kajXBEIcGXnTjq/ikb5fTLJvMpKdrEs/jeT4X87fyAtralh1Nq1zKmoYLzbjbWVHz5t7e08n07zEHBbKMRY2+Y38TjzolGGdxtI3tAnfGsoxEG2zTupFFe1t+M3jK7l+6Bpck9FBUMti/fTaa5qayNoGFzcWTfAJ5kMi1MpHvL5sH75yx39FpaMVCrV0Tal+Yy8UevU9qupqeHkk09mwYIFXHjhhZhbWGUtR/tYFm7DYHUm06NNalsYhsHeLhd/S6U4rdtK8l+7rfTuZFkMNU0+zWQ4dQBXm0WKkYKGSC8VwSBBw+jq+92cI9xuLux2sb46FmMPy+Iw28YwDHbt9rWDOi8IKrdyn9AxlxGLxXjKspjmdnN65w+yK4NBliWTXe/aQd/6hC/vVuNwy+JCv5/fJhI9gkYKuDcQYFA6DXvssbVvUdlIxONYpomtc0vyJhQK8fnnnztdRtGqq6vj6aefZtmyZYwfP97pcgpK0DS5yO/nxnCYLHCYbdOey7EilSLY7c2XrTnf7+eHbW2Mtm3G2jbPxuP8M51mt26v3VcHg1zX3k6FaTLe7SaRy/F2Ok1LNsv0bq+tIuVOQUNkO43udfF5utfL6S0tjFu/nvEeD8e73XxrG98JzwGRSASfz0dDPM65vR5jjG2zvHP3lO59wle1tXXdJg1UdFsxeT4e51fRKJ9mMkRyOdJAqNeKyi6m2RWGpEOOjkMSPV6v2qbySK1TO+bggw/mgAMOoL6+XkFjE34UCDDINLkvEuGzbJYKw2CUy8Vl23Dxf7LXyyeZDDeHwyRyOU7yeKj1+Xit285V03w+fIbBA5EIs8Jh/HTMfZyvc45EelDQENlOvXtvR9k2/zNoEK8mk/wpmWR6aytHut08WFXV5/s06GiPMAC2sud6X/qE30qluKS1lauCQY52u6kwjK4WrC09F4F0Ok0mk1HbVJ6pdWrHGIZBXV0dP/7xj1m9ejW77LKL0yU56ny/v8fFvWEYfN/v5/ubueD/ptvN/w4d2uPPDrTtjf7sskBgo3ByXa/7OsXr5ZTN7Cg13LI2uk+RcqS3MEXyKGSanOz1MruigrmVlR2D4Z2BwDYMMlv5eugICYZhsJfLxcpeO0Ct3Eyf8AiXq8evDUv8K5JJdrUsfhgIMNq2Gely8UVnPbJl8Xgc0zRxa+/7vNKKxo479dRT8fv9LFy40OlSRES2SEFDJE/mRSL8Jh7no3Saf6XTvJBIMMQ0u9qYdjVNlieTNGYytPbhYv9cn4/H43GejMVoSKe5OxLhvXS6RxvP1cEg90WjPBSN8q90mvdSKRbFYsyLRADYw+VidSbDb+NxPkmneSgaZWki0R9Pv+QkEgk8Ho/apvJsw4qGTrjefoFAgNNOO43HHntM21GLSEFT0BDJk6Bpcn8kwsSmJk5sauLzTIaF3U7zvikUYlkyyaHr1nF8U9NW7+9Un49L/X5mhcOc0NTEZ5kMU3w+vN3anKb5fNxZUcETsRjjm5o4tbmZRbFY14rGCR4PF/j9/Gd7O8c3NbEileIKDSpuVTqTIZ1O6zTwfhAKhUilUiR7ndQs26a2tpa1a9eydOlSp0sREdksI6e3lUR6WrUKxo0DtxsKbLeh05ubGWKa3FdZmf87T6UgmYTly2HUqPzff18UyPc+HIkQiUQYMmTIwKxoFML3foC89NJLnH322bz99tsM3sJZMrJ1p5xyCpZl8fTTTztdSn4VyOvAgCuj1wEpH1rREClQsVyOeZEI/0yn+SidZnY4zJ+TyT5v0SjbT21T/ScYDAJoIDwP6urqeP311/nwww+dLkVEZJO065SIQ45ev54vMpseD/9Z59kYLyeT3BONksjl2NOy+FVlJUdqOLlfpTMZUqkUAW1T2S9CoRCABsLzYNKkSQwaNIj58+dz8803O12OiMhGFDREHLKwqor0ZjoXB5smXsPgyerqAa7KOd/97nc5cKedmLWZzw9bs4ZfV1YycTPbSfb2ejLJ/2tu5v2ddqJiG84ISSQSGIaBW/MZ/UJBI3/cbjdTp06lvr6eH//4x/gVjkWkwKh1SsQhu1rWRtvSbvgV1OF5G1k5eDDHDMDFfyIex+12dw3xS36pdSq/zjrrLMLhML/5zW+cLkVEZCO6mhGRojDEsnD388V/JpslmUrh7eOqSX8o9e1KtaKRX8OHD+fYY4+lvr5eWwaLSMFR65SIFIxsNsstiQSPRiK4DYNan48Zne+A926dejOZ5Jr2dj7KZNjP5eJyv59zW1v5Q00NB3bbqebtdJpbwmE+SKc50OXi7ooK9nRt+qUvkUhgQNdp4HdHIjwUjRLP5TjZ66XGMHg1meSPgwZ1fc1jsRhzo1E+y2QYbpqc5/dzdrcWllva21maSPB/2SxDTJNTvV6uCASwO0PTneEwS+NxznW5uGfaNL5oamL16tV5/b5uUUMDDODqgjuX46BcDusf/4B99hmwx91IMAgjRzr3+HlUW1tLbW0tK1euZMyYMU6XIyLSRUFDRArGUy+9xAXAkooK3szluLytjbG2zVG9Wqbas1lqW1s51u3mgcpKvshkuGEz75DfFg5zYzDIINNkZlsbV7S18XxNzSZvG4/HsTvbpp6NxbgnEuG2UIixts1v4nHmRaMM7zyjBODZWIw7wmFuDYU4yLZ5J5XiqvZ2/IbRtTtY0DS5p6KCoZbF++k0V7W1ETQMLu52nsknmQyLczke+slPsPbffwe/i9ugoQHGju3YVnOAGMCz4TD2T38Ks2cP2ONuxLZhxYqSCBvjx49n1113Zf78+QoaIlJQFDREpGDsv8cezGhuBstipG3zcCzGn5PJjYLGc/E4BjC7ogKPYbCPy8WX2SxXtbVtdJ/XBIN8o3Onrh8EApzV0kIil8PTqw0rm8uRSiYJVVQA8OtYjGleL6d3BoYrg0GWJZNEurWnzI5EuLFzhzCA3SyLDzIZFsRiXUHj8m6BYrhlcaHfz28TiR5BIwXc6/UyaO+9YSCDRjjcETJMEzazytMfkqaJYZp4nNpBLZ3ueN4lMidiWRZnnXUWd911FzfeeCNVVVVOlyQiAihoiEgBOWCPPeCtt7o+HmKarNtE3/nHmQwHuFw9wsKYzVwo79/tz4d2Dtmvy2bZpdvKBEAmnca0rK62qY/Saep6nVkyxrZZ3nmidTSX45NMhivb2noEnDRQ0a2u5+NxfhWN8mkmQySXIw2EeoWcXUyTQU5uAOByDejBaFnTJGOazh7GVmInk0+dOpXZs2ezaNEipk+f7nQ5IiKAgoaIFBBXr7BgwA4PuHa/lN1weZ/d1O1sm8GDB/f5kL5ItuNeZodCHNzrgnnDjlVvpVJc0trKVcEgR7vdVBhGVwtWd/4y2+HKNE0NLufZ4MGDmTRpEvPnz+f888/H1M51IlIA9EokIkVnT8vivXSaZLeL1ZXp9A7fb/fL/b1cLlb2ml3o/vFOlsVQ0+TTTGaj7Yl361wtWZFMsqtl8cNAgNG2zUiXiy+ym4o55cUwDLL6PuTd2WefTUNDA//93//tdCkiIoCChogUoVO8XrLA1W1tfJhO81oiwS86VwnytTZwrs/H4/E4T8ZiNKTT3B2J8F463eP+rw4GuS8a5aFolH+l07yXSrEoFmNeJALAHi4XqzMZfhuP80k6zUPRKEsTiTxVWLwMw9CKRj847LDD2Hfffamvr3e6FBERQEFDRIpQyDSZX1nJO+k0xzU1cVskwpWdw9W9h7y316k+H5f6/cwKhzmhqYnPMhmm+Hx4u93/NJ+POysqeCIWY3xTE6c2N7MoFuta0TjB4+ECv5//bG/n+KYmVqRSXNFtCLxcmQoa/cIwDGpra/n973/Pl19+6XQ5IiIYOb3ai/S0ahWMGwdut7PDqgMtleoYkF2+HEaNcqaGHfjePxuLcUV7O//caaceYSCfTm9uZohpcl9lZX7v2KnvvUP/1tvDYRLxOIMHDx6wx+yhEP6t95P29nbGjBnDRRddxIwZM5wuZ/voNbjk/l1K+dKKhogUpadiMf6STPJZJsOL8Ti3hMN82+PJW8iI5XLMi0T4ZzrNR+k0s8Nh/pxMdm1bK9vPMAyyeo+rX4RCIU499VQWLlxY8qfMi0jh065TIlKU1maz3BGJ0JjNMtQ0+bbXyzWdp4j3xdHr1/NFJrPJz/2s82yMl5NJ7olGSeRy7GlZ/KqykiOdOvuhhGhGo3/V1dWxcOFC/vCHPzBp0iSnyxGRMqagISJF6eJAoMehd9tqYVUV6c1c7A42TbyGwZPV1dt9/7J5G7a3zZG/4X35twMPPJBDDjmE+vp6BQ0RcZSChoiUpV17HdgnA8fobG/L5XJdv5f8qqur47LLLqOhoYGRI0c6XY6IlCnNaIiIyIDqCho6S6PffPvb36a6upr58+c7XYqIlDEFDRERGVAbgoYGwvuPx+PhjDPO4IknniAejztdjoiUKQUNEREZUKbZ8aNHA+H966yzzqK1tZXnn3/e6VJEpExpRkNERAZU9xkN6T8jRozgW9/6FvX19UyZMsXpcrZdOu10BQOr3J6vlAUFDRERGVCa0Rg4dXV1nHvuuaxatYpRxXIIXDDYcVDfhgPsyoltdzx/kRKhoCEiIgNKMxoD57jjjuMrX/kK9fX1zJ492+ly+mbkSFixAsJhpysZeMFgx/MXKREKGiIiMqAMwNShfQPC5XJx5pln8vOf/5wbbriBiooKp0vqG11si5QEDYOLbE463bF0Xy6/1B8sA8joPLRP+t+0adNIpVI8/fTTTpciImVGKxoivak/2OkqpAwYhkFWMxoDYujQoUycOJH6+nrOOeccHZIoIgNGQUOkN/UHO12FlAFDrVMDqra2lilTpvDGG2/wzW9+0+lyRKRMKGiIbIoutkX6lWY0BtYRRxzBnnvuyfz58xU0RGTAaEZDREQGnGGa2t52ABmGQV1dHUuWLKGxsdHpckSkTChoiIjIgDMMo8/b265IJjlm/Xp2W7OGc1pa+rewEjZlyhRcLhePP/6406WISJlQ0BARkQG3LTMaN4XDHOhy8T+DB3NPsWzPWoAqKir4zne+w8KFC8lkMk6XIyJlQEFDRET6RWoLQcLchu1tP8lkGOd28xXLosLUj60dUVdXx+rVq3nllVecLkVEyoBesUVEJC++29zMtW1t3NDezoFr1zK1pYU3kkkmNTWxe2MjX1u7llvb20nnch0rGtksyVyO69rbGbV2LSMaG5nc1MTKVAqAzzMZhq1ZQ3M2yxVtbQxbs4ZFsZjDz7K4jR49mtGjR1NfX+90KSJSBhQ0REQkb56Mx3EDz1dXMyMQ4MyWFr7mcvFyTQ23hUI8Ho9zTyTSNaMxKxxmcTzOPRUVvFRTw0jLYmpzMy3ZLMNMk5WDBxMyDGaFQqwcPJjJXq/TT7Ho1dXV8eqrr/Lpp586XYqIlDgFDRERyZuRlsV1oRB7ulwsSyYZZprcGgqxl8vFRK+XqwIB5kajAERzORZEo9wQCnGMx8M+Lhd3VFTgNQwei8WwDIMhloUBhDp/79Vhczts8uTJhEIhFixY4HQpIlLiFDRERCRvvmrbXb//MJ3mULe7x0nUh9k2kVyOL4EvgBQwttvX2IbBGNvmQw0r9xufz8fpp5/OE088QTKZdLocESlhChoiIpI3/j6uONi2TVVlZT9XI5tTW1tLU1MTv/vd75wuRURKmIKGiIj0i71dLt5MJnvsLvWXVIqgYbCLZbGX14ttGKzoHP6Gjp2qVqZS7GNZTpRcNvbcc0/GjRunoXAR6VcupwsQEdlIOu10BQOrRJ/v2T4fD0ajXNfezjl+Px9nMsyORJju92MaBn6gzudjVns7VZ3h44FIhFguxzSfz+nyS15tbS0XXHAB7733Hvvvv7/T5YhICVLQEJHCEQyCbUMqBeXWO27bHc+/hOxsWSysquLmcJiFTU1UGwZTvV5+GAh03ebaYJAscGlbG+FcjtEuF49XV1Op8zL63YQJExg6dCj19fXcdtttTpcjIiXIyPX1xCQRkYHQ0ADhsNNVDLxgEEaOHNjHXLUKxo0Dt7sj6JSLDUF2+XIYNcrpahx1xx138Mtf/pK//e1vBEss6IqI87SiISKFZaAvtkXK2Jlnnsm9997LM888Q11dndPliEiJ0dq0iIhImfrKV77C8ccfz/z581GDg4jkm4KGiIhIGaurq+O9997jzTffdLoUESkxChoiIiJl7Mgjj2TEiBHa6lZE8k5BQ0REpIyZpslZZ53FCy+8wPr1650uR0RKiIKGiIhImTvjjDMwDIMnnnjC6VJEpIQoaIiIiJS56upqJk+ezIIFC8hms06XIyIlQkFDREREqK2t5bPPPuO1115zuhQRKREKGiIiIsLBBx/MgQceqKFwEckbBQ0RERHBMAzq6up4+eWXWb16tdPliEgJUNAQERERAE455RT8fj8LFy50uhQRKQEKGiIiIgJAIBDgtNNO49FHHyWVSjldjogUOQUNERER6VJbW8u6detYunSp06WISJFT0BAREZEu++67L4cffriGwkVkhyloiIiISA91dXW88cYbfPDBB06XIiJFTEFDREREepg0aRKDBw9m/vz5TpciIkVMQUNERER6sG2bqVOn8tRTTxGNRp0uR0SKlIKGiIiIbOTMM88kHA7z3HPPOV2KiBQpBQ0RERHZyPDhwzn22GOpr68nl8s5XY6IFCEFDREREdmkuro63nnnHf72t785XYqIFCEFDREREdmko48+muHDh2soXES2i4KGiIiIbJJlWZx11ln89re/paWlxelyRKTIKGiIiIjIZp1xxhlks1kWLVrkdCkiUmQUNERERGSzBg8ezEknncT8+fPJZrNOlyMiRURBQ0RERLaorq6OhoYGli9f7nQpIlJEXE4XICIiDkunna5gYJXb882Dww47jH333Zf6+nqOOuoop8sRkSKhoCEiUq6CQbBtSKUgmXS6moFl2x3PX/rEMAzq6uq4/vrr+fLLL9l5552dLklEioCR0yk8IiLlq6EBwmGnqxh4wSCMHOl0FQNrB/+uI5EIp512Gqeffjp1dXV5LKyflePftUiBUNAQEREpdQ0NMHZsx+rVDojH46TTaQKBAIZh5Km4fmbbsGKFwoaIA9Q6JSIiUurC4Y6QYZrg2v4f/ZZlEW5pwTYMPG53HgvsJ+l0x/Mux1U7kQKgoCEiIlIuXK6Od/i3k23bGG43kWQSTyCQx8L6UbnNH4kUEG1vKyIiIn3m9/tJJpOkMxmnSxGRAqegISIiIn3m9XoxTZNYNOp0KSJS4BQ0REREpM8MwOfzEYvH0X4yIrIlChoiIiKyTXw+H9lslngi4XQpIlLAFDRERERkm7gsC4/HQ1TtUyKyBQoaIiIiss18Ph+pVIpUOu10KSJSoBQ0REREZJt5PB4sy9KqhohsloKGiIiIbLMNQ+HxeJyshsJFZBMUNERERGS7+Hw+yOWIxWJOlyIiBUhBQ0RERLaLZZp4vF5isRha0xCR3hQ0REREZLv5fT7S6TSpZNLpUkSkwChoiIiIyHaz3W5cLhdRtU+JSC8KGiIiIrLdNgyFJ+JxMtms0+WISAFR0BAREZEd4vP5wDA0FC4iPShoiIiIyA4xDQOvhsJFpBcFDREREdlhfr+fTCZDIpFwuhQRKRAKGiIiIrLDbJcL27bVPiUiXRQ0REREJC/8fj+JRIJ0JuN0KSJSABQ0REREJC+8Hg+mhsJFpJOChoiIiOSFYRh4fT4NhYsIoKAhIiIieeT3+8lms8TjcadLERGHKWiIiIhI3rgsC7fbTTQadboUEXGYgoaIiEiJWbRoEfvtt59jj+/3+0mlUqTSacdqEBHnKWiIiIhIXnk8HizTLNqh8FwuR1ohSWSHKWiIiIgUmFdffZXJkyez3377ccABB1BbW8snn3wCwOuvv86wYcNoa2vruv27777LsGHD+Pzzz3n99de54ooraGtrY9iwYQwbNow76+sBaM1muay1lf0aG9mjsZHvNTfT0McL6uZslotaWxmzdi17NDZyzPr1/KbXHEY2l+OBSIQj1q3jkEyGI8Nh7g6Huz7/f5kMF7W2sn9jI3s2NjKxqYm/plJdn/99IsEJ69czorGRw9et485wmHSuY6w8l8txZzjMoevWsXtjI2PWruW69vaur30kGuWb69YxorGRr65dy/ktLV2fS6ZSXHfddYwaNYoRI0YwefJkVq5c2fX5Dd/TV155hQkTJjBixAieeeYZdtllF/7+97/3eI4PPvggY8eOJZvN9un7JlLOFDREREQKTDQaZfr06bz44os89dRTGIbBeeed16eL27FjxzJr1ixCoRArV65k5cqVXHjaaQD8MBLh7+k09VVV/K66mhzwvZYWUrmt7xGVyOX4qsvFwqoqXq2p4Uyfj0tbW/lbt6Dw03CY+yIRLg8EeHXQIG4zDCo7z9SIZLOc2tzMl5kM9VVVvFxTw8V+f9fuVP+TTHJZayvf9/tZVlPDz0IhnozHuScSAWBxIsEvo1F+Fgrx+qBB/Lqqiv1dLgD+nkpxfXs7PwoGWT5oEI9VVXG4291V183z5rF48WLuueceXnrpJUaOHMnUqVNp6RZGAG699VauvfZali1bxoQJEzjyyCNZtGhRj9s88cQTnH766ZimLqFEtsbldAEiIiLS00knndTj4zlz5nDQQQfxwQcfbPVrbdsmFAphGAZDhgzp+MM1a2jIZnkpleL56moO7bwIv7+ykkPWrePFRIJve71bvN+dLYuLAoGuj891uXgtmeSFeJwxtk04m+VXsRg/DYWY4vMBMMjrJZ1OkwOei8dZn82ytKaGqs6L9BGuf1+G3BmJ8INAoOtrd3e5+FE2yy3hMDOCQVZnMuxkmhzpdmMbBrtYFmNsG4DVmQx+w+A4t5ugabKrZXGQbUMqRTSXY/4LL3D3/fdzzDHHAHDHHXewbNkyHnvsMS6++OKuGn70ox9x1FFHdX08bdo0Zs6cyU033YTb7WbVqlW8//77PPLII1v9exARrWiIiIgUnIaGBi666CIOP/xw9tlnHw477DAAVq9evd33+WE2iwu6Ls4Bqk2TPS2LD/vQPpXJ5ZgTDnPM+vXs39jIXo2NvJZIsLpzleXDTIZkLse4bisJPr+fdDpNKpXi3XSag1yurpDR2z/Sae6KRNir8773amzkqvZ21mSzxHI5vu31EgcOX7eOq9raWBqPd7VVHeV2s6tlcfj69Vza2sqzsRixzs99ms2SSqcZO3Zs12PZts2YMWP48MMPe9QwevToHh+feOKJWJbF0qVLgY4h+yOOOILhw4dv9fslIlrREBERKTi1tbXsuuuuzJ49m6FDh5LNZhk/fjypVIpA56pCrlu7U6pb+1J/+UU0yq9iMWYFg+zvcuE3DG4Ih0l21rGp9RC32011dTW2bePdyrkakVyOqwMBJnk8G33OAwyzLJYPGsSfk0mWJZP8uL2dB6JRnq2uJmia/L6mhjdSKZYlEtwRiTA7EuHFUGibnqPf7+/xsW3bnHbaaTzxxBNMmjSJ5557jptvvnmb7lOknGlFQ0REpIA0Nzfz8ccfc/nllzNu3Dj23ntvWltbuz4/aNAgANasWdP1Z++++26P+3C73WQ6ZyM22Ns0SUOPmYrmbJaPMxn2cW39fccVqRQTPB6+6/NxgG2zm2XxcbeVkD1cLryGwfJksuvPDMDjdmMA+7tcvJtO07KZOZNRLhcfZTKMcLk2+mUaBgBew+B4j4dbQiGeqa7mrVSK9ztrcBkGR7rdXBcK8fKgQXyRybA8nWZ308R2uVixYkXXY6VSKVauXMk+++yz1ec9bdo0/vznP/PII4+QyWSYNGnSVr9GRDpoRUNERKSAVFZWUl1dzcKFCxk6dChffPEFP/3pT7s+P2LEiI6dpO68k2uuuYaPP/6YuXPn9riP4cOHE4lEWL58OQcccAC+eJyRpskE2+aq9nZ+FgoRNAxuDYfZ2TSZsIlVhN5GWha/SyR4M5mk0jSZF42yLptlw6W6xzC4xO/n5vZ2bGCsbbM+l+ODdJqpPh+neL3cG4lwTksL/xkMMsQ0eSedZmfL4hDb5spAgNqWFnYxTf7D68Wko53q/XSamcEgi2IxsnS0fvmAZ+JxvIbBrpbFHxIJPstkONy2qTRNXk4kyAJ7mib+bJa6k09m1qxZVFVVscsuu/DAAw8Qi8WYNm3aVp/33nvvzcEHH8wtt9zC1KlT8W5llkVE/k0rGiIiIgXENE3mzp3L22+/zdFHH82NN97I9ddf3/V527b5xS9+wUcffcQxxxzD/fffz8yZM3vcx6GHHkptbS3Tp0/noIMO4oHOnZPuDgT4qstFbUsL/9HcTA54tKoKu3PFYEsuDwQY5XIxtaWF7zY3M8Q0mdgroFwRCHBhIMAdkQjfamriwtZW1nWuYNiGwRPV1Qw2Tc5saeGYpiZ+Ho12XYgc7fEwv6qKZckkJzY18R9NTfwyGmVXywKg0jB4NBZjclMTxzY18edkkvlVVVSbJpWGwZJEgv/X3MxR69czPxbjgcpK9u1cqbn2/PM56aSTuPTSSznhhBNoaGjg8ccfp7Kysk9/J9OmTSOVSnHGGWf06fYi0sHI5fqwp52IiIgUr1WrYNw4cLuh2zB4yUulIJmE5cth1Kjtvps5c+bwu9/9jpdffjmPxYmUPq1oiIiIiGxCJBLh/fff5+GHH+bcc891uhyRoqMZDREREeF7zc38z2Z2r7osEOCybmdolItrr72W5557jhNPPFFtUyLbQa1TIiIipa4PrVNfZjLEN3NJUGWamz3/oqDlqXVKRLaPVjRERESEnTuHrkVE8qUI354QEREREZFCp6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ5p6AhIiIiIiJ553K6ABERERkg6bTTFQyscnu+IgVGQUNERKTUBYNg25BKQTLpdDUDy7Y7nr+IDDgjl8vlnC5CRERE+llDA4TDTlcx8IJBGDnS6SpEypKChoiIiIiI5J2GwUVEREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO8UNEREREREJO/+P0WxMmR6iaiCAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 1000x1000 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,10))\n",
    "plt.axis('off')\n",
    "\n",
    "nx.draw_networkx(G,with_labels=True,node_size=2000,\n",
    "                 font_size=10,\n",
    "                 node_shape='s',\n",
    "                 alpha=0.9,\n",
    "                 node_color='red')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1b34436a",
   "metadata": {},
   "source": [
    "Можно посчитать близость между синсетами (зависит от того, какой путь нужно пройти по этим связям от одного объекта до другого и есть ли он вообще)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "aa6f95a4",
   "metadata": {},
   "outputs": [],
   "source": [
    "car = wn.synsets('car')[0]\n",
    "bike = wn.synsets('bike')[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "a6af6e45",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.3333333333333333"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bike.path_similarity(car)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "151c9c0f",
   "metadata": {},
   "source": [
    "Есть и другие метрики:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "9f01bcf7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.538973871058276"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wn.lch_similarity(bike, car)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "a675de99",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9166666666666666"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wn.wup_similarity(bike, car)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0693e61e-495e-44fc-95f6-316d9732dff6",
   "metadata": {},
   "source": [
    "### использование WordNet для WSD"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8aac2778",
   "metadata": {},
   "source": [
    "Датасеты для обучения и оценки WSD моделей обычно используют смыслы из WordNet. Например, http://lcl.uniroma1.it/wsdeval/ Там очень много данных, поэтому я взял небольшой кусок"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "ad1d8e05",
   "metadata": {},
   "outputs": [],
   "source": [
    "corpus_wsd = []\n",
    "corpus = open('data/corpus_wsd_50k.txt').read().split('\\n\\n')\n",
    "for sent in corpus:\n",
    "    corpus_wsd.append([s.split('\\t') for s in sent.split('\\n')])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "936b2d5b",
   "metadata": {},
   "source": [
    "Корпус состоит из предложений, где у каждого слова три поля - значение, лемма и само слово. Значение пустое, когда слово однозначное, а у многозначных слов стоит тэг вида **'long%3:00:02::'** Это тэг wordnet'ного формата"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "7239f516",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[['', 'have', 'Have'],\n",
       " ['', 'you', 'you'],\n",
       " ['permit%2:41:00::', 'permit', 'permitted'],\n",
       " ['', 'it', 'it'],\n",
       " ['', 'to', 'to'],\n",
       " ['become%2:42:01::', 'become', 'become'],\n",
       " ['', 'a', 'a'],\n",
       " ['giveaway%1:21:00::', 'giveaway', 'giveaway'],\n",
       " ['program%1:09:01::', 'program', 'program'],\n",
       " ['rather%4:02:02::', 'rather', 'rather'],\n",
       " ['', 'than', 'than'],\n",
       " ['', 'one', 'one'],\n",
       " ['', 'that', 'that'],\n",
       " ['have%2:42:00::', 'have', 'has'],\n",
       " ['', 'the', 'the'],\n",
       " ['goal%1:09:00::', 'goal', 'goal'],\n",
       " ['', 'of', 'of'],\n",
       " ['improved%3:00:00::', 'improved', 'improved'],\n",
       " ['employee%1:18:00::', 'employee', 'employee'],\n",
       " ['morale%1:26:00::', 'morale', 'morale'],\n",
       " ['', 'and', 'and'],\n",
       " ['', ',', ','],\n",
       " ['consequently%4:02:00::', 'consequently', 'consequently'],\n",
       " ['', ',', ','],\n",
       " ['increased%3:00:00::', 'increased', 'increased'],\n",
       " ['productivity%1:07:00::', 'productivity', 'productivity'],\n",
       " ['', '?', '?']]"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corpus_wsd[1]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2885c3c5",
   "metadata": {},
   "source": [
    "Вот так можно достать соответствующий этому тэгу синсет:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "3aafc59d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Synset('program.n.02')"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wn.lemma_from_key('program%1:09:01::').synset()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "8633e438",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'a system of projects or services intended to meet a public need'"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wn.lemma_from_key('program%1:09:01::').synset().definition()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "af7341b8",
   "metadata": {},
   "source": [
    "### Алгоритм Леска"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "63bae13a",
   "metadata": {},
   "source": [
    "Как теперь разрешить предсказать правильный смысл для каждого такого случая? Классический способ это сделать - алгоритм Леска, который основан на уже существующей базе значений и их определений (например, на wordnet). Алгоритм у него достаточно простой: 1) для неоднозначного слова находятся определения всех его значений; 2) выбирается контекст, в котором встретилось неоднозначное слово - обычно это какое-то окно вокруг слова (5 слов слева + 5 слов справа); 3) контекст сравнивается с каждым из определений и выбирается значение с наибольшей близостью к контексту. \n",
    "\n",
    "\n",
    "Сама имплементация этого алгоритма будет вашим домашним заданием, но мы разберем ключевой пункт этого алгоритма - расчет близости контекста и определения."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "82263be4",
   "metadata": {},
   "source": [
    "Рассмотрим несколько способов."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2f30a022",
   "metadata": {},
   "source": [
    "## Пересечение слов\n",
    "\n",
    "Самый простой способ - это просто посчитать общие слова. Чем больше общих слова - тем более похожи определение и контекст. Такое удобно написать через множества (set) в питоне"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "id": "f304c31e",
   "metadata": {},
   "outputs": [],
   "source": [
    "definition1 = 'a system of projects or services intended to meet a public need'\n",
    "definition2 = '(computer science) a sequence of instructions that a computer can interpret and execute'\n",
    "context = 'i wrote a computer _ that predicts the sense of a word' # само слово не program не используем"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "id": "a93df84c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'a', 'of'}"
      ]
     },
     "execution_count": 158,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(set(definition1.split()) & set(context.split()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "id": "e1f94102",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'a', 'computer', 'of', 'that'}"
      ]
     },
     "execution_count": 159,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(set(definition2.split()) & set(context.split()))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9b43e4eb",
   "metadata": {},
   "source": [
    "Можно еще конечно почистить от стоп-слов, но уже видно, что это как-то работает. Единственный недостаток такого подхода в том, что он зависит от длины текстов. Можно представить ситуацию, где определение состоит из трех слов и все они есть в контексте, но при этом в другом определении, состоящим из 30 слов, есть 4 пересечения. Хотелось бы метрику, которая устраняет влияние длины текста."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4f1046ce",
   "metadata": {},
   "source": [
    "## Jaccard\n",
    "\n",
    "Пример такой метрики - метрика Жаккара (Jaccard). В ней к пересечению еще добавляется объединение и итоговое значение - это отношение пересечения к объединению. Логика тут в том, чтобы оценить какая доля слов пересекается. Значение будет от 0 до 1, так как пересечение в худшем случае равно 0, а в лучшем равно объединению. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "id": "e0c92e46",
   "metadata": {},
   "outputs": [],
   "source": [
    "definition1 = 'a system of projects or services intended to meet a public need'\n",
    "definition2 = '(computer science) a sequence of instructions that a computer can interpret and execute'\n",
    "context = 'i wrote a computer _ that predicts the sense of a word'\n",
    "\n",
    "intersection1 = (set(definition1.split()) & set(context.split()))\n",
    "union1 = (set(definition1.split()) | set(context.split()))\n",
    "jaccard1 =  len(intersection1) / len(union1)\n",
    "\n",
    "intersection2 = (set(definition2.split()) & set(context.split()))\n",
    "union2 = (set(definition2.split()) | set(context.split()))\n",
    "jaccard2 =  len(intersection2) / len(union2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "id": "3a348e87",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.1, 0.21052631578947367)"
      ]
     },
     "execution_count": 161,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "jaccard1, jaccard2"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "61df87a6",
   "metadata": {},
   "source": [
    "### Косинусная близость"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ba750b88",
   "metadata": {},
   "source": [
    "Другая метрика, которую нужно разобрать - косинусная близость (или косинусное расстояние). В каком-то смысле это то же самое, что и Жаккар, но обобщенное на работу с векторами. Саму векторизацию мы пока пропустим и представим, что они у нас уже есть."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2fa0fd5a",
   "metadata": {},
   "source": [
    "Есть геометрическое объяснение косинусного расстояния/близости: угол между двумя одинаковыми векторами равен 0, а косинус 0 - равен 1; угол между перпендикулярными векторами векторами равен 90 градусов, а косинус - 0; угол между векторами направленными в противоположные стороны равен 180, а косинус -1. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5f622418",
   "metadata": {},
   "source": [
    "Картинка для примера:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "19a92ccd",
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.display import Image\n",
    "from IPython.core.display import HTML "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "0dc51d38",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src=\"https://i.ibb.co/2qFsPLQ/Screenshot-2022-11-16-at-15-36-40.png\" width=\"500\" height=\"500\"/>"
      ],
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(url=\"https://i.ibb.co/2qFsPLQ/Screenshot-2022-11-16-at-15-36-40.png\",\n",
    "     width=500, height=500)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4bd5385f",
   "metadata": {},
   "source": [
    "Допустим в нашем словаре 2 слова: Hi и Mark. На графике изображены три текста: \"Hi\", \"Mark\", \"Hi Mark\". Между Hi и Mark нет ничего общего, они перпендикулярны, поэтому их близость равна 0, а между Hi Mark и Mark уже есть общее, поэтому близость будет ненулевая. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "016f52ae",
   "metadata": {},
   "source": [
    "Но геометрическое объяснение сложновато обобщить на реальные примеры. Уникальных слов обычно сильно больше 2 и, соответственно, нужно к графику выше мысленно добавлять еще сотни или даже тысячи размерностей. Нарисовать такое или даже представить сложновато. \n",
    "\n",
    "Поэтому давайте лучше посмотрим на алгебраическую интерпретацию."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5dc9ec1c",
   "metadata": {},
   "source": [
    "Формула вот такая:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "b2fd8228",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,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\" width=\"500\" height=\"500\"/>"
      ],
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(url=\"data:image/png;base64,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\",\n",
    "     width=500, height=500)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "14f9d44c",
   "metadata": {},
   "source": [
    "В числителе (вверху) здесь скалярное произдведение - это такая операция с векторами, когда числа, стоящие в двух векторах напротив друг друга перемножаются, а результат складывается. \n",
    "\n",
    "В знаменателе (внизу) тут произведение длин векторов, а длина вектора считается как произведение корней из суммы квадратов значений в первом и втором векторах.\n",
    "\n",
    "Вот пошаговая визуализация расчета близости для двух векторов с абсолютными частотностями. Вектора 1 и 2 здесь это представление двух текстов мешком слов - это такой метод, который показывает какие слова из словаря встречаются в каждом из текстов и в каком количестве, но порядок слов при этом теряется. В этом примере в словаре 6 уникальных слов, поэтому размерность векторов - 6. Каждая из размерностей показывает частотность каждого из слов в словаре в конкретном тексте. Можно увидеть, что первому вектору соответствует текст из 3 слов, которые встретились по 1 разу, а второму вектору соответствует текст тоже из 3 уникальных слов, два из которых такие же как и в первом тексте, но первое с другой частотностью."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "0c5bf47a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src=\"https://i.ibb.co/Qn87BXW/Screenshot-2022-11-16-at-15-46-54.png\" width=\"700\" height=\"700\"/>"
      ],
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(url=\"https://i.ibb.co/Qn87BXW/Screenshot-2022-11-16-at-15-46-54.png\",\n",
    "     width=700, height=700)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "229f585b",
   "metadata": {},
   "source": [
    "Выглядит наверное сложно, но важно понять, что тут происходит. \n",
    "\n",
    "Вверху (скалярное произведение) по сути происходит поиск общих слов в первом и втором векторе (=тексте). Если слово есть в одном тексте, а в другом нет - то результат произведения будет нулевой. Чем больше пересечений, тем больше будет итоговое скалярное произведение. **Чем больше значение скалярного произведения, тем больше будет итоговая близость.**\n",
    "\n",
    "Внизу - просто нормализация. Результат скалярного произведения может быть любым (отрицательным, положительным, большим, маленьким), деление на длину вектра приводит итоговую метрику в удобный интервал от -1 до 1 (а в нашем случае без отрицательных значений от 0 до 1). \n",
    "\n",
    "Почему внизу не просто сумма? Потому что значения могут быть отрицательными и может получиться ноль. Для обхода проблемы с нулями часто используется корень из квадрата \n",
    "(почему не абсолютное значение - как я понимаю сложный математический вопрос, часто пишут что абсолютное значение сложнее математически, а квадрат проще)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "925d8402",
   "metadata": {},
   "source": [
    "Можно заметить, что это очень похоже на Жаккара. И для этих векторов (посчитанных мешком слов) мы смогли бы посчитать и Жаккара - пересечение это количество ненулевых размерностей, которые есть в обоих векторах, а объединение - это сумма ненулевых размерностей в обоих векторах. Но косинусное расстояние подходит еще и для векторов, которые построены не напрямую из слов. Это могут быть вектора из трансформерной модели, где каждая размерность может принимать как отрицательные так и положительные значений и как правило нулевых размерностей там нет. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "606558d6",
   "metadata": {},
   "source": [
    "В таких скалярное произведение будет штрафовать пары значений с разными знаками (будет получаться отрицательное число и сумма будет уменьшаться)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9bc83cb5",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "d9cb2f4c",
   "metadata": {},
   "source": [
    "## Sentence transformers\n",
    "\n",
    "Давайте посмотрим как считать близость между реальными векторами текстов. Для этого забежим вперед и возьмем предобученную трансформерную модель, которая умеет создавать эмбеддинг для небольшого текста. \n",
    "\n",
    "Модель возьмем из библиотеки [sentence_transformers](https://www.sbert.net/index.html). Это библиотека содержит большое количество моделей, которые предназначены для построение эмбединнгов текстов. Многие из них - специально обученые производить хорошие эмбединги, между которыми удобно рассчитывать близость. На сайте библиотеки есть раздел с лидербордом по моделям. Мы возьмем лучшую (по усредненному скору) модель - `all-mpnet-base-v2`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "27cb1c61-134a-4e48-9fe6-2e8a49ff384f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# библиотеки могут заработать не сразу\n",
    "# можете попробовать установить более новую версию питона если инсталлов ниже не хватает"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "8062ee1f",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "!python -m pip install torch torchvision torchaudio\n",
    "!python -m pip install sentence_transformers transformers accelerate -U"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "ce73180c",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics.pairwise import cosine_distances, cosine_similarity\n",
    "from sentence_transformers import SentenceTransformer"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aac0ae91-31bf-4dba-9d4e-24ed9bbe7061",
   "metadata": {},
   "source": [
    "Интерфейс у модели достаточно простой, есть функция encode, которая сама предобрабатывает текст и превращает его в 1 вектор"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "3de62fae",
   "metadata": {},
   "outputs": [],
   "source": [
    "# embedding model\n",
    "model = SentenceTransformer('sentence-transformers/all-mpnet-base-v2')\n",
    "embed = model.encode"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "52fa6009-48c0-4135-ad7a-7913d3cfe088",
   "metadata": {},
   "source": [
    "Рассчитаем вектора для определений и контекста, которые мы использовали выше"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "14451973",
   "metadata": {},
   "outputs": [],
   "source": [
    "definition1 = 'a system of projects or services intended to meet a public need'\n",
    "definition2 = '(computer science) a sequence of instructions that a computer can interpret and execute'\n",
    "context = 'i wrote a computer _ that predicts the sense of a word'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "57044bab",
   "metadata": {},
   "outputs": [],
   "source": [
    "definition1_emb = embed(definition1)\n",
    "definition2_emb = embed(definition2)\n",
    "context_emb = embed(context)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a2cd2860-f995-4d08-b31a-1bfa518eb999",
   "metadata": {},
   "source": [
    "Посмотрим на получившийся вектор. В нем 768 не очень больших, положительных и отрицательных чисел"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "3989291a-2f00-487b-a4b1-786fbeff0834",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(768,)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "context_emb.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "cbd2ce6a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 2.60696858e-02,  2.81297602e-02, -2.86605358e-02,  2.73304842e-02,\n",
       "       -2.03045104e-02, -2.45557521e-02, -3.44669889e-03,  1.29649406e-02,\n",
       "        3.22340988e-02, -1.32162180e-02,  3.78281027e-02, -4.27398421e-02,\n",
       "        1.42525053e-02,  8.37171823e-02,  3.09127476e-02,  1.17423786e-02,\n",
       "        6.91081583e-02,  1.30820498e-02, -1.07209831e-02, -2.10680906e-03,\n",
       "        4.95719211e-03, -1.02009624e-03, -5.77510986e-03,  4.73414510e-02,\n",
       "       -3.00503504e-02, -3.59182991e-02, -2.45586336e-02, -2.17193905e-02,\n",
       "       -2.08961219e-02, -2.29210742e-02, -2.80032977e-02,  1.52165564e-02,\n",
       "       -1.65842436e-02,  4.57603335e-02,  1.44828766e-06, -5.61360009e-02,\n",
       "       -3.94317359e-02, -1.73282661e-02,  1.10615166e-02, -2.93712430e-02,\n",
       "        8.55174810e-02,  3.19165811e-02,  4.22134530e-03, -2.52286508e-03,\n",
       "       -6.43932000e-02,  4.01510037e-02,  4.70006950e-02, -4.48253714e-02,\n",
       "       -1.41513813e-03,  3.66889797e-02, -8.11715703e-03, -4.46259156e-02,\n",
       "        2.88627185e-02,  2.14875434e-02,  5.65632023e-02, -7.84562319e-04,\n",
       "        5.81800705e-03,  2.19479781e-02,  3.44444998e-02, -5.85746132e-02,\n",
       "       -3.14422622e-02,  9.82944295e-03, -2.18724068e-02,  2.72080489e-02,\n",
       "        5.10382988e-02,  2.24636458e-02,  2.58076116e-02, -6.15553483e-02,\n",
       "       -5.12524135e-03, -2.24703737e-02,  2.49784882e-03, -3.52916196e-02,\n",
       "       -1.16497632e-02,  2.93239839e-02, -3.97467874e-02, -2.59171855e-02,\n",
       "        1.10591054e-02, -1.20284809e-02, -1.90009009e-02, -2.27706600e-02,\n",
       "        6.18728735e-02,  1.14581380e-02, -1.07487626e-02,  4.07132506e-03,\n",
       "       -3.67226219e-03,  8.43563527e-02, -2.39394954e-03, -3.37652229e-02,\n",
       "        1.20446840e-02,  2.25450788e-02,  1.10169696e-02, -1.64049696e-02,\n",
       "        5.59397414e-02,  1.17083488e-03, -9.31880530e-03,  2.62642857e-02,\n",
       "       -2.74639539e-02, -2.27501988e-02,  8.93268362e-03, -7.66405538e-02,\n",
       "        3.55793536e-02,  1.20696472e-02,  5.86612076e-02,  3.63901295e-02,\n",
       "       -1.65351499e-02, -5.73400129e-03,  1.45409070e-02,  1.96258314e-02,\n",
       "       -7.27106109e-02,  2.72692945e-02, -8.13957080e-02, -2.11496335e-02,\n",
       "       -3.26966271e-02,  7.29898438e-02,  3.90958190e-02, -4.96897846e-02,\n",
       "       -6.63388595e-02, -1.01196701e-02,  6.42161875e-04, -1.88012272e-02,\n",
       "       -1.35782529e-02,  1.69789605e-02, -9.91971418e-03,  2.28155497e-02,\n",
       "       -4.12469879e-02, -3.77093293e-02, -7.57177593e-03,  2.83327959e-02,\n",
       "       -3.02903876e-02, -2.85167098e-02,  2.07434315e-02,  2.25604214e-02,\n",
       "       -9.28365719e-03, -1.51147833e-02, -8.98459088e-03,  6.70140386e-02,\n",
       "       -1.28544942e-02, -5.78452088e-02,  2.47775298e-03, -7.97296911e-02,\n",
       "       -2.81679835e-02, -2.84210779e-02, -3.25452425e-02, -6.56164214e-02,\n",
       "       -2.47204620e-02,  2.01430097e-02,  1.01629477e-02, -5.48637062e-02,\n",
       "        8.94648675e-03,  4.05956618e-02, -4.44052890e-02,  1.02671042e-01,\n",
       "       -7.56809115e-02,  1.26552181e-02,  1.37866579e-03, -1.13331880e-02,\n",
       "        1.51405660e-02,  2.56871488e-02, -5.24983183e-03,  4.51805294e-02,\n",
       "        4.59146686e-02,  5.85043104e-03,  6.26491150e-03,  7.20547792e-03,\n",
       "       -5.88717051e-02, -1.00635946e-01, -6.11553602e-02, -1.08829951e-02,\n",
       "       -5.83557002e-02,  4.36093472e-02, -3.92436385e-02,  4.10913564e-02,\n",
       "        5.64262178e-03,  5.37047759e-02,  2.42957957e-02,  1.05013326e-01,\n",
       "        6.05127029e-02,  2.84899641e-02,  5.55243762e-03, -3.09707951e-02,\n",
       "       -1.55448066e-02, -1.46294916e-02, -1.43787283e-02, -3.49355722e-03,\n",
       "       -8.83707330e-02,  2.19522649e-03, -2.93946192e-02, -2.35651769e-02,\n",
       "        1.60144772e-02, -5.43737002e-02, -3.02929245e-02, -7.79132126e-03,\n",
       "        1.89319085e-02, -3.13578583e-02,  4.60174680e-02, -5.50660631e-03,\n",
       "       -1.92169063e-02, -4.88135181e-02,  8.63954145e-03, -4.77152430e-02,\n",
       "        2.17958838e-02, -4.88891192e-02,  7.63864294e-02,  1.30082771e-01,\n",
       "        3.28815654e-02, -5.33933975e-02,  8.30223709e-02,  7.57148629e-03,\n",
       "       -2.32078675e-02,  4.61315811e-02,  2.62386221e-02,  1.36163775e-02,\n",
       "       -4.62059826e-02, -2.14045327e-02, -1.82300620e-02,  2.83758324e-02,\n",
       "        1.28799668e-02,  2.15108078e-02, -4.15302766e-03,  3.15550305e-02,\n",
       "        1.82425436e-02, -3.74762639e-02,  2.08042860e-02,  2.08608937e-02,\n",
       "       -2.82959430e-03, -5.35163693e-02, -4.71888147e-02,  1.35965738e-02,\n",
       "        1.38235912e-02,  9.01751686e-03, -3.65398708e-03,  5.73994629e-02,\n",
       "        2.57573053e-02, -2.01233313e-03, -5.14804944e-02, -2.00652517e-02,\n",
       "       -4.66603264e-02,  3.39639969e-02,  1.17061762e-02, -5.82824051e-02,\n",
       "        2.77182721e-02,  3.31477113e-02,  6.75387383e-02,  1.22633018e-02,\n",
       "        5.67695983e-02, -6.65890053e-02,  5.69646060e-02, -7.09204301e-02,\n",
       "        3.76802348e-02, -1.43325031e-02,  4.47984375e-02,  7.28084426e-03,\n",
       "        2.85710488e-02,  1.60099398e-02,  3.72759327e-02, -2.62777582e-02,\n",
       "       -2.46258285e-02, -2.06304975e-02,  3.72276478e-03, -4.67101634e-02,\n",
       "        8.38323124e-03, -1.35120293e-02,  3.39612849e-02,  3.49291484e-03,\n",
       "        4.99470299e-03, -2.49034129e-02,  7.93784333e-04, -3.46210860e-02,\n",
       "       -4.31761034e-02,  3.18461545e-02, -4.72946931e-03,  1.70705188e-02,\n",
       "       -2.31329408e-02, -4.39915759e-03,  1.13313077e-02,  2.25792192e-02,\n",
       "        3.83138545e-02, -1.36187421e-02,  1.68776456e-02,  3.40576060e-02,\n",
       "       -3.42961587e-02, -2.40255930e-02, -4.41743918e-02, -3.07683609e-02,\n",
       "       -4.96942215e-02,  5.87186776e-02,  3.34627815e-02, -3.17160413e-02,\n",
       "       -7.49719329e-03,  3.18871289e-02, -2.39836518e-02,  2.70108543e-02,\n",
       "        2.77622230e-02, -4.32922244e-02,  2.50458159e-02, -2.90878844e-02,\n",
       "       -5.16465493e-03, -3.24391341e-03,  1.94145795e-02,  4.15609553e-02,\n",
       "        4.53335270e-02, -4.18853667e-03,  8.84133205e-03, -3.26370895e-02,\n",
       "        2.16132705e-03,  2.40324698e-02,  5.22514284e-02, -3.98226120e-02,\n",
       "       -4.63361889e-02,  3.35683487e-02,  8.99985433e-02,  2.09509823e-02,\n",
       "       -3.21147405e-02,  4.97947820e-02, -5.90330772e-02, -3.41108479e-02,\n",
       "       -7.22306669e-02, -1.34600475e-02, -4.47860770e-02,  6.01749718e-02,\n",
       "       -3.98624782e-03, -5.05394153e-02,  8.97321180e-02, -7.15939677e-04,\n",
       "        1.99675821e-02,  3.08200088e-03, -2.99170837e-02,  2.10373215e-02,\n",
       "        7.66325668e-02, -2.41141766e-02,  1.48944026e-02,  2.33445171e-04,\n",
       "       -3.57108861e-02, -2.00601127e-02, -2.78608240e-02,  3.36323306e-02,\n",
       "       -6.52984530e-02,  2.57370248e-03, -2.62748804e-02,  2.23488789e-02,\n",
       "        1.96563937e-02,  2.94369552e-02, -1.44276265e-02, -2.28903387e-02,\n",
       "       -7.76160210e-02,  1.66622142e-03,  8.07723925e-02,  3.26537266e-02,\n",
       "       -2.33838372e-02,  4.45028879e-02,  1.28065404e-02,  1.65519267e-02,\n",
       "       -5.23447176e-04, -2.67643575e-02, -2.80446447e-02, -1.69686824e-02,\n",
       "       -1.78932026e-02, -2.25125272e-02, -3.50899086e-03, -9.78177879e-03,\n",
       "        7.08928751e-03,  1.03520146e-02, -4.23031040e-02, -1.43829808e-02,\n",
       "       -8.08087960e-02,  2.69559585e-02,  2.88853701e-02, -3.09571866e-02,\n",
       "       -3.26542407e-02,  2.75605381e-03,  1.02422247e-02, -2.35671792e-02,\n",
       "       -2.33161505e-02,  5.81696406e-02,  2.19720807e-02, -1.17143495e-02,\n",
       "        1.01174209e-02, -2.66882428e-03, -6.03606440e-02, -2.37942976e-03,\n",
       "        1.21415406e-02, -4.11002859e-02, -4.51718643e-02, -1.60665307e-02,\n",
       "        2.70442460e-02,  3.09809074e-02, -3.05432342e-02,  3.15632485e-03,\n",
       "       -6.68478161e-02, -1.29532870e-02, -4.91549075e-03, -3.23550291e-02,\n",
       "       -3.58997770e-02, -4.72941734e-02, -1.03332186e-02,  4.78735194e-04,\n",
       "       -7.98223615e-02,  4.37410586e-02,  2.22239434e-03,  4.04256135e-02,\n",
       "        6.82347082e-03, -1.96469575e-02,  3.53155695e-02,  2.63391398e-02,\n",
       "        3.70172337e-02, -7.35288300e-03,  1.09556749e-01,  5.15761897e-02,\n",
       "       -5.43018468e-02, -4.83204499e-02, -5.97818531e-02, -5.58765978e-03,\n",
       "        6.76242262e-02, -1.12661384e-02,  4.74265181e-02, -6.16931589e-03,\n",
       "       -8.67325999e-03, -1.69603527e-02, -1.58700030e-02, -6.41136691e-02,\n",
       "       -4.75617908e-02,  1.82289444e-02,  6.15704432e-02, -1.81176811e-02,\n",
       "       -4.63466346e-02, -1.21487258e-02,  3.53394002e-02,  1.43153798e-02,\n",
       "       -1.51322945e-03, -4.43292446e-02,  8.88635230e-04, -3.44144069e-02,\n",
       "       -1.51215037e-02,  7.00254133e-03, -2.48295208e-03, -6.46390244e-02,\n",
       "       -5.57476804e-02,  7.15366192e-03,  7.20427837e-03, -8.45418051e-02,\n",
       "       -1.68811772e-02,  6.69234549e-04,  4.24135625e-02,  9.27098654e-03,\n",
       "        5.17644221e-03,  1.82485823e-02,  1.43265156e-02, -1.22386374e-01,\n",
       "       -1.10516297e-02,  2.88278852e-02,  6.96114302e-02, -4.52436022e-02,\n",
       "       -3.79772182e-03,  1.01903854e-02, -6.72446284e-03, -2.29705274e-02,\n",
       "        6.06019422e-02,  7.16592092e-03, -2.32902430e-02,  4.78281453e-02,\n",
       "        4.67650816e-02,  3.98564376e-02, -5.32545783e-02, -1.64170079e-02,\n",
       "        8.58898915e-04,  5.65902935e-03,  3.52774933e-02, -2.19531581e-02,\n",
       "        1.59286782e-02,  1.41221192e-02, -2.00276058e-02, -9.17218905e-03,\n",
       "       -7.49452505e-03,  2.47107036e-02, -2.48925504e-03, -9.48616304e-03,\n",
       "        4.92534135e-03, -2.19699480e-02, -6.24688230e-02,  2.18430795e-02,\n",
       "        2.42757834e-02,  5.86127490e-02, -5.02397008e-02,  3.77099402e-03,\n",
       "       -1.80244390e-02, -2.82009076e-02, -2.79995631e-02, -1.48973903e-02,\n",
       "       -3.07533308e-03, -2.61211246e-02,  2.79021319e-02, -2.04358483e-03,\n",
       "       -2.52628494e-02, -3.60019244e-02, -3.50774080e-02, -4.76166494e-02,\n",
       "        1.70159491e-03,  5.43572940e-02,  5.06276004e-02,  4.04567569e-02,\n",
       "        1.00103272e-02, -2.36803927e-02,  1.92706790e-02,  5.35508106e-03,\n",
       "       -1.91528480e-02,  2.40947725e-03, -2.84320526e-02,  2.67460588e-02,\n",
       "       -2.68261265e-02,  1.71120428e-02, -1.85380150e-02,  5.52691147e-03,\n",
       "       -3.09624802e-02, -3.61106507e-02, -4.97021191e-02,  6.04570918e-02,\n",
       "        1.61241926e-02, -4.77947854e-02, -3.17425579e-02, -7.21791480e-03,\n",
       "        2.30332687e-02, -1.35370474e-02,  6.81777075e-02, -2.20571272e-02,\n",
       "        1.39310835e-02, -7.71866413e-03,  1.99044030e-02,  5.11281611e-03,\n",
       "       -7.50671551e-02,  3.81815545e-02, -8.49779770e-02,  1.39642861e-02,\n",
       "        1.07679078e-02, -6.52924925e-02, -1.85855962e-02,  2.07402129e-02,\n",
       "        4.84880898e-03,  3.73609737e-02, -3.75557207e-02,  1.61507502e-02,\n",
       "       -4.15056869e-02, -3.32206041e-02,  5.85096329e-02,  4.62735929e-02,\n",
       "        5.73100746e-02,  2.75436435e-02,  4.90288064e-02,  5.45493104e-02,\n",
       "       -6.84350030e-03, -4.08428535e-03, -3.86421382e-02,  2.65611894e-02,\n",
       "       -2.24672351e-02,  1.40256025e-02,  3.93021060e-03, -4.97294685e-33,\n",
       "       -3.59437205e-02, -5.74628450e-02, -1.62865128e-02,  9.00254678e-03,\n",
       "        1.60071328e-02, -1.28748419e-03, -1.16924280e-02,  5.32272682e-02,\n",
       "        1.11453226e-02, -1.07059609e-02, -2.21116026e-03,  1.72310579e-03,\n",
       "        2.94323582e-02, -2.58768201e-02,  4.24533933e-02,  1.73520483e-02,\n",
       "       -1.25585869e-02,  9.55471490e-03,  1.68609172e-02, -8.00950527e-02,\n",
       "        1.16722668e-02,  2.90282853e-02,  7.11322501e-02, -4.01582569e-02,\n",
       "        1.51641283e-03,  1.55115621e-02, -3.80987749e-02, -1.73638910e-02,\n",
       "        7.47093046e-03,  2.56600287e-02, -1.90735031e-02,  9.33269411e-02,\n",
       "        8.26741336e-04,  3.29675004e-02, -1.44453831e-02,  2.23382246e-02,\n",
       "       -9.78794787e-03, -5.95429912e-02,  2.51312107e-02,  7.84727465e-03,\n",
       "       -3.35632339e-02,  2.06995346e-02,  2.14453470e-02, -3.03651635e-02,\n",
       "       -3.38789374e-02,  3.42657045e-02,  8.02799389e-02, -2.27102488e-02,\n",
       "       -1.45141222e-02,  1.35908630e-02, -3.66875641e-02,  5.68338810e-03,\n",
       "        1.32126007e-02, -3.84590738e-02,  2.29348205e-02, -3.02357171e-02,\n",
       "        5.74902864e-03, -6.41770810e-02, -3.17523740e-02,  4.55485955e-02,\n",
       "       -1.53187504e-02,  3.03061772e-02,  6.04593642e-02,  1.04117710e-02,\n",
       "        2.54260544e-02,  5.37067838e-02,  8.20452720e-02, -4.19960497e-03,\n",
       "        6.58357795e-03,  6.83037518e-03,  3.24908309e-02,  1.74423326e-02,\n",
       "        6.07533753e-02,  8.56453925e-02,  4.81057130e-02, -9.37992185e-02,\n",
       "        4.93356995e-02,  2.36274602e-04,  1.70228854e-02,  3.32530811e-02,\n",
       "        3.10481396e-02,  2.79911049e-02, -2.56981142e-02, -3.17143947e-02,\n",
       "        1.03422655e-02, -1.64685361e-02, -4.46704216e-02, -4.09626588e-02,\n",
       "        1.29443863e-02,  2.18537985e-04,  1.89387016e-02,  2.98334397e-02,\n",
       "       -3.28490287e-02, -4.15589474e-02,  2.17367448e-02, -2.91569363e-02,\n",
       "       -3.56767070e-03, -4.43830248e-03, -1.84462015e-02,  5.87964244e-03,\n",
       "        1.65849365e-02, -3.31492499e-02, -2.50569899e-02,  2.18881853e-02,\n",
       "        3.69399004e-02,  4.92964080e-03, -5.22426777e-02,  1.24047222e-02,\n",
       "       -5.31707965e-02, -2.49955952e-02,  7.60653813e-04, -4.11682203e-02,\n",
       "        4.60143015e-02, -2.12279912e-02,  2.31682360e-02, -4.59492579e-02,\n",
       "       -1.29493950e-02,  1.74553448e-03, -4.47838940e-03,  3.03012095e-02,\n",
       "       -3.20938304e-02, -5.02362698e-02, -1.07084615e-02,  6.86716195e-03,\n",
       "       -5.05223386e-02,  6.73037115e-03, -1.80894174e-02,  3.50810401e-02,\n",
       "        5.49725480e-02, -1.24864271e-02, -7.71613931e-03,  1.99220050e-02,\n",
       "        2.18156728e-07,  4.85302508e-02,  5.60069419e-02, -2.00262796e-02,\n",
       "        5.47962673e-02, -2.99007986e-02, -1.53709818e-02,  3.61846900e-03,\n",
       "        7.77632650e-03, -9.64170974e-03, -1.96539946e-02,  5.73286321e-03,\n",
       "        7.15564843e-03,  8.37491546e-03,  2.20809933e-02, -4.68556099e-02,\n",
       "       -2.16812119e-02,  2.27468014e-02,  3.64910699e-02, -7.91412592e-03,\n",
       "        2.46194284e-02,  9.09671858e-02,  9.61854309e-02,  5.92731088e-02,\n",
       "       -1.38591742e-02, -6.40719943e-03, -5.35838231e-02, -1.11716166e-02,\n",
       "       -1.05274292e-02,  3.35113294e-02, -2.62926258e-02,  3.67871076e-02,\n",
       "       -7.31061120e-03,  2.26462316e-02,  8.51114765e-02, -2.73483116e-02,\n",
       "       -4.49619032e-02,  8.26552585e-02,  2.72787884e-02, -2.61984356e-02,\n",
       "        6.16995990e-02, -9.19354148e-04, -2.93136034e-02,  1.78981456e-03,\n",
       "       -1.66344941e-02,  1.99334715e-02,  2.98925079e-02,  3.50755975e-02,\n",
       "       -1.89114921e-02, -3.96314077e-02, -5.20372938e-04,  5.12182713e-02,\n",
       "       -1.46389212e-02,  8.99124425e-03, -3.68658081e-03,  7.81543646e-03,\n",
       "        2.73975078e-02, -2.42043957e-02, -3.49381031e-03, -6.62653008e-03,\n",
       "        3.11317313e-02, -3.71047445e-02,  3.21330990e-05,  1.61609314e-02,\n",
       "       -5.71104772e-02,  1.29826050e-02, -6.20196834e-02, -2.54406221e-02,\n",
       "        1.13652132e-34,  4.66185063e-03,  2.03389302e-02, -2.14782474e-03,\n",
       "       -1.75573342e-02,  2.12787408e-02,  1.70300659e-02,  7.33859986e-02,\n",
       "        1.07082278e-02,  3.89475189e-02, -2.34981850e-02, -5.90743609e-02],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "context_emb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "e2c25576-12fe-4d65-97c4-66b51947d80a",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0.026,  0.028, -0.029,  0.027, -0.02 , -0.025, -0.003,  0.013,\n",
       "        0.032, -0.013,  0.038, -0.043,  0.014,  0.084,  0.031,  0.012,\n",
       "        0.069,  0.013, -0.011, -0.002,  0.005, -0.001, -0.006,  0.047,\n",
       "       -0.03 , -0.036, -0.025, -0.022, -0.021, -0.023, -0.028,  0.015,\n",
       "       -0.017,  0.046,  0.   , -0.056, -0.039, -0.017,  0.011, -0.029,\n",
       "        0.086,  0.032,  0.004, -0.003, -0.064,  0.04 ,  0.047, -0.045,\n",
       "       -0.001,  0.037, -0.008, -0.045,  0.029,  0.021,  0.057, -0.001,\n",
       "        0.006,  0.022,  0.034, -0.059, -0.031,  0.01 , -0.022,  0.027,\n",
       "        0.051,  0.022,  0.026, -0.062, -0.005, -0.022,  0.002, -0.035,\n",
       "       -0.012,  0.029, -0.04 , -0.026,  0.011, -0.012, -0.019, -0.023,\n",
       "        0.062,  0.011, -0.011,  0.004, -0.004,  0.084, -0.002, -0.034,\n",
       "        0.012,  0.023,  0.011, -0.016,  0.056,  0.001, -0.009,  0.026,\n",
       "       -0.027, -0.023,  0.009, -0.077,  0.036,  0.012,  0.059,  0.036,\n",
       "       -0.017, -0.006,  0.015,  0.02 , -0.073,  0.027, -0.081, -0.021,\n",
       "       -0.033,  0.073,  0.039, -0.05 , -0.066, -0.01 ,  0.001, -0.019,\n",
       "       -0.014,  0.017, -0.01 ,  0.023, -0.041, -0.038, -0.008,  0.028,\n",
       "       -0.03 , -0.029,  0.021,  0.023, -0.009, -0.015, -0.009,  0.067,\n",
       "       -0.013, -0.058,  0.002, -0.08 , -0.028, -0.028, -0.033, -0.066,\n",
       "       -0.025,  0.02 ,  0.01 , -0.055,  0.009,  0.041, -0.044,  0.103,\n",
       "       -0.076,  0.013,  0.001, -0.011,  0.015,  0.026, -0.005,  0.045,\n",
       "        0.046,  0.006,  0.006,  0.007, -0.059, -0.101, -0.061, -0.011,\n",
       "       -0.058,  0.044, -0.039,  0.041,  0.006,  0.054,  0.024,  0.105,\n",
       "        0.061,  0.028,  0.006, -0.031, -0.016, -0.015, -0.014, -0.003,\n",
       "       -0.088,  0.002, -0.029, -0.024,  0.016, -0.054, -0.03 , -0.008,\n",
       "        0.019, -0.031,  0.046, -0.006, -0.019, -0.049,  0.009, -0.048,\n",
       "        0.022, -0.049,  0.076,  0.13 ,  0.033, -0.053,  0.083,  0.008,\n",
       "       -0.023,  0.046,  0.026,  0.014, -0.046, -0.021, -0.018,  0.028,\n",
       "        0.013,  0.022, -0.004,  0.032,  0.018, -0.037,  0.021,  0.021,\n",
       "       -0.003, -0.054, -0.047,  0.014,  0.014,  0.009, -0.004,  0.057,\n",
       "        0.026, -0.002, -0.051, -0.02 , -0.047,  0.034,  0.012, -0.058,\n",
       "        0.028,  0.033,  0.068,  0.012,  0.057, -0.067,  0.057, -0.071,\n",
       "        0.038, -0.014,  0.045,  0.007,  0.029,  0.016,  0.037, -0.026,\n",
       "       -0.025, -0.021,  0.004, -0.047,  0.008, -0.014,  0.034,  0.003,\n",
       "        0.005, -0.025,  0.001, -0.035, -0.043,  0.032, -0.005,  0.017,\n",
       "       -0.023, -0.004,  0.011,  0.023,  0.038, -0.014,  0.017,  0.034,\n",
       "       -0.034, -0.024, -0.044, -0.031, -0.05 ,  0.059,  0.033, -0.032,\n",
       "       -0.007,  0.032, -0.024,  0.027,  0.028, -0.043,  0.025, -0.029,\n",
       "       -0.005, -0.003,  0.019,  0.042,  0.045, -0.004,  0.009, -0.033,\n",
       "        0.002,  0.024,  0.052, -0.04 , -0.046,  0.034,  0.09 ,  0.021,\n",
       "       -0.032,  0.05 , -0.059, -0.034, -0.072, -0.013, -0.045,  0.06 ,\n",
       "       -0.004, -0.051,  0.09 , -0.001,  0.02 ,  0.003, -0.03 ,  0.021,\n",
       "        0.077, -0.024,  0.015,  0.   , -0.036, -0.02 , -0.028,  0.034,\n",
       "       -0.065,  0.003, -0.026,  0.022,  0.02 ,  0.029, -0.014, -0.023,\n",
       "       -0.078,  0.002,  0.081,  0.033, -0.023,  0.045,  0.013,  0.017,\n",
       "       -0.001, -0.027, -0.028, -0.017, -0.018, -0.023, -0.004, -0.01 ,\n",
       "        0.007,  0.01 , -0.042, -0.014, -0.081,  0.027,  0.029, -0.031,\n",
       "       -0.033,  0.003,  0.01 , -0.024, -0.023,  0.058,  0.022, -0.012,\n",
       "        0.01 , -0.003, -0.06 , -0.002,  0.012, -0.041, -0.045, -0.016,\n",
       "        0.027,  0.031, -0.031,  0.003, -0.067, -0.013, -0.005, -0.032,\n",
       "       -0.036, -0.047, -0.01 ,  0.   , -0.08 ,  0.044,  0.002,  0.04 ,\n",
       "        0.007, -0.02 ,  0.035,  0.026,  0.037, -0.007,  0.11 ,  0.052,\n",
       "       -0.054, -0.048, -0.06 , -0.006,  0.068, -0.011,  0.047, -0.006,\n",
       "       -0.009, -0.017, -0.016, -0.064, -0.048,  0.018,  0.062, -0.018,\n",
       "       -0.046, -0.012,  0.035,  0.014, -0.002, -0.044,  0.001, -0.034,\n",
       "       -0.015,  0.007, -0.002, -0.065, -0.056,  0.007,  0.007, -0.085,\n",
       "       -0.017,  0.001,  0.042,  0.009,  0.005,  0.018,  0.014, -0.122,\n",
       "       -0.011,  0.029,  0.07 , -0.045, -0.004,  0.01 , -0.007, -0.023,\n",
       "        0.061,  0.007, -0.023,  0.048,  0.047,  0.04 , -0.053, -0.016,\n",
       "        0.001,  0.006,  0.035, -0.022,  0.016,  0.014, -0.02 , -0.009,\n",
       "       -0.007,  0.025, -0.002, -0.009,  0.005, -0.022, -0.062,  0.022,\n",
       "        0.024,  0.059, -0.05 ,  0.004, -0.018, -0.028, -0.028, -0.015,\n",
       "       -0.003, -0.026,  0.028, -0.002, -0.025, -0.036, -0.035, -0.048,\n",
       "        0.002,  0.054,  0.051,  0.04 ,  0.01 , -0.024,  0.019,  0.005,\n",
       "       -0.019,  0.002, -0.028,  0.027, -0.027,  0.017, -0.019,  0.006,\n",
       "       -0.031, -0.036, -0.05 ,  0.06 ,  0.016, -0.048, -0.032, -0.007,\n",
       "        0.023, -0.014,  0.068, -0.022,  0.014, -0.008,  0.02 ,  0.005,\n",
       "       -0.075,  0.038, -0.085,  0.014,  0.011, -0.065, -0.019,  0.021,\n",
       "        0.005,  0.037, -0.038,  0.016, -0.042, -0.033,  0.059,  0.046,\n",
       "        0.057,  0.028,  0.049,  0.055, -0.007, -0.004, -0.039,  0.027,\n",
       "       -0.022,  0.014,  0.004, -0.   , -0.036, -0.057, -0.016,  0.009,\n",
       "        0.016, -0.001, -0.012,  0.053,  0.011, -0.011, -0.002,  0.002,\n",
       "        0.029, -0.026,  0.042,  0.017, -0.013,  0.01 ,  0.017, -0.08 ,\n",
       "        0.012,  0.029,  0.071, -0.04 ,  0.002,  0.016, -0.038, -0.017,\n",
       "        0.007,  0.026, -0.019,  0.093,  0.001,  0.033, -0.014,  0.022,\n",
       "       -0.01 , -0.06 ,  0.025,  0.008, -0.034,  0.021,  0.021, -0.03 ,\n",
       "       -0.034,  0.034,  0.08 , -0.023, -0.015,  0.014, -0.037,  0.006,\n",
       "        0.013, -0.038,  0.023, -0.03 ,  0.006, -0.064, -0.032,  0.046,\n",
       "       -0.015,  0.03 ,  0.06 ,  0.01 ,  0.025,  0.054,  0.082, -0.004,\n",
       "        0.007,  0.007,  0.032,  0.017,  0.061,  0.086,  0.048, -0.094,\n",
       "        0.049,  0.   ,  0.017,  0.033,  0.031,  0.028, -0.026, -0.032,\n",
       "        0.01 , -0.016, -0.045, -0.041,  0.013,  0.   ,  0.019,  0.03 ,\n",
       "       -0.033, -0.042,  0.022, -0.029, -0.004, -0.004, -0.018,  0.006,\n",
       "        0.017, -0.033, -0.025,  0.022,  0.037,  0.005, -0.052,  0.012,\n",
       "       -0.053, -0.025,  0.001, -0.041,  0.046, -0.021,  0.023, -0.046,\n",
       "       -0.013,  0.002, -0.004,  0.03 , -0.032, -0.05 , -0.011,  0.007,\n",
       "       -0.051,  0.007, -0.018,  0.035,  0.055, -0.012, -0.008,  0.02 ,\n",
       "        0.   ,  0.049,  0.056, -0.02 ,  0.055, -0.03 , -0.015,  0.004,\n",
       "        0.008, -0.01 , -0.02 ,  0.006,  0.007,  0.008,  0.022, -0.047,\n",
       "       -0.022,  0.023,  0.036, -0.008,  0.025,  0.091,  0.096,  0.059,\n",
       "       -0.014, -0.006, -0.054, -0.011, -0.011,  0.034, -0.026,  0.037,\n",
       "       -0.007,  0.023,  0.085, -0.027, -0.045,  0.083,  0.027, -0.026,\n",
       "        0.062, -0.001, -0.029,  0.002, -0.017,  0.02 ,  0.03 ,  0.035,\n",
       "       -0.019, -0.04 , -0.001,  0.051, -0.015,  0.009, -0.004,  0.008,\n",
       "        0.027, -0.024, -0.003, -0.007,  0.031, -0.037,  0.   ,  0.016,\n",
       "       -0.057,  0.013, -0.062, -0.025,  0.   ,  0.005,  0.02 , -0.002,\n",
       "       -0.018,  0.021,  0.017,  0.073,  0.011,  0.039, -0.023, -0.059],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# для наглядности\n",
    "context_emb.round(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "98ca36d5-7b87-42c9-b362-24eba2ea51f9",
   "metadata": {},
   "source": [
    "Рассчитаем косинусную близость с помощью функции из sklearn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "eca47d2b-ff8e-4ce8-9b7d-fd69c1f186c3",
   "metadata": {},
   "outputs": [
    {
     "ename": "ValueError",
     "evalue": "Expected 2D array, got 1D array instead:\narray=[ 2.60696858e-02  2.81297602e-02 -2.86605358e-02  2.73304842e-02\n -2.03045104e-02 -2.45557521e-02 -3.44669889e-03  1.29649406e-02\n  3.22340988e-02 -1.32162180e-02  3.78281027e-02 -4.27398421e-02\n  1.42525053e-02  8.37171823e-02  3.09127476e-02  1.17423786e-02\n  6.91081583e-02  1.30820498e-02 -1.07209831e-02 -2.10680906e-03\n  4.95719211e-03 -1.02009624e-03 -5.77510986e-03  4.73414510e-02\n -3.00503504e-02 -3.59182991e-02 -2.45586336e-02 -2.17193905e-02\n -2.08961219e-02 -2.29210742e-02 -2.80032977e-02  1.52165564e-02\n -1.65842436e-02  4.57603335e-02  1.44828766e-06 -5.61360009e-02\n -3.94317359e-02 -1.73282661e-02  1.10615166e-02 -2.93712430e-02\n  8.55174810e-02  3.19165811e-02  4.22134530e-03 -2.52286508e-03\n -6.43932000e-02  4.01510037e-02  4.70006950e-02 -4.48253714e-02\n -1.41513813e-03  3.66889797e-02 -8.11715703e-03 -4.46259156e-02\n  2.88627185e-02  2.14875434e-02  5.65632023e-02 -7.84562319e-04\n  5.81800705e-03  2.19479781e-02  3.44444998e-02 -5.85746132e-02\n -3.14422622e-02  9.82944295e-03 -2.18724068e-02  2.72080489e-02\n  5.10382988e-02  2.24636458e-02  2.58076116e-02 -6.15553483e-02\n -5.12524135e-03 -2.24703737e-02  2.49784882e-03 -3.52916196e-02\n -1.16497632e-02  2.93239839e-02 -3.97467874e-02 -2.59171855e-02\n  1.10591054e-02 -1.20284809e-02 -1.90009009e-02 -2.27706600e-02\n  6.18728735e-02  1.14581380e-02 -1.07487626e-02  4.07132506e-03\n -3.67226219e-03  8.43563527e-02 -2.39394954e-03 -3.37652229e-02\n  1.20446840e-02  2.25450788e-02  1.10169696e-02 -1.64049696e-02\n  5.59397414e-02  1.17083488e-03 -9.31880530e-03  2.62642857e-02\n -2.74639539e-02 -2.27501988e-02  8.93268362e-03 -7.66405538e-02\n  3.55793536e-02  1.20696472e-02  5.86612076e-02  3.63901295e-02\n -1.65351499e-02 -5.73400129e-03  1.45409070e-02  1.96258314e-02\n -7.27106109e-02  2.72692945e-02 -8.13957080e-02 -2.11496335e-02\n -3.26966271e-02  7.29898438e-02  3.90958190e-02 -4.96897846e-02\n -6.63388595e-02 -1.01196701e-02  6.42161875e-04 -1.88012272e-02\n -1.35782529e-02  1.69789605e-02 -9.91971418e-03  2.28155497e-02\n -4.12469879e-02 -3.77093293e-02 -7.57177593e-03  2.83327959e-02\n -3.02903876e-02 -2.85167098e-02  2.07434315e-02  2.25604214e-02\n -9.28365719e-03 -1.51147833e-02 -8.98459088e-03  6.70140386e-02\n -1.28544942e-02 -5.78452088e-02  2.47775298e-03 -7.97296911e-02\n -2.81679835e-02 -2.84210779e-02 -3.25452425e-02 -6.56164214e-02\n -2.47204620e-02  2.01430097e-02  1.01629477e-02 -5.48637062e-02\n  8.94648675e-03  4.05956618e-02 -4.44052890e-02  1.02671042e-01\n -7.56809115e-02  1.26552181e-02  1.37866579e-03 -1.13331880e-02\n  1.51405660e-02  2.56871488e-02 -5.24983183e-03  4.51805294e-02\n  4.59146686e-02  5.85043104e-03  6.26491150e-03  7.20547792e-03\n -5.88717051e-02 -1.00635946e-01 -6.11553602e-02 -1.08829951e-02\n -5.83557002e-02  4.36093472e-02 -3.92436385e-02  4.10913564e-02\n  5.64262178e-03  5.37047759e-02  2.42957957e-02  1.05013326e-01\n  6.05127029e-02  2.84899641e-02  5.55243762e-03 -3.09707951e-02\n -1.55448066e-02 -1.46294916e-02 -1.43787283e-02 -3.49355722e-03\n -8.83707330e-02  2.19522649e-03 -2.93946192e-02 -2.35651769e-02\n  1.60144772e-02 -5.43737002e-02 -3.02929245e-02 -7.79132126e-03\n  1.89319085e-02 -3.13578583e-02  4.60174680e-02 -5.50660631e-03\n -1.92169063e-02 -4.88135181e-02  8.63954145e-03 -4.77152430e-02\n  2.17958838e-02 -4.88891192e-02  7.63864294e-02  1.30082771e-01\n  3.28815654e-02 -5.33933975e-02  8.30223709e-02  7.57148629e-03\n -2.32078675e-02  4.61315811e-02  2.62386221e-02  1.36163775e-02\n -4.62059826e-02 -2.14045327e-02 -1.82300620e-02  2.83758324e-02\n  1.28799668e-02  2.15108078e-02 -4.15302766e-03  3.15550305e-02\n  1.82425436e-02 -3.74762639e-02  2.08042860e-02  2.08608937e-02\n -2.82959430e-03 -5.35163693e-02 -4.71888147e-02  1.35965738e-02\n  1.38235912e-02  9.01751686e-03 -3.65398708e-03  5.73994629e-02\n  2.57573053e-02 -2.01233313e-03 -5.14804944e-02 -2.00652517e-02\n -4.66603264e-02  3.39639969e-02  1.17061762e-02 -5.82824051e-02\n  2.77182721e-02  3.31477113e-02  6.75387383e-02  1.22633018e-02\n  5.67695983e-02 -6.65890053e-02  5.69646060e-02 -7.09204301e-02\n  3.76802348e-02 -1.43325031e-02  4.47984375e-02  7.28084426e-03\n  2.85710488e-02  1.60099398e-02  3.72759327e-02 -2.62777582e-02\n -2.46258285e-02 -2.06304975e-02  3.72276478e-03 -4.67101634e-02\n  8.38323124e-03 -1.35120293e-02  3.39612849e-02  3.49291484e-03\n  4.99470299e-03 -2.49034129e-02  7.93784333e-04 -3.46210860e-02\n -4.31761034e-02  3.18461545e-02 -4.72946931e-03  1.70705188e-02\n -2.31329408e-02 -4.39915759e-03  1.13313077e-02  2.25792192e-02\n  3.83138545e-02 -1.36187421e-02  1.68776456e-02  3.40576060e-02\n -3.42961587e-02 -2.40255930e-02 -4.41743918e-02 -3.07683609e-02\n -4.96942215e-02  5.87186776e-02  3.34627815e-02 -3.17160413e-02\n -7.49719329e-03  3.18871289e-02 -2.39836518e-02  2.70108543e-02\n  2.77622230e-02 -4.32922244e-02  2.50458159e-02 -2.90878844e-02\n -5.16465493e-03 -3.24391341e-03  1.94145795e-02  4.15609553e-02\n  4.53335270e-02 -4.18853667e-03  8.84133205e-03 -3.26370895e-02\n  2.16132705e-03  2.40324698e-02  5.22514284e-02 -3.98226120e-02\n -4.63361889e-02  3.35683487e-02  8.99985433e-02  2.09509823e-02\n -3.21147405e-02  4.97947820e-02 -5.90330772e-02 -3.41108479e-02\n -7.22306669e-02 -1.34600475e-02 -4.47860770e-02  6.01749718e-02\n -3.98624782e-03 -5.05394153e-02  8.97321180e-02 -7.15939677e-04\n  1.99675821e-02  3.08200088e-03 -2.99170837e-02  2.10373215e-02\n  7.66325668e-02 -2.41141766e-02  1.48944026e-02  2.33445171e-04\n -3.57108861e-02 -2.00601127e-02 -2.78608240e-02  3.36323306e-02\n -6.52984530e-02  2.57370248e-03 -2.62748804e-02  2.23488789e-02\n  1.96563937e-02  2.94369552e-02 -1.44276265e-02 -2.28903387e-02\n -7.76160210e-02  1.66622142e-03  8.07723925e-02  3.26537266e-02\n -2.33838372e-02  4.45028879e-02  1.28065404e-02  1.65519267e-02\n -5.23447176e-04 -2.67643575e-02 -2.80446447e-02 -1.69686824e-02\n -1.78932026e-02 -2.25125272e-02 -3.50899086e-03 -9.78177879e-03\n  7.08928751e-03  1.03520146e-02 -4.23031040e-02 -1.43829808e-02\n -8.08087960e-02  2.69559585e-02  2.88853701e-02 -3.09571866e-02\n -3.26542407e-02  2.75605381e-03  1.02422247e-02 -2.35671792e-02\n -2.33161505e-02  5.81696406e-02  2.19720807e-02 -1.17143495e-02\n  1.01174209e-02 -2.66882428e-03 -6.03606440e-02 -2.37942976e-03\n  1.21415406e-02 -4.11002859e-02 -4.51718643e-02 -1.60665307e-02\n  2.70442460e-02  3.09809074e-02 -3.05432342e-02  3.15632485e-03\n -6.68478161e-02 -1.29532870e-02 -4.91549075e-03 -3.23550291e-02\n -3.58997770e-02 -4.72941734e-02 -1.03332186e-02  4.78735194e-04\n -7.98223615e-02  4.37410586e-02  2.22239434e-03  4.04256135e-02\n  6.82347082e-03 -1.96469575e-02  3.53155695e-02  2.63391398e-02\n  3.70172337e-02 -7.35288300e-03  1.09556749e-01  5.15761897e-02\n -5.43018468e-02 -4.83204499e-02 -5.97818531e-02 -5.58765978e-03\n  6.76242262e-02 -1.12661384e-02  4.74265181e-02 -6.16931589e-03\n -8.67325999e-03 -1.69603527e-02 -1.58700030e-02 -6.41136691e-02\n -4.75617908e-02  1.82289444e-02  6.15704432e-02 -1.81176811e-02\n -4.63466346e-02 -1.21487258e-02  3.53394002e-02  1.43153798e-02\n -1.51322945e-03 -4.43292446e-02  8.88635230e-04 -3.44144069e-02\n -1.51215037e-02  7.00254133e-03 -2.48295208e-03 -6.46390244e-02\n -5.57476804e-02  7.15366192e-03  7.20427837e-03 -8.45418051e-02\n -1.68811772e-02  6.69234549e-04  4.24135625e-02  9.27098654e-03\n  5.17644221e-03  1.82485823e-02  1.43265156e-02 -1.22386374e-01\n -1.10516297e-02  2.88278852e-02  6.96114302e-02 -4.52436022e-02\n -3.79772182e-03  1.01903854e-02 -6.72446284e-03 -2.29705274e-02\n  6.06019422e-02  7.16592092e-03 -2.32902430e-02  4.78281453e-02\n  4.67650816e-02  3.98564376e-02 -5.32545783e-02 -1.64170079e-02\n  8.58898915e-04  5.65902935e-03  3.52774933e-02 -2.19531581e-02\n  1.59286782e-02  1.41221192e-02 -2.00276058e-02 -9.17218905e-03\n -7.49452505e-03  2.47107036e-02 -2.48925504e-03 -9.48616304e-03\n  4.92534135e-03 -2.19699480e-02 -6.24688230e-02  2.18430795e-02\n  2.42757834e-02  5.86127490e-02 -5.02397008e-02  3.77099402e-03\n -1.80244390e-02 -2.82009076e-02 -2.79995631e-02 -1.48973903e-02\n -3.07533308e-03 -2.61211246e-02  2.79021319e-02 -2.04358483e-03\n -2.52628494e-02 -3.60019244e-02 -3.50774080e-02 -4.76166494e-02\n  1.70159491e-03  5.43572940e-02  5.06276004e-02  4.04567569e-02\n  1.00103272e-02 -2.36803927e-02  1.92706790e-02  5.35508106e-03\n -1.91528480e-02  2.40947725e-03 -2.84320526e-02  2.67460588e-02\n -2.68261265e-02  1.71120428e-02 -1.85380150e-02  5.52691147e-03\n -3.09624802e-02 -3.61106507e-02 -4.97021191e-02  6.04570918e-02\n  1.61241926e-02 -4.77947854e-02 -3.17425579e-02 -7.21791480e-03\n  2.30332687e-02 -1.35370474e-02  6.81777075e-02 -2.20571272e-02\n  1.39310835e-02 -7.71866413e-03  1.99044030e-02  5.11281611e-03\n -7.50671551e-02  3.81815545e-02 -8.49779770e-02  1.39642861e-02\n  1.07679078e-02 -6.52924925e-02 -1.85855962e-02  2.07402129e-02\n  4.84880898e-03  3.73609737e-02 -3.75557207e-02  1.61507502e-02\n -4.15056869e-02 -3.32206041e-02  5.85096329e-02  4.62735929e-02\n  5.73100746e-02  2.75436435e-02  4.90288064e-02  5.45493104e-02\n -6.84350030e-03 -4.08428535e-03 -3.86421382e-02  2.65611894e-02\n -2.24672351e-02  1.40256025e-02  3.93021060e-03 -4.97294685e-33\n -3.59437205e-02 -5.74628450e-02 -1.62865128e-02  9.00254678e-03\n  1.60071328e-02 -1.28748419e-03 -1.16924280e-02  5.32272682e-02\n  1.11453226e-02 -1.07059609e-02 -2.21116026e-03  1.72310579e-03\n  2.94323582e-02 -2.58768201e-02  4.24533933e-02  1.73520483e-02\n -1.25585869e-02  9.55471490e-03  1.68609172e-02 -8.00950527e-02\n  1.16722668e-02  2.90282853e-02  7.11322501e-02 -4.01582569e-02\n  1.51641283e-03  1.55115621e-02 -3.80987749e-02 -1.73638910e-02\n  7.47093046e-03  2.56600287e-02 -1.90735031e-02  9.33269411e-02\n  8.26741336e-04  3.29675004e-02 -1.44453831e-02  2.23382246e-02\n -9.78794787e-03 -5.95429912e-02  2.51312107e-02  7.84727465e-03\n -3.35632339e-02  2.06995346e-02  2.14453470e-02 -3.03651635e-02\n -3.38789374e-02  3.42657045e-02  8.02799389e-02 -2.27102488e-02\n -1.45141222e-02  1.35908630e-02 -3.66875641e-02  5.68338810e-03\n  1.32126007e-02 -3.84590738e-02  2.29348205e-02 -3.02357171e-02\n  5.74902864e-03 -6.41770810e-02 -3.17523740e-02  4.55485955e-02\n -1.53187504e-02  3.03061772e-02  6.04593642e-02  1.04117710e-02\n  2.54260544e-02  5.37067838e-02  8.20452720e-02 -4.19960497e-03\n  6.58357795e-03  6.83037518e-03  3.24908309e-02  1.74423326e-02\n  6.07533753e-02  8.56453925e-02  4.81057130e-02 -9.37992185e-02\n  4.93356995e-02  2.36274602e-04  1.70228854e-02  3.32530811e-02\n  3.10481396e-02  2.79911049e-02 -2.56981142e-02 -3.17143947e-02\n  1.03422655e-02 -1.64685361e-02 -4.46704216e-02 -4.09626588e-02\n  1.29443863e-02  2.18537985e-04  1.89387016e-02  2.98334397e-02\n -3.28490287e-02 -4.15589474e-02  2.17367448e-02 -2.91569363e-02\n -3.56767070e-03 -4.43830248e-03 -1.84462015e-02  5.87964244e-03\n  1.65849365e-02 -3.31492499e-02 -2.50569899e-02  2.18881853e-02\n  3.69399004e-02  4.92964080e-03 -5.22426777e-02  1.24047222e-02\n -5.31707965e-02 -2.49955952e-02  7.60653813e-04 -4.11682203e-02\n  4.60143015e-02 -2.12279912e-02  2.31682360e-02 -4.59492579e-02\n -1.29493950e-02  1.74553448e-03 -4.47838940e-03  3.03012095e-02\n -3.20938304e-02 -5.02362698e-02 -1.07084615e-02  6.86716195e-03\n -5.05223386e-02  6.73037115e-03 -1.80894174e-02  3.50810401e-02\n  5.49725480e-02 -1.24864271e-02 -7.71613931e-03  1.99220050e-02\n  2.18156728e-07  4.85302508e-02  5.60069419e-02 -2.00262796e-02\n  5.47962673e-02 -2.99007986e-02 -1.53709818e-02  3.61846900e-03\n  7.77632650e-03 -9.64170974e-03 -1.96539946e-02  5.73286321e-03\n  7.15564843e-03  8.37491546e-03  2.20809933e-02 -4.68556099e-02\n -2.16812119e-02  2.27468014e-02  3.64910699e-02 -7.91412592e-03\n  2.46194284e-02  9.09671858e-02  9.61854309e-02  5.92731088e-02\n -1.38591742e-02 -6.40719943e-03 -5.35838231e-02 -1.11716166e-02\n -1.05274292e-02  3.35113294e-02 -2.62926258e-02  3.67871076e-02\n -7.31061120e-03  2.26462316e-02  8.51114765e-02 -2.73483116e-02\n -4.49619032e-02  8.26552585e-02  2.72787884e-02 -2.61984356e-02\n  6.16995990e-02 -9.19354148e-04 -2.93136034e-02  1.78981456e-03\n -1.66344941e-02  1.99334715e-02  2.98925079e-02  3.50755975e-02\n -1.89114921e-02 -3.96314077e-02 -5.20372938e-04  5.12182713e-02\n -1.46389212e-02  8.99124425e-03 -3.68658081e-03  7.81543646e-03\n  2.73975078e-02 -2.42043957e-02 -3.49381031e-03 -6.62653008e-03\n  3.11317313e-02 -3.71047445e-02  3.21330990e-05  1.61609314e-02\n -5.71104772e-02  1.29826050e-02 -6.20196834e-02 -2.54406221e-02\n  1.13652132e-34  4.66185063e-03  2.03389302e-02 -2.14782474e-03\n -1.75573342e-02  2.12787408e-02  1.70300659e-02  7.33859986e-02\n  1.07082278e-02  3.89475189e-02 -2.34981850e-02 -5.90743609e-02].\nReshape your data either using array.reshape(-1, 1) if your data has a single feature or array.reshape(1, -1) if it contains a single sample.",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[8], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m \u001b[43mcosine_distances\u001b[49m\u001b[43m(\u001b[49m\u001b[43mcontext_emb\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mdefinition1_emb\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/miniforge3/envs/myenv/lib/python3.11/site-packages/sklearn/utils/_param_validation.py:211\u001b[0m, in \u001b[0;36mvalidate_params.<locals>.decorator.<locals>.wrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m    205\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[1;32m    206\u001b[0m     \u001b[38;5;28;01mwith\u001b[39;00m config_context(\n\u001b[1;32m    207\u001b[0m         skip_parameter_validation\u001b[38;5;241m=\u001b[39m(\n\u001b[1;32m    208\u001b[0m             prefer_skip_nested_validation \u001b[38;5;129;01mor\u001b[39;00m global_skip_validation\n\u001b[1;32m    209\u001b[0m         )\n\u001b[1;32m    210\u001b[0m     ):\n\u001b[0;32m--> 211\u001b[0m         \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    212\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m InvalidParameterError \u001b[38;5;28;01mas\u001b[39;00m e:\n\u001b[1;32m    213\u001b[0m     \u001b[38;5;66;03m# When the function is just a wrapper around an estimator, we allow\u001b[39;00m\n\u001b[1;32m    214\u001b[0m     \u001b[38;5;66;03m# the function to delegate validation to the estimator, but we replace\u001b[39;00m\n\u001b[1;32m    215\u001b[0m     \u001b[38;5;66;03m# the name of the estimator by the name of the function in the error\u001b[39;00m\n\u001b[1;32m    216\u001b[0m     \u001b[38;5;66;03m# message to avoid confusion.\u001b[39;00m\n\u001b[1;32m    217\u001b[0m     msg \u001b[38;5;241m=\u001b[39m re\u001b[38;5;241m.\u001b[39msub(\n\u001b[1;32m    218\u001b[0m         \u001b[38;5;124mr\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mparameter of \u001b[39m\u001b[38;5;124m\\\u001b[39m\u001b[38;5;124mw+ must be\u001b[39m\u001b[38;5;124m\"\u001b[39m,\n\u001b[1;32m    219\u001b[0m         \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mparameter of \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mfunc\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__qualname__\u001b[39m\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m must be\u001b[39m\u001b[38;5;124m\"\u001b[39m,\n\u001b[1;32m    220\u001b[0m         \u001b[38;5;28mstr\u001b[39m(e),\n\u001b[1;32m    221\u001b[0m     )\n",
      "File \u001b[0;32m~/miniforge3/envs/myenv/lib/python3.11/site-packages/sklearn/metrics/pairwise.py:1105\u001b[0m, in \u001b[0;36mcosine_distances\u001b[0;34m(X, Y)\u001b[0m\n\u001b[1;32m   1079\u001b[0m \u001b[38;5;250m\u001b[39m\u001b[38;5;124;03m\"\"\"Compute cosine distance between samples in X and Y.\u001b[39;00m\n\u001b[1;32m   1080\u001b[0m \n\u001b[1;32m   1081\u001b[0m \u001b[38;5;124;03mCosine distance is defined as 1.0 minus the cosine similarity.\u001b[39;00m\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   1102\u001b[0m \u001b[38;5;124;03mscipy.spatial.distance.cosine : Dense matrices only.\u001b[39;00m\n\u001b[1;32m   1103\u001b[0m \u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m   1104\u001b[0m \u001b[38;5;66;03m# 1.0 - cosine_similarity(X, Y) without copy\u001b[39;00m\n\u001b[0;32m-> 1105\u001b[0m S \u001b[38;5;241m=\u001b[39m \u001b[43mcosine_similarity\u001b[49m\u001b[43m(\u001b[49m\u001b[43mX\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mY\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1106\u001b[0m S \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m=\u001b[39m \u001b[38;5;241m-\u001b[39m\u001b[38;5;241m1\u001b[39m\n\u001b[1;32m   1107\u001b[0m S \u001b[38;5;241m+\u001b[39m\u001b[38;5;241m=\u001b[39m \u001b[38;5;241m1\u001b[39m\n",
      "File \u001b[0;32m~/miniforge3/envs/myenv/lib/python3.11/site-packages/sklearn/utils/_param_validation.py:184\u001b[0m, in \u001b[0;36mvalidate_params.<locals>.decorator.<locals>.wrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m    182\u001b[0m global_skip_validation \u001b[38;5;241m=\u001b[39m get_config()[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mskip_parameter_validation\u001b[39m\u001b[38;5;124m\"\u001b[39m]\n\u001b[1;32m    183\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m global_skip_validation:\n\u001b[0;32m--> 184\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    186\u001b[0m func_sig \u001b[38;5;241m=\u001b[39m signature(func)\n\u001b[1;32m    188\u001b[0m \u001b[38;5;66;03m# Map *args/**kwargs to the function signature\u001b[39;00m\n",
      "File \u001b[0;32m~/miniforge3/envs/myenv/lib/python3.11/site-packages/sklearn/metrics/pairwise.py:1578\u001b[0m, in \u001b[0;36mcosine_similarity\u001b[0;34m(X, Y, dense_output)\u001b[0m\n\u001b[1;32m   1543\u001b[0m \u001b[38;5;250m\u001b[39m\u001b[38;5;124;03m\"\"\"Compute cosine similarity between samples in X and Y.\u001b[39;00m\n\u001b[1;32m   1544\u001b[0m \n\u001b[1;32m   1545\u001b[0m \u001b[38;5;124;03mCosine similarity, or the cosine kernel, computes similarity as the\u001b[39;00m\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   1574\u001b[0m \u001b[38;5;124;03m    Returns the cosine similarity between samples in X and Y.\u001b[39;00m\n\u001b[1;32m   1575\u001b[0m \u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m   1576\u001b[0m \u001b[38;5;66;03m# to avoid recursive import\u001b[39;00m\n\u001b[0;32m-> 1578\u001b[0m X, Y \u001b[38;5;241m=\u001b[39m \u001b[43mcheck_pairwise_arrays\u001b[49m\u001b[43m(\u001b[49m\u001b[43mX\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mY\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1580\u001b[0m X_normalized \u001b[38;5;241m=\u001b[39m normalize(X, copy\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mTrue\u001b[39;00m)\n\u001b[1;32m   1581\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m X \u001b[38;5;129;01mis\u001b[39;00m Y:\n",
      "File \u001b[0;32m~/miniforge3/envs/myenv/lib/python3.11/site-packages/sklearn/metrics/pairwise.py:165\u001b[0m, in \u001b[0;36mcheck_pairwise_arrays\u001b[0;34m(X, Y, precomputed, dtype, accept_sparse, force_all_finite, copy)\u001b[0m\n\u001b[1;32m    156\u001b[0m     X \u001b[38;5;241m=\u001b[39m Y \u001b[38;5;241m=\u001b[39m check_array(\n\u001b[1;32m    157\u001b[0m         X,\n\u001b[1;32m    158\u001b[0m         accept_sparse\u001b[38;5;241m=\u001b[39maccept_sparse,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    162\u001b[0m         estimator\u001b[38;5;241m=\u001b[39mestimator,\n\u001b[1;32m    163\u001b[0m     )\n\u001b[1;32m    164\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m--> 165\u001b[0m     X \u001b[38;5;241m=\u001b[39m \u001b[43mcheck_array\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    166\u001b[0m \u001b[43m        \u001b[49m\u001b[43mX\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    167\u001b[0m \u001b[43m        \u001b[49m\u001b[43maccept_sparse\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43maccept_sparse\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    168\u001b[0m \u001b[43m        \u001b[49m\u001b[43mdtype\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mdtype\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    169\u001b[0m \u001b[43m        \u001b[49m\u001b[43mcopy\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mcopy\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    170\u001b[0m \u001b[43m        \u001b[49m\u001b[43mforce_all_finite\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mforce_all_finite\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    171\u001b[0m \u001b[43m        \u001b[49m\u001b[43mestimator\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mestimator\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    172\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    173\u001b[0m     Y \u001b[38;5;241m=\u001b[39m check_array(\n\u001b[1;32m    174\u001b[0m         Y,\n\u001b[1;32m    175\u001b[0m         accept_sparse\u001b[38;5;241m=\u001b[39maccept_sparse,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    179\u001b[0m         estimator\u001b[38;5;241m=\u001b[39mestimator,\n\u001b[1;32m    180\u001b[0m     )\n\u001b[1;32m    182\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m precomputed:\n",
      "File \u001b[0;32m~/miniforge3/envs/myenv/lib/python3.11/site-packages/sklearn/utils/validation.py:938\u001b[0m, in \u001b[0;36mcheck_array\u001b[0;34m(array, accept_sparse, accept_large_sparse, dtype, order, copy, force_all_finite, ensure_2d, allow_nd, ensure_min_samples, ensure_min_features, estimator, input_name)\u001b[0m\n\u001b[1;32m    936\u001b[0m     \u001b[38;5;66;03m# If input is 1D raise error\u001b[39;00m\n\u001b[1;32m    937\u001b[0m     \u001b[38;5;28;01mif\u001b[39;00m array\u001b[38;5;241m.\u001b[39mndim \u001b[38;5;241m==\u001b[39m \u001b[38;5;241m1\u001b[39m:\n\u001b[0;32m--> 938\u001b[0m         \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mValueError\u001b[39;00m(\n\u001b[1;32m    939\u001b[0m             \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mExpected 2D array, got 1D array instead:\u001b[39m\u001b[38;5;130;01m\\n\u001b[39;00m\u001b[38;5;124marray=\u001b[39m\u001b[38;5;132;01m{}\u001b[39;00m\u001b[38;5;124m.\u001b[39m\u001b[38;5;130;01m\\n\u001b[39;00m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    940\u001b[0m             \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mReshape your data either using array.reshape(-1, 1) if \u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    941\u001b[0m             \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124myour data has a single feature or array.reshape(1, -1) \u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    942\u001b[0m             \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mif it contains a single sample.\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;241m.\u001b[39mformat(array)\n\u001b[1;32m    943\u001b[0m         )\n\u001b[1;32m    945\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m dtype_numeric \u001b[38;5;129;01mand\u001b[39;00m \u001b[38;5;28mhasattr\u001b[39m(array\u001b[38;5;241m.\u001b[39mdtype, \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mkind\u001b[39m\u001b[38;5;124m\"\u001b[39m) \u001b[38;5;129;01mand\u001b[39;00m array\u001b[38;5;241m.\u001b[39mdtype\u001b[38;5;241m.\u001b[39mkind \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mUSV\u001b[39m\u001b[38;5;124m\"\u001b[39m:\n\u001b[1;32m    946\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mValueError\u001b[39;00m(\n\u001b[1;32m    947\u001b[0m         \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mdtype=\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mnumeric\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m is not compatible with arrays of bytes/strings.\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    948\u001b[0m         \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mConvert your data to numeric values explicitly instead.\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    949\u001b[0m     )\n",
      "\u001b[0;31mValueError\u001b[0m: Expected 2D array, got 1D array instead:\narray=[ 2.60696858e-02  2.81297602e-02 -2.86605358e-02  2.73304842e-02\n -2.03045104e-02 -2.45557521e-02 -3.44669889e-03  1.29649406e-02\n  3.22340988e-02 -1.32162180e-02  3.78281027e-02 -4.27398421e-02\n  1.42525053e-02  8.37171823e-02  3.09127476e-02  1.17423786e-02\n  6.91081583e-02  1.30820498e-02 -1.07209831e-02 -2.10680906e-03\n  4.95719211e-03 -1.02009624e-03 -5.77510986e-03  4.73414510e-02\n -3.00503504e-02 -3.59182991e-02 -2.45586336e-02 -2.17193905e-02\n -2.08961219e-02 -2.29210742e-02 -2.80032977e-02  1.52165564e-02\n -1.65842436e-02  4.57603335e-02  1.44828766e-06 -5.61360009e-02\n -3.94317359e-02 -1.73282661e-02  1.10615166e-02 -2.93712430e-02\n  8.55174810e-02  3.19165811e-02  4.22134530e-03 -2.52286508e-03\n -6.43932000e-02  4.01510037e-02  4.70006950e-02 -4.48253714e-02\n -1.41513813e-03  3.66889797e-02 -8.11715703e-03 -4.46259156e-02\n  2.88627185e-02  2.14875434e-02  5.65632023e-02 -7.84562319e-04\n  5.81800705e-03  2.19479781e-02  3.44444998e-02 -5.85746132e-02\n -3.14422622e-02  9.82944295e-03 -2.18724068e-02  2.72080489e-02\n  5.10382988e-02  2.24636458e-02  2.58076116e-02 -6.15553483e-02\n -5.12524135e-03 -2.24703737e-02  2.49784882e-03 -3.52916196e-02\n -1.16497632e-02  2.93239839e-02 -3.97467874e-02 -2.59171855e-02\n  1.10591054e-02 -1.20284809e-02 -1.90009009e-02 -2.27706600e-02\n  6.18728735e-02  1.14581380e-02 -1.07487626e-02  4.07132506e-03\n -3.67226219e-03  8.43563527e-02 -2.39394954e-03 -3.37652229e-02\n  1.20446840e-02  2.25450788e-02  1.10169696e-02 -1.64049696e-02\n  5.59397414e-02  1.17083488e-03 -9.31880530e-03  2.62642857e-02\n -2.74639539e-02 -2.27501988e-02  8.93268362e-03 -7.66405538e-02\n  3.55793536e-02  1.20696472e-02  5.86612076e-02  3.63901295e-02\n -1.65351499e-02 -5.73400129e-03  1.45409070e-02  1.96258314e-02\n -7.27106109e-02  2.72692945e-02 -8.13957080e-02 -2.11496335e-02\n -3.26966271e-02  7.29898438e-02  3.90958190e-02 -4.96897846e-02\n -6.63388595e-02 -1.01196701e-02  6.42161875e-04 -1.88012272e-02\n -1.35782529e-02  1.69789605e-02 -9.91971418e-03  2.28155497e-02\n -4.12469879e-02 -3.77093293e-02 -7.57177593e-03  2.83327959e-02\n -3.02903876e-02 -2.85167098e-02  2.07434315e-02  2.25604214e-02\n -9.28365719e-03 -1.51147833e-02 -8.98459088e-03  6.70140386e-02\n -1.28544942e-02 -5.78452088e-02  2.47775298e-03 -7.97296911e-02\n -2.81679835e-02 -2.84210779e-02 -3.25452425e-02 -6.56164214e-02\n -2.47204620e-02  2.01430097e-02  1.01629477e-02 -5.48637062e-02\n  8.94648675e-03  4.05956618e-02 -4.44052890e-02  1.02671042e-01\n -7.56809115e-02  1.26552181e-02  1.37866579e-03 -1.13331880e-02\n  1.51405660e-02  2.56871488e-02 -5.24983183e-03  4.51805294e-02\n  4.59146686e-02  5.85043104e-03  6.26491150e-03  7.20547792e-03\n -5.88717051e-02 -1.00635946e-01 -6.11553602e-02 -1.08829951e-02\n -5.83557002e-02  4.36093472e-02 -3.92436385e-02  4.10913564e-02\n  5.64262178e-03  5.37047759e-02  2.42957957e-02  1.05013326e-01\n  6.05127029e-02  2.84899641e-02  5.55243762e-03 -3.09707951e-02\n -1.55448066e-02 -1.46294916e-02 -1.43787283e-02 -3.49355722e-03\n -8.83707330e-02  2.19522649e-03 -2.93946192e-02 -2.35651769e-02\n  1.60144772e-02 -5.43737002e-02 -3.02929245e-02 -7.79132126e-03\n  1.89319085e-02 -3.13578583e-02  4.60174680e-02 -5.50660631e-03\n -1.92169063e-02 -4.88135181e-02  8.63954145e-03 -4.77152430e-02\n  2.17958838e-02 -4.88891192e-02  7.63864294e-02  1.30082771e-01\n  3.28815654e-02 -5.33933975e-02  8.30223709e-02  7.57148629e-03\n -2.32078675e-02  4.61315811e-02  2.62386221e-02  1.36163775e-02\n -4.62059826e-02 -2.14045327e-02 -1.82300620e-02  2.83758324e-02\n  1.28799668e-02  2.15108078e-02 -4.15302766e-03  3.15550305e-02\n  1.82425436e-02 -3.74762639e-02  2.08042860e-02  2.08608937e-02\n -2.82959430e-03 -5.35163693e-02 -4.71888147e-02  1.35965738e-02\n  1.38235912e-02  9.01751686e-03 -3.65398708e-03  5.73994629e-02\n  2.57573053e-02 -2.01233313e-03 -5.14804944e-02 -2.00652517e-02\n -4.66603264e-02  3.39639969e-02  1.17061762e-02 -5.82824051e-02\n  2.77182721e-02  3.31477113e-02  6.75387383e-02  1.22633018e-02\n  5.67695983e-02 -6.65890053e-02  5.69646060e-02 -7.09204301e-02\n  3.76802348e-02 -1.43325031e-02  4.47984375e-02  7.28084426e-03\n  2.85710488e-02  1.60099398e-02  3.72759327e-02 -2.62777582e-02\n -2.46258285e-02 -2.06304975e-02  3.72276478e-03 -4.67101634e-02\n  8.38323124e-03 -1.35120293e-02  3.39612849e-02  3.49291484e-03\n  4.99470299e-03 -2.49034129e-02  7.93784333e-04 -3.46210860e-02\n -4.31761034e-02  3.18461545e-02 -4.72946931e-03  1.70705188e-02\n -2.31329408e-02 -4.39915759e-03  1.13313077e-02  2.25792192e-02\n  3.83138545e-02 -1.36187421e-02  1.68776456e-02  3.40576060e-02\n -3.42961587e-02 -2.40255930e-02 -4.41743918e-02 -3.07683609e-02\n -4.96942215e-02  5.87186776e-02  3.34627815e-02 -3.17160413e-02\n -7.49719329e-03  3.18871289e-02 -2.39836518e-02  2.70108543e-02\n  2.77622230e-02 -4.32922244e-02  2.50458159e-02 -2.90878844e-02\n -5.16465493e-03 -3.24391341e-03  1.94145795e-02  4.15609553e-02\n  4.53335270e-02 -4.18853667e-03  8.84133205e-03 -3.26370895e-02\n  2.16132705e-03  2.40324698e-02  5.22514284e-02 -3.98226120e-02\n -4.63361889e-02  3.35683487e-02  8.99985433e-02  2.09509823e-02\n -3.21147405e-02  4.97947820e-02 -5.90330772e-02 -3.41108479e-02\n -7.22306669e-02 -1.34600475e-02 -4.47860770e-02  6.01749718e-02\n -3.98624782e-03 -5.05394153e-02  8.97321180e-02 -7.15939677e-04\n  1.99675821e-02  3.08200088e-03 -2.99170837e-02  2.10373215e-02\n  7.66325668e-02 -2.41141766e-02  1.48944026e-02  2.33445171e-04\n -3.57108861e-02 -2.00601127e-02 -2.78608240e-02  3.36323306e-02\n -6.52984530e-02  2.57370248e-03 -2.62748804e-02  2.23488789e-02\n  1.96563937e-02  2.94369552e-02 -1.44276265e-02 -2.28903387e-02\n -7.76160210e-02  1.66622142e-03  8.07723925e-02  3.26537266e-02\n -2.33838372e-02  4.45028879e-02  1.28065404e-02  1.65519267e-02\n -5.23447176e-04 -2.67643575e-02 -2.80446447e-02 -1.69686824e-02\n -1.78932026e-02 -2.25125272e-02 -3.50899086e-03 -9.78177879e-03\n  7.08928751e-03  1.03520146e-02 -4.23031040e-02 -1.43829808e-02\n -8.08087960e-02  2.69559585e-02  2.88853701e-02 -3.09571866e-02\n -3.26542407e-02  2.75605381e-03  1.02422247e-02 -2.35671792e-02\n -2.33161505e-02  5.81696406e-02  2.19720807e-02 -1.17143495e-02\n  1.01174209e-02 -2.66882428e-03 -6.03606440e-02 -2.37942976e-03\n  1.21415406e-02 -4.11002859e-02 -4.51718643e-02 -1.60665307e-02\n  2.70442460e-02  3.09809074e-02 -3.05432342e-02  3.15632485e-03\n -6.68478161e-02 -1.29532870e-02 -4.91549075e-03 -3.23550291e-02\n -3.58997770e-02 -4.72941734e-02 -1.03332186e-02  4.78735194e-04\n -7.98223615e-02  4.37410586e-02  2.22239434e-03  4.04256135e-02\n  6.82347082e-03 -1.96469575e-02  3.53155695e-02  2.63391398e-02\n  3.70172337e-02 -7.35288300e-03  1.09556749e-01  5.15761897e-02\n -5.43018468e-02 -4.83204499e-02 -5.97818531e-02 -5.58765978e-03\n  6.76242262e-02 -1.12661384e-02  4.74265181e-02 -6.16931589e-03\n -8.67325999e-03 -1.69603527e-02 -1.58700030e-02 -6.41136691e-02\n -4.75617908e-02  1.82289444e-02  6.15704432e-02 -1.81176811e-02\n -4.63466346e-02 -1.21487258e-02  3.53394002e-02  1.43153798e-02\n -1.51322945e-03 -4.43292446e-02  8.88635230e-04 -3.44144069e-02\n -1.51215037e-02  7.00254133e-03 -2.48295208e-03 -6.46390244e-02\n -5.57476804e-02  7.15366192e-03  7.20427837e-03 -8.45418051e-02\n -1.68811772e-02  6.69234549e-04  4.24135625e-02  9.27098654e-03\n  5.17644221e-03  1.82485823e-02  1.43265156e-02 -1.22386374e-01\n -1.10516297e-02  2.88278852e-02  6.96114302e-02 -4.52436022e-02\n -3.79772182e-03  1.01903854e-02 -6.72446284e-03 -2.29705274e-02\n  6.06019422e-02  7.16592092e-03 -2.32902430e-02  4.78281453e-02\n  4.67650816e-02  3.98564376e-02 -5.32545783e-02 -1.64170079e-02\n  8.58898915e-04  5.65902935e-03  3.52774933e-02 -2.19531581e-02\n  1.59286782e-02  1.41221192e-02 -2.00276058e-02 -9.17218905e-03\n -7.49452505e-03  2.47107036e-02 -2.48925504e-03 -9.48616304e-03\n  4.92534135e-03 -2.19699480e-02 -6.24688230e-02  2.18430795e-02\n  2.42757834e-02  5.86127490e-02 -5.02397008e-02  3.77099402e-03\n -1.80244390e-02 -2.82009076e-02 -2.79995631e-02 -1.48973903e-02\n -3.07533308e-03 -2.61211246e-02  2.79021319e-02 -2.04358483e-03\n -2.52628494e-02 -3.60019244e-02 -3.50774080e-02 -4.76166494e-02\n  1.70159491e-03  5.43572940e-02  5.06276004e-02  4.04567569e-02\n  1.00103272e-02 -2.36803927e-02  1.92706790e-02  5.35508106e-03\n -1.91528480e-02  2.40947725e-03 -2.84320526e-02  2.67460588e-02\n -2.68261265e-02  1.71120428e-02 -1.85380150e-02  5.52691147e-03\n -3.09624802e-02 -3.61106507e-02 -4.97021191e-02  6.04570918e-02\n  1.61241926e-02 -4.77947854e-02 -3.17425579e-02 -7.21791480e-03\n  2.30332687e-02 -1.35370474e-02  6.81777075e-02 -2.20571272e-02\n  1.39310835e-02 -7.71866413e-03  1.99044030e-02  5.11281611e-03\n -7.50671551e-02  3.81815545e-02 -8.49779770e-02  1.39642861e-02\n  1.07679078e-02 -6.52924925e-02 -1.85855962e-02  2.07402129e-02\n  4.84880898e-03  3.73609737e-02 -3.75557207e-02  1.61507502e-02\n -4.15056869e-02 -3.32206041e-02  5.85096329e-02  4.62735929e-02\n  5.73100746e-02  2.75436435e-02  4.90288064e-02  5.45493104e-02\n -6.84350030e-03 -4.08428535e-03 -3.86421382e-02  2.65611894e-02\n -2.24672351e-02  1.40256025e-02  3.93021060e-03 -4.97294685e-33\n -3.59437205e-02 -5.74628450e-02 -1.62865128e-02  9.00254678e-03\n  1.60071328e-02 -1.28748419e-03 -1.16924280e-02  5.32272682e-02\n  1.11453226e-02 -1.07059609e-02 -2.21116026e-03  1.72310579e-03\n  2.94323582e-02 -2.58768201e-02  4.24533933e-02  1.73520483e-02\n -1.25585869e-02  9.55471490e-03  1.68609172e-02 -8.00950527e-02\n  1.16722668e-02  2.90282853e-02  7.11322501e-02 -4.01582569e-02\n  1.51641283e-03  1.55115621e-02 -3.80987749e-02 -1.73638910e-02\n  7.47093046e-03  2.56600287e-02 -1.90735031e-02  9.33269411e-02\n  8.26741336e-04  3.29675004e-02 -1.44453831e-02  2.23382246e-02\n -9.78794787e-03 -5.95429912e-02  2.51312107e-02  7.84727465e-03\n -3.35632339e-02  2.06995346e-02  2.14453470e-02 -3.03651635e-02\n -3.38789374e-02  3.42657045e-02  8.02799389e-02 -2.27102488e-02\n -1.45141222e-02  1.35908630e-02 -3.66875641e-02  5.68338810e-03\n  1.32126007e-02 -3.84590738e-02  2.29348205e-02 -3.02357171e-02\n  5.74902864e-03 -6.41770810e-02 -3.17523740e-02  4.55485955e-02\n -1.53187504e-02  3.03061772e-02  6.04593642e-02  1.04117710e-02\n  2.54260544e-02  5.37067838e-02  8.20452720e-02 -4.19960497e-03\n  6.58357795e-03  6.83037518e-03  3.24908309e-02  1.74423326e-02\n  6.07533753e-02  8.56453925e-02  4.81057130e-02 -9.37992185e-02\n  4.93356995e-02  2.36274602e-04  1.70228854e-02  3.32530811e-02\n  3.10481396e-02  2.79911049e-02 -2.56981142e-02 -3.17143947e-02\n  1.03422655e-02 -1.64685361e-02 -4.46704216e-02 -4.09626588e-02\n  1.29443863e-02  2.18537985e-04  1.89387016e-02  2.98334397e-02\n -3.28490287e-02 -4.15589474e-02  2.17367448e-02 -2.91569363e-02\n -3.56767070e-03 -4.43830248e-03 -1.84462015e-02  5.87964244e-03\n  1.65849365e-02 -3.31492499e-02 -2.50569899e-02  2.18881853e-02\n  3.69399004e-02  4.92964080e-03 -5.22426777e-02  1.24047222e-02\n -5.31707965e-02 -2.49955952e-02  7.60653813e-04 -4.11682203e-02\n  4.60143015e-02 -2.12279912e-02  2.31682360e-02 -4.59492579e-02\n -1.29493950e-02  1.74553448e-03 -4.47838940e-03  3.03012095e-02\n -3.20938304e-02 -5.02362698e-02 -1.07084615e-02  6.86716195e-03\n -5.05223386e-02  6.73037115e-03 -1.80894174e-02  3.50810401e-02\n  5.49725480e-02 -1.24864271e-02 -7.71613931e-03  1.99220050e-02\n  2.18156728e-07  4.85302508e-02  5.60069419e-02 -2.00262796e-02\n  5.47962673e-02 -2.99007986e-02 -1.53709818e-02  3.61846900e-03\n  7.77632650e-03 -9.64170974e-03 -1.96539946e-02  5.73286321e-03\n  7.15564843e-03  8.37491546e-03  2.20809933e-02 -4.68556099e-02\n -2.16812119e-02  2.27468014e-02  3.64910699e-02 -7.91412592e-03\n  2.46194284e-02  9.09671858e-02  9.61854309e-02  5.92731088e-02\n -1.38591742e-02 -6.40719943e-03 -5.35838231e-02 -1.11716166e-02\n -1.05274292e-02  3.35113294e-02 -2.62926258e-02  3.67871076e-02\n -7.31061120e-03  2.26462316e-02  8.51114765e-02 -2.73483116e-02\n -4.49619032e-02  8.26552585e-02  2.72787884e-02 -2.61984356e-02\n  6.16995990e-02 -9.19354148e-04 -2.93136034e-02  1.78981456e-03\n -1.66344941e-02  1.99334715e-02  2.98925079e-02  3.50755975e-02\n -1.89114921e-02 -3.96314077e-02 -5.20372938e-04  5.12182713e-02\n -1.46389212e-02  8.99124425e-03 -3.68658081e-03  7.81543646e-03\n  2.73975078e-02 -2.42043957e-02 -3.49381031e-03 -6.62653008e-03\n  3.11317313e-02 -3.71047445e-02  3.21330990e-05  1.61609314e-02\n -5.71104772e-02  1.29826050e-02 -6.20196834e-02 -2.54406221e-02\n  1.13652132e-34  4.66185063e-03  2.03389302e-02 -2.14782474e-03\n -1.75573342e-02  2.12787408e-02  1.70300659e-02  7.33859986e-02\n  1.07082278e-02  3.89475189e-02 -2.34981850e-02 -5.90743609e-02].\nReshape your data either using array.reshape(-1, 1) if your data has a single feature or array.reshape(1, -1) if it contains a single sample."
     ]
    }
   ],
   "source": [
    "cosine_distances(context_emb, definition1_emb)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b35e9dc7-9967-4d53-88d7-081c572b3097",
   "metadata": {},
   "source": [
    "Сначала у нас будет ошибка, потому что эта функция ожидает двумерные массивы. А у нас два одномерных вектора. Но это можно поправить функцией reshape - она превратит вектор с размерностью (768) в массив с размерностью (1, 768)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "a589fa78-cc29-44e9-955c-08f23a9749fa",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.8994292]], dtype=float32)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_distances(context_emb.reshape(1, -1), definition1_emb.reshape(1, -1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "b04efedb-2848-4ee9-92d4-acb6cfa57e91",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.6817975]], dtype=float32)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_distances(context_emb.reshape(1, -1), definition2_emb.reshape(1, -1))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6300a220-b500-41fa-ac54-6160b6065783",
   "metadata": {},
   "source": [
    "Вместо расстояния можно рассчитать близость (то же самое только вычтенное из единицы). Еще при работе с векторами удобно (и быстрее) считать близость не по одному вектору, а между всеми сразу."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "4ba8470c-8e1e-4c83-bdca-8f3460e4b1bd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.10057081, 0.3182025 ]], dtype=float32)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(context_emb.reshape(1, -1), [definition1_emb, definition2_emb])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7b233f46-ecec-40da-aca1-4395ec06da29",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "8514594f-76f1-4d0e-a17d-e3734b1f841d",
   "metadata": {},
   "source": [
    "# WSI"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "db809372-23da-40fc-96a1-2a1da4677f51",
   "metadata": {},
   "source": [
    "WSI можно представить как задачу кластеризации контекстов, в которых встречается слово.\n",
    "\n",
    "Кластеризация - задача разбиения объектов на группы (кластеры). Кластеризация это задача обучения без учителя. \n",
    "\n",
    "Основное предположение в кластеризации - объекты, близкие в признаковом пространстве, близки на самом деле. Все алгоритмы кластеризации так или иначе вычисляют близость между объектами (это может быть, например, косинусное расстояние) и стараются организовать кластеры так, чтобы между объектами одного кластера было небольшое расстояние.\n",
    "\n",
    "Так как считать близости между всеми объектами может быть долго, данные могут распологаться в пространстве странным образом (без четких границ, например) или иметь выбросы, в конкретных алгоритмах кластеризации используются дополнительные предположения и разные трюки. Часто к ним привязывается гиперпараметр(ы), который нужно настраивать самому. Сделать это обычно непросто из-за того, что у нас нет конкретного критерия качества, а сами параметры могут быть непонятными. \n",
    "\n",
    "Возьмем данные диалоговского соревнования 2018 года - https://github.com/nlpub/russe-wsi-kit/blob/initial/data/main/wiki-wiki/train.csv (только train часть для простоты)\n",
    "\n",
    "В данных примеры использования целевых слов и индексы значений, в которых они использованы\n",
    "\n",
    "В качестве метрики используется Adjusted Rand Index. https://en.wikipedia.org/wiki/Rand_index , http://scikit-learn.org/stable/modules/generated/sklearn.metrics.adjusted_rand_score.html  \n",
    "Это специальная метрика для кластеризации. Ее нужно использовать, потому что нельзя напрямую сравнить индекс кластера и индекс значения. Индексы значений проставлены в случайном порядке и индексы кластеров тоже рассчитываются случайно. Поэтому нужно проверять то, что два контекста находятся в одном кластере и одном значении или разных кластерах и разных значениях, все остальные случаи - ошибка.  ARI именно это и считает, плюс там есть нормализация, которая позволяет исключить, что случайная кластеризация получит хорошие метрики."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "5a41783a-a141-4392-90e1-f13d20e0e21a",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# !pip install pandas"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d85e951b-0026-4d1d-b588-89b1b6c72564",
   "metadata": {},
   "source": [
    "Посмотрим на данные"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "fc1725c8-dc59-462c-a965-7a974eaeac17",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "2e8c9c76",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('https://raw.githubusercontent.com/nlpub/russe-wsi-kit/initial/data/main/wiki-wiki/train.csv', sep='\\t')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "e857f6e6",
   "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>context_id</th>\n",
       "      <th>word</th>\n",
       "      <th>gold_sense_id</th>\n",
       "      <th>predict_sense_id</th>\n",
       "      <th>positions</th>\n",
       "      <th>context</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>замок</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0-5, 339-344</td>\n",
       "      <td>замок владимира мономаха в любече . многочисле...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>замок</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>11-16, 17-22, 188-193</td>\n",
       "      <td>шильонский замок замок шильйон ( ) , известный...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>замок</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>299-304</td>\n",
       "      <td>проведения архитектурно - археологических рабо...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>замок</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>111-116</td>\n",
       "      <td>топи с . , л . белокуров легенда о завещании м...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>замок</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>134-139, 262-267</td>\n",
       "      <td>великий князь литовский гедимин после успешной...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>95</th>\n",
       "      <td>96</td>\n",
       "      <td>замок</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>163-168, 213-218, 364-369</td>\n",
       "      <td>без каминов и очагов . у входа в башню было по...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>96</th>\n",
       "      <td>97</td>\n",
       "      <td>замок</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>221-226, 276-281</td>\n",
       "      <td>v . в том сражении шотландцы потерпели сокруши...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>97</th>\n",
       "      <td>98</td>\n",
       "      <td>замок</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0-5, 16-21, 179-184</td>\n",
       "      <td>замок данноттар замок данноттар ( ) расположен...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98</th>\n",
       "      <td>99</td>\n",
       "      <td>замок</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>232-237</td>\n",
       "      <td>. благодаря поэме байрона заключение бонивара ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99</th>\n",
       "      <td>100</td>\n",
       "      <td>замок</td>\n",
       "      <td>1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>5-10, 150-155, 289-294</td>\n",
       "      <td>года замок утратил свое военное значение , так...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>100 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    context_id   word  gold_sense_id  predict_sense_id  \\\n",
       "0            1  замок              1               NaN   \n",
       "1            2  замок              1               NaN   \n",
       "2            3  замок              1               NaN   \n",
       "3            4  замок              1               NaN   \n",
       "4            5  замок              1               NaN   \n",
       "..         ...    ...            ...               ...   \n",
       "95          96  замок              1               NaN   \n",
       "96          97  замок              1               NaN   \n",
       "97          98  замок              1               NaN   \n",
       "98          99  замок              1               NaN   \n",
       "99         100  замок              1               NaN   \n",
       "\n",
       "                    positions  \\\n",
       "0                0-5, 339-344   \n",
       "1       11-16, 17-22, 188-193   \n",
       "2                     299-304   \n",
       "3                     111-116   \n",
       "4            134-139, 262-267   \n",
       "..                        ...   \n",
       "95  163-168, 213-218, 364-369   \n",
       "96           221-226, 276-281   \n",
       "97        0-5, 16-21, 179-184   \n",
       "98                    232-237   \n",
       "99     5-10, 150-155, 289-294   \n",
       "\n",
       "                                              context  \n",
       "0   замок владимира мономаха в любече . многочисле...  \n",
       "1   шильонский замок замок шильйон ( ) , известный...  \n",
       "2   проведения архитектурно - археологических рабо...  \n",
       "3   топи с . , л . белокуров легенда о завещании м...  \n",
       "4   великий князь литовский гедимин после успешной...  \n",
       "..                                                ...  \n",
       "95  без каминов и очагов . у входа в башню было по...  \n",
       "96  v . в том сражении шотландцы потерпели сокруши...  \n",
       "97  замок данноттар замок данноттар ( ) расположен...  \n",
       "98  . благодаря поэме байрона заключение бонивара ...  \n",
       "99  года замок утратил свое военное значение , так...  \n",
       "\n",
       "[100 rows x 6 columns]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(100)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4dc6fbab",
   "metadata": {},
   "source": [
    "С помощью функции groupby мы можем сделать словарь: слово, значения которого надо найти - всего его контексты и размеченные значения."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "5cbc002e",
   "metadata": {},
   "outputs": [],
   "source": [
    "grouped_df = df.groupby('word')[['word', 'context', 'gold_sense_id']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "900d5088",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    word                                            context  gold_sense_id\n",
      "383  бор  бор ( элемент ) бор — элемент тринадцатой груп...              1\n",
      "384  бор  бор - углерод - кремний семейство сплавов на о...              1\n",
      "385  бор  с большим выделением теплоты , образуется окси...              1\n",
      "386  бор  это объясняется прежде всего тем , что у компл...              1\n",
      "387  бор  совсем малочисленна . элементарный бор в приро...              1\n",
      "388  бор  действующем при месторождении горно - химическ...              1\n",
      "389  бор  b c ) . при нагревании в атмосфере кислорода и...              1\n",
      "390  бор  собственных минералов бора в чужих минералах о...              1\n",
      "391  бор  бор - углерод - кремний семейство сплавов на о...              1\n",
      "392  бор  с другими галогенами с образованием тригалоген...              1\n",
      "393  бор  и в стабильны и входят в состав природного бор...              1\n",
      "394  бор  номером . обозначается символом b ( ) . в своб...              1\n",
      "395  бор  и взаимные переходы которых определяются темпе...              1\n",
      "396  бор  . оксид бора formula_  — типичный кислотный ок...              1\n",
      "397  бор  бор ( растение ) бор ( ) — род растений семейс...              2\n",
      "398  бор  каштакский бор каштакский бор — бор , находящи...              2\n",
      "399  бор  бор ( растение ) бор ( ) — род растений семейс...              2\n",
      "400  бор  сосновый бор ( город ) сосно́вый бор — город (...              2\n",
      "401  бор  образом искусственный остров , поэтому для сое...              2\n",
      "402  бор   км к северу от города бузулук в обширной прир...              2\n",
      "403  бор  образовавшийся таким образом сросткинский сосн...              2\n",
      "404  бор  григоровский бор григоровский бор — историческ...              2\n",
      "405  бор  ичалковский бор площадью гектаров отнесен к ка...              2\n",
      "406  бор  его трассе ( на старых картах отчетливо видно ...              2\n",
      "407  бор  . в xix веке в серебряном бору размещался арти...              2\n",
      "408  бор  пески , мощность которых местами достигает мет...              2\n",
      "409  бор  заказника и его статус определяются положением...              2\n",
      "410  бор  каштакский бор каштакский бор — бор , находящи...              2\n",
      "411  бор  на территории бора действовал государственный ...              2\n",
      "412  бор  . фридлянда и инженера а . а . белоголового бы...              2\n",
      "413  бор  до $ . в настоящее время на территории серебря...              2\n",
      "414  бор  серебряный бор есть теория , что название сере...              2\n",
      "415  бор  бузулукский бор бузулу́кский бор — островной м...              2\n",
      "416  бор  . во второй половине xx века , сразу после вел...              2\n",
      "417  бор  , лента бора срастается с соседней лентой касм...              2\n",
      "418  бор  была существенно увеличена в дальнейшем на тер...              2\n",
      "419  бор  бузулук в обширной приречной котловине , имеющ...              2\n",
      "420  бор  ичалковский бор специальным распоряжением сове...              2\n",
      "421  бор  к металлургическому району города . древостой ...              2\n",
      "422  бор  , которые понравятся и детям , и взрослым . в ...              2\n",
      "423  бор  колонии цапли , а в лесу гнездились соколы — п...              2\n",
      "424  бор  серебряном бору , в отдельный подвид сосны обы...              2\n",
      "425  бор  окрестностях семипалатинска . ширина бора на з...              2\n",
      "426  бор  года №  ичалковский бор , его пещеры и карсты ...              2\n",
      "427  бор  серебряный бор есть теория , что название сере...              2\n",
      "428  бор  нашлось необходимого обоснования ( по краеведч...              2\n",
      "429  бор  бор ( лес ) различают следующие основные групп...              2\n",
      "430  бор  ичалковский бор специальным распоряжением сове...              2\n",
      "431  бор  , км² . от самого села до бора можно дойти по ...              2\n",
      "432  бор  озеро на дне холодной пещеры даже в теплое вре...              2\n",
      "433  бор  серебряный бор есть теория , что название сере...              2\n",
      "434  бор  ленточный бор ле́нточные бо́ры — сосновые трав...              2\n",
      "435  бор  в окрестностях барнаула , составляет —  км . н...              2\n",
      "436  бор  также в сосновом бору открыта секция биатлона ...              2\n",
      "437  бор  экспресс банк , мособлбанк , внешпромбанк , ба...              2\n",
      "438  бор  каштакский бор каштакский бор — бор , находящи...              2 \n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "\n",
    "for key, _ in grouped_df:\n",
    "    print(grouped_df.get_group(key), \"\\n\\n\")\n",
    "    break"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a2c8896f",
   "metadata": {},
   "source": [
    "Давайте рассмотрим несколько популярных алгоритмов кластеризации."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "12ab73ba",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.cluster import KMeans, DBSCAN, AffinityPropagation\n",
    "import numpy as np\n",
    "from sklearn.metrics import adjusted_rand_score\n",
    "\n",
    "from IPython.display import Image\n",
    "from IPython.core.display import HTML"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "95684cf1",
   "metadata": {},
   "source": [
    "### DBSCAN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "72cafda9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src=\"https://media.giphy.com/media/lCL2GQewp7fkk/giphy.gif\"/>"
      ],
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(url= \"https://media.giphy.com/media/lCL2GQewp7fkk/giphy.gif\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "04959085",
   "metadata": {},
   "source": [
    "DBSCAN - это density based алгоритм. Он работает итеративно и объединяет в кластеры объекты, которые лежат плотными кучками. Объекты, лежащие отдельно, считаются выбросами и не кластеризуются. \n",
    "\n",
    "Если более детально, то в DBSCAN есть два основных параметра - порог близости (epsilon) и минимальное количество объектов (min_samples). Вначале выбирают объекты, в радиусе epsilon от которых лежит больше min_samples объектов. Такие точки формируют кластеры и объекты попавшие в радиус добавляют в эти кластеры. Далее для каждого объекта попавшего в кластер находятся объекты, которые лежат в радиусе epsilon и они также добавляются в кластер (и так далее пока новых объектов не будет). Визуально DBSCAN похож на расспространение болезни передающейся по воздуху.\n",
    "\n",
    "DBSCAN сам находит количество кластеров и умеет отфильтровывать выбросы. Параметрами можно настраивать объём окрестности и таким оброзом влиять на кластеры."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "868c932a",
   "metadata": {},
   "source": [
    "Интерактивная визуализация есть тут - https://www.naftaliharris.com/blog/visualizing-dbscan-clustering/"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2bb517a5-277e-4481-8558-bba73b351999",
   "metadata": {},
   "source": [
    "### Affinity Propagation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "bb16d70c-15b9-4eaa-accf-11f8c066ab92",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src=\"http://www.ritchievink.com/img/post-14-affinity_propagation/preference_median.gif\"/>"
      ],
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(url=\"http://www.ritchievink.com/img/post-14-affinity_propagation/preference_median.gif\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4577a8aa-42af-4474-ac5e-9d22fe01fece",
   "metadata": {},
   "source": [
    "В AP хитрым образом ищутся \"экземпляры\", объекты в выборке, которые могут использоваться как прототипы. Ближайшие к прототипам объекты записываются в этот кластер.\n",
    "\n",
    "Приемущества AP - не нужно подбирать количество кластеров.  \n",
    "Недостатки - медленно и тратится много памяти, параметры не очевидные.\n",
    "\n",
    "AffinityPropagation работает с не очень большими дасетами, поэтому код ниже может долго работать, когда выбран этот алгоритм."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "65edb59a-c322-4938-8b03-9be8e3a8161f",
   "metadata": {},
   "source": [
    "### Agglomerative clustering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "37bf32b3-7f36-4517-906c-8e45f285de01",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src=\"https://www.datanovia.com/en/wp-content/uploads/dn-tutorials/003-hierarchical-clustering-in-r/figures/002-agglomerative-clustering-cutree-cut-dendrogram-1.png\"/>"
      ],
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(url= \"https://www.datanovia.com/en/wp-content/uploads/dn-tutorials/003-hierarchical-clustering-in-r/figures/002-agglomerative-clustering-cutree-cut-dendrogram-1.png\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4ece5bfb-e70e-45d8-9d68-8230f5c9be53",
   "metadata": {},
   "source": [
    "В агломеративной кластеризации каждому объекту припывается свой кластер и потом ближайшие кластеры постепенно объединяются, пока не получится 1 кластер или не получится количество, заданное как параметр. "
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "41333ba0",
   "metadata": {},
   "source": [
    "### Kmeans"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "2364359f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src=\"https://media.giphy.com/media/hgCM9JNzlqAr6/giphy.gif\"/>"
      ],
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(url= \"https://media.giphy.com/media/hgCM9JNzlqAr6/giphy.gif\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "573e349f",
   "metadata": {},
   "source": [
    "KMeans наверное самый известный алгоритм кластеризации. Выбирается количество кластеров, случайно раскидываются центры этих кластеров, объекты относятся к ближайшим кластерам, центры кластеров пересчитываются в среднее значение попавших в него объектов."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fcd64707",
   "metadata": {},
   "source": [
    "Более интерактивную версию можно посмотреть вот тут - https://www.naftaliharris.com/blog/visualizing-k-means-clustering/"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c460df00",
   "metadata": {},
   "source": [
    "KMeans достаточно быстрый, но выбор количества кластеров часто очень не тривиальная проблема."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8e979e78-d6b4-4f03-b65d-9194f96c7dbf",
   "metadata": {},
   "source": [
    "Про эти и другие алгоритмы можно подробнее почитать вот тут - https://scikit-learn.org/stable/modules/clustering.html"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "58c8a8a9-cfe8-45f3-87b2-5cb3d233c1f7",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "e3636005-9215-4399-a030-c8e1d4383876",
   "metadata": {},
   "source": [
    "Теперь давайте попробуем решить нашу задачу с помощью алгоритмов кластеризации и sentence transformers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "2d5f9eeb",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/mnefedov/miniforge3/envs/myenv/lib/python3.11/site-packages/sklearn/cluster/_kmeans.py:1416: FutureWarning: The default value of `n_init` will change from 10 to 'auto' in 1.4. Set the value of `n_init` explicitly to suppress the warning\n",
      "  super()._check_params_vs_input(X, default_n_init=10)\n",
      "/Users/mnefedov/miniforge3/envs/myenv/lib/python3.11/site-packages/sklearn/cluster/_kmeans.py:1416: FutureWarning: The default value of `n_init` will change from 10 to 'auto' in 1.4. Set the value of `n_init` explicitly to suppress the warning\n",
      "  super()._check_params_vs_input(X, default_n_init=10)\n",
      "/Users/mnefedov/miniforge3/envs/myenv/lib/python3.11/site-packages/sklearn/cluster/_kmeans.py:1416: FutureWarning: The default value of `n_init` will change from 10 to 'auto' in 1.4. Set the value of `n_init` explicitly to suppress the warning\n",
      "  super()._check_params_vs_input(X, default_n_init=10)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.07963918459393977\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/mnefedov/miniforge3/envs/myenv/lib/python3.11/site-packages/sklearn/cluster/_kmeans.py:1416: FutureWarning: The default value of `n_init` will change from 10 to 'auto' in 1.4. Set the value of `n_init` explicitly to suppress the warning\n",
      "  super()._check_params_vs_input(X, default_n_init=10)\n"
     ]
    }
   ],
   "source": [
    "ARI = []\n",
    "\n",
    "for key, _ in grouped_df:\n",
    "    # вытаскиваем контексты\n",
    "    texts = grouped_df.get_group(key)['context'].values\n",
    "\n",
    "    # создаем пустую матрицу для векторов \n",
    "    X = np.zeros((len(texts), 768))\n",
    "\n",
    "    # переводим тексты в векторы и кладем в матрицу\n",
    "    for i, text in enumerate(texts):\n",
    "        X[i] = embed(text)\n",
    "\n",
    "    # выбираем один из алгоритмов\n",
    "    # cluster = AffinityPropagation(damping=0.9)\n",
    "    cluster = KMeans(3)\n",
    "#     cluster = DBSCAN(min_samples=1, eps=0.1)\n",
    "    \n",
    "    cluster.fit(X)\n",
    "    labels = np.array(cluster.labels_)+1 \n",
    "\n",
    "    # расчитываем метрику для отдельного слова\n",
    "    ARI.append(adjusted_rand_score(grouped_df.get_group(key)['gold_sense_id'], labels))\n",
    "    \n",
    "print(np.mean(ARI)) # усредненная метрика"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7bd5bdd5-2fed-4053-8e01-681db7b19cc5",
   "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": "3.11.5"
  },
  "vscode": {
   "interpreter": {
    "hash": "02e6746e832aa8eb1abf7db22a4c6670d44d17f146748c440a1ed1ace17345ab"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}