1.2 Spectral Clustering News Stories.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Spectral Clustering News Stories\n",
    "\n",
    "The first step is to gather some data.  Our goal here will be to use goose3 to parse a Google News page, and use a small custom function to generate a list of news articles that we want to parse.\n",
    "\n",
    "The goal of clustering these news articles is to identify topics. Google will likely give us a few potential tags we could use for each article including the overarching list of topics such as:\n",
    "  * U.S.\n",
    "  * World\n",
    "  * Sports\n",
    "  * Technology\n",
    "  * Entertainment\n",
    "  * etc..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# For Scraping Google News\n",
    "import bs4\n",
    "import lxml #xml parser\n",
    "import ssl\n",
    "from bs4 import BeautifulSoup as soup\n",
    "from urllib.request import urlopen\n",
    "\n",
    "# For Scraping Articles\n",
    "from goose3 import Goose\n",
    "\n",
    "from tqdm import tqdm\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.cluster import SpectralClustering\n",
    "from sklearn.manifold import TSNE\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "\n",
    "def scape_news_links(xml_news_url):\n",
    "    '''\n",
    "    Provide a url which will return an XML file to parse. Returns the links to news stories.\n",
    "    \n",
    "    Params\n",
    "    ------\n",
    "    xml_news_url: string\n",
    "        A URL which will return \n",
    "    '''\n",
    "\n",
    "    context = ssl._create_unverified_context()\n",
    "    Client=urlopen(xml_news_url, context=context)\n",
    "    xml_page=Client.read()\n",
    "    Client.close()\n",
    "\n",
    "    soup_page=soup(xml_page,\"xml\")\n",
    "\n",
    "    news_list=soup_page.findAll(\"item\")\n",
    "\n",
    "    return [news.link.text for news in news_list]\n",
    "\n",
    "def scrape_news_articles(topics=[]):\n",
    "    if topics == []:\n",
    "        topics = ['WORLD', 'SPORTS', 'TECHNOLOGY', 'BUSINESS', 'ENTERTAINMENT', 'SCIENCE', 'HEALTH']\n",
    "    stem = \"https://news.google.com/news/rss/headlines/section/topic/\"\n",
    "    g = Goose()\n",
    "    articles = []\n",
    "    for topic in topics:\n",
    "        links = scape_news_links(stem + topic)\n",
    "        for link in tqdm(links):\n",
    "            try:\n",
    "                article = g.extract(url=link)\n",
    "                articles.append({\n",
    "                    'title': article.title,\n",
    "                    'tag': topic,\n",
    "                    'text': article.cleaned_text\n",
    "                })\n",
    "            except:\n",
    "                pass\n",
    "    return articles"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "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>tag</th>\n",
       "      <th>text</th>\n",
       "      <th>title</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6</td>\n",
       "      <td>U.S. To Lift Tariffs On Steel And Aluminum Fro...</td>\n",
       "      <td>U.S. To Lift Tariffs On Steel And Aluminum Fro...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6</td>\n",
       "      <td>Rep. Tulsi Gabbard said she looks forward to a...</td>\n",
       "      <td>Rep. Tulsi Gabbard warns war with Iran would m...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>6</td>\n",
       "      <td>Get YouTube without the ads.</td>\n",
       "      <td>Brexit talks between Labour and the Conservati...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>6</td>\n",
       "      <td>LONDON/OSLO (Reuters) - Iran’s elite Revolutio...</td>\n",
       "      <td>Exclusive: Insurer says Iran's Guards likely t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6</td>\n",
       "      <td>Thousands of marriage-equality advocates celeb...</td>\n",
       "      <td>Taiwan becomes first in Asia to legalize same-...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   tag                                               text  \\\n",
       "0    6  U.S. To Lift Tariffs On Steel And Aluminum Fro...   \n",
       "1    6  Rep. Tulsi Gabbard said she looks forward to a...   \n",
       "2    6                       Get YouTube without the ads.   \n",
       "3    6  LONDON/OSLO (Reuters) - Iran’s elite Revolutio...   \n",
       "4    6  Thousands of marriage-equality advocates celeb...   \n",
       "\n",
       "                                               title  \n",
       "0  U.S. To Lift Tariffs On Steel And Aluminum Fro...  \n",
       "1  Rep. Tulsi Gabbard warns war with Iran would m...  \n",
       "2  Brexit talks between Labour and the Conservati...  \n",
       "3  Exclusive: Insurer says Iran's Guards likely t...  \n",
       "4  Taiwan becomes first in Asia to legalize same-...  "
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pull_new = False\n",
    "if pull_new:\n",
    "    data = scrape_news_articles()\n",
    "    df = pd.DataFrame(data)\n",
    "    df.to_pickle('data/raw_articles.pickle')\n",
    "    df.to_csv('data/raw_articles.csv')\n",
    "else:\n",
    "    df = pd.read_pickle('data/raw_articles.pickle')\n",
    "encoder = LabelEncoder()\n",
    "df['tag'] = encoder.fit_transform(df['tag'])\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Transform Data with TF-IDF\n",
    "The next steps are to do so initial data processing.  We will first want a bag of words representation of the all of our articles, and then we'll scale that matrix by TF-IDF to complete our feature engineering in this case study.\n",
    "\n",
    "The case study wants us to use MITIE, which looks like a MIT homebrewed solution and having never heard of it I'm not super interested in using this one.  Instead I will leverage scikit learn to perform these relatively simple tasks all in one pass."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(490, 19755)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stop_words = [\"a\", \"about\", \"above\", \"above\", \"across\", \"after\", \"afterwards\", \"again\", \"against\", \"all\", \"almost\", \"alone\", \"along\", \"already\", \"also\",\"although\",\"always\",\"am\",\"among\", \"amongst\", \"amoungst\", \"amount\",  \"an\", \"and\", \"another\", \"any\",\"anyhow\",\"anyone\",\"anything\",\"anyway\", \"anywhere\", \"are\", \"around\", \"as\",  \"at\", \"back\",\"be\",\"became\", \"because\",\"become\",\"becomes\", \"becoming\", \"been\", \"before\", \"beforehand\", \"behind\", \"being\", \"below\", \"beside\", \"besides\", \"between\", \"beyond\", \"bill\", \"both\", \"bottom\",\"but\", \"by\", \"call\", \"can\", \"cannot\", \"cant\", \"co\", \"con\", \"could\", \"couldnt\", \"cry\", \"de\", \"describe\", \"detail\", \"do\", \"done\", \"down\", \"due\", \"during\", \"each\", \"eg\", \"eight\", \"either\", \"eleven\",\"else\", \"elsewhere\", \"empty\", \"enough\", \"etc\", \"even\", \"ever\", \"every\", \"everyone\", \"everything\", \"everywhere\", \"except\", \"few\", \"fifteen\", \"fify\", \"fill\", \"find\", \"fire\", \"first\", \"five\", \"for\", \"former\", \"formerly\", \"forty\", \"found\", \"four\", \"from\", \"front\", \"full\", \"further\", \"get\", \"give\", \"go\", \"had\", \"has\", \"hasnt\", \"have\", \"he\", \"hence\", \"her\", \"here\", \"hereafter\", \"hereby\", \"herein\", \"hereupon\", \"hers\", \"herself\", \"him\", \"himself\", \"his\", \"how\", \"however\", \"hundred\", \"ie\", \"if\", \"in\", \"inc\", \"indeed\", \"interest\", \"into\", \"is\", \"it\", \"its\", \"itself\", \"keep\", \"last\", \"latter\", \"latterly\", \"least\", \"less\", \"ltd\", \"made\", \"many\", \"may\", \"me\", \"meanwhile\", \"might\", \"mill\", \"mine\", \"more\", \"moreover\", \"most\", \"mostly\", \"move\", \"much\", \"must\", \"my\", \"myself\", \"name\", \"namely\", \"neither\", \"never\", \"nevertheless\", \"next\", \"nine\", \"no\", \"nobody\", \"none\", \"noone\", \"nor\", \"not\", \"nothing\", \"now\", \"nowhere\", \"of\", \"off\", \"often\", \"on\", \"once\", \"one\", \"only\", \"onto\", \"or\", \"other\", \"others\", \"otherwise\", \"our\", \"ours\", \"ourselves\", \"out\", \"over\", \"own\",\"part\", \"per\", \"perhaps\", \"please\", \"put\", \"rather\", \"re\", \"same\", \"see\", \"seem\", \"seemed\", \"seeming\", \"seems\", \"serious\", \"several\", \"she\", \"should\", \"show\", \"side\", \"since\", \"sincere\", \"six\", \"sixty\", \"so\", \"some\", \"somehow\", \"someone\", \"something\", \"sometime\", \"sometimes\", \"somewhere\", \"still\", \"such\", \"system\", \"take\", \"ten\", \"than\", \"that\", \"the\", \"their\", \"them\", \"themselves\", \"then\", \"thence\", \"there\", \"thereafter\", \"thereby\", \"therefore\", \"therein\", \"thereupon\", \"these\", \"they\", \"thick\", \"thin\", \"third\", \"this\", \"those\", \"though\", \"three\", \"through\", \"throughout\", \"thru\", \"thus\", \"to\", \"together\", \"too\", \"top\", \"toward\", \"towards\", \"twelve\", \"twenty\", \"two\", \"un\", \"under\", \"until\", \"up\", \"upon\", \"us\", \"very\", \"via\", \"was\", \"we\", \"well\", \"were\", \"what\", \"whatever\", \"when\", \"whence\", \"whenever\", \"where\", \"whereafter\", \"whereas\", \"whereby\", \"wherein\", \"whereupon\", \"wherever\", \"whether\", \"which\", \"while\", \"whither\", \"who\", \"whoever\", \"whole\", \"whom\", \"whose\", \"why\", \"will\", \"with\", \"within\", \"without\", \"would\", \"yet\", \"you\", \"your\", \"yours\", \"yourself\", \"yourselves\", \"the\"]\n",
    "vectorizer = TfidfVectorizer(stop_words=stop_words)\n",
    "vectorized = vectorizer.fit_transform(df['text'])\n",
    "dense = vectorized.todense()\n",
    "vectorized.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Some quick spot checks verify that this very large and sparse array looks like it could be the TF-IDF representation of our documents. At the very least, we see a reasonable shape with the correct number of rows. \n",
    "\n",
    "### Time to Cluster: Spectral Clustering\n",
    "With spectral clustering we first calculate the laplacian of a graph representing our data.  Clustering is then done on the relevant eiganvectors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([5, 1, 3, 1, 1, 1, 3, 5, 1, 5, 1, 1, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "       1, 5, 1, 1, 1, 1, 1, 1, 1, 6, 1, 1, 1, 1, 1, 1, 1, 3, 1, 1, 5, 1,\n",
       "       1, 1, 1, 1, 1, 1, 1, 1, 5, 5, 1, 1, 1, 1, 1, 1, 1, 1, 5, 1, 1, 1,\n",
       "       1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 3, 0, 0, 0, 1, 0, 0, 0, 0, 0,\n",
       "       0, 1, 0, 0, 0, 0, 0, 1, 3, 0, 1, 0, 1, 0, 0, 0, 3, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0, 3, 0, 4, 0, 4, 6, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 1,\n",
       "       4, 4, 4, 4, 1, 4, 4, 0, 4, 4, 6, 0, 4, 4, 1, 3, 4, 1, 4, 4, 1, 4,\n",
       "       4, 4, 4, 3, 4, 1, 4, 4, 4, 4, 4, 4, 0, 4, 1, 4, 4, 1, 4, 4, 3, 4,\n",
       "       1, 1, 5, 4, 3, 0, 4, 4, 0, 0, 1, 1, 5, 5, 1, 5, 1, 4, 5, 1, 5, 5,\n",
       "       1, 5, 5, 1, 5, 1, 5, 1, 5, 1, 4, 5, 4, 1, 1, 5, 1, 5, 5, 1, 1, 5,\n",
       "       5, 3, 5, 1, 5, 5, 1, 4, 5, 1, 5, 5, 5, 5, 1, 1, 1, 5, 5, 5, 1, 5,\n",
       "       1, 5, 1, 4, 5, 1, 5, 5, 5, 1, 1, 5, 1, 1, 5, 5, 0, 1, 1, 1, 1, 3,\n",
       "       5, 4, 0, 3, 1, 1, 1, 0, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1,\n",
       "       1, 1, 1, 0, 1, 1, 3, 1, 1, 1, 1, 1, 1, 3, 3, 1, 1, 1, 0, 1, 3, 1,\n",
       "       3, 1, 0, 1, 3, 1, 1, 0, 1, 1, 3, 1, 3, 3, 1, 3, 0, 3, 1, 1, 6, 6,\n",
       "       6, 6, 6, 6, 2, 4, 1, 6, 2, 6, 2, 6, 6, 2, 2, 6, 1, 2, 2, 6, 6, 2,\n",
       "       6, 6, 2, 2, 2, 2, 2, 6, 2, 2, 6, 2, 2, 2, 1, 6, 6, 6, 6, 6, 6, 2,\n",
       "       6, 2, 6, 2, 6, 2, 6, 6, 4, 2, 6, 1, 2, 6, 2, 6, 1, 6, 2, 6, 6, 6,\n",
       "       6, 6, 2, 1, 1, 2, 1, 1, 2, 2, 1, 1, 1, 1, 1, 2, 1, 1, 1, 2, 1, 2,\n",
       "       1, 2, 1, 1, 1, 2, 1, 1, 1, 1, 2, 1, 1, 2, 5, 2, 1, 2, 1, 1, 2, 2,\n",
       "       1, 1, 2, 1, 2, 1, 1, 1, 1, 2, 1, 1, 2, 2, 1, 1, 2, 2, 2, 2, 1, 1,\n",
       "       1, 2, 2, 2, 1, 3])"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = SpectralClustering(n_clusters=7,\n",
    "                           affinity=\"nearest_neighbors\",\n",
    "                           n_neighbors = 30)\n",
    "\n",
    "model.fit(vectorized)\n",
    "base_predictions = model.labels_\n",
    "base_predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So we've successfully clustered our articles, and there is some indication that we're even doing a pretty decent job for some of the possible labels.  However, currently we can only look at the set of labels.  Might there be a way to visualize this?  I will try out both PCA and T-SNE to process the TF-IDF representation to reduce the dimensions to 2."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "pca = PCA(n_components=2)\n",
    "p_embedding = pca.fit_transform(dense)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAs8AAAHiCAYAAAAXqCHCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzs3XeYXGXd//H3d9r2Te/dFFqoCaF3EAhVRAELRRBB87M+iogPoGBXQB94UBTRB1CQItJbEJBOQockkIT03nazbdq5f3+cSZjdmdlsspM5y+Tzuq69MnOfs+d8puSe795zn3PMOYeIiIiIiGxZKOgAIiIiIiIfFyqeRURERES6SMWziIiIiEgXqXgWEREREekiFc8iIiIiIl2k4llEREREpItUPMtWMzNnZuOKtK2nzeyCAstGZ/YVydx/xMzOKcZ+t4WZfd7MHu9k+eFmtqSUmT4OzOw2M7sy6BwiPZX61ILLd4g+1cyuNLPbuvH7C8zs6GJm2sYcnb6e5UTFc0Ayb/ZWM2sys5VmdouZ1WYtP9bMnjWzjWa22syeMbOTO2zj8ExH+L0t7OtwM/My+8r+OWB7Pb7twTl3vHPurwHu/3bn3Cc33e/uB17mQ64t81o0ZF7v3YuTtkv7/0HWe6HNzNJZ99/dxm1eYGZPFzmqyBapT916Zdin7mZmj5vZejPbYGYzzWxqcdIW3GfBP1aKuI96M7vOzBZl3mdzM/f7F3Ef3SrgIff1LGcqnoN1knOuFtgH2Bf4IYCZnQ7cBfwfMBwYBFwOnNTh988B1mX+3ZJlzrnaDj8vFulxyLablnkP9AOeBm4t1Y6dcz/d9F4ALgJezHpv7NZx/U2jVSI9mPrUHdsDwBP4r+9A4OtAY5CButtvmlkMmA7sBhwH1AMHAmuBKd0OWCQ72ueDiucewDm3FHgEmGhmBlwDXOWc+5NzrsE55znnnnHOfXnT75hZNXA68DVgvJlN3tb9Z/5yvtrMXsj8VfuAmfUzs9vNrNHMXjWz0R1+baqZzTezNWb2KzMLZW3vS2Y2K/PX/2NmNipr2TFmNjsz0no9YFnLwmb268w25wMn5Ml5Qeb2uWb2XGb99Wb2oZkdn7XumKxRpifN7IZCf1VnRqA+nbl9cGb0Y2rm/tFm9kb2PjO3n838+puZ5+yMrO19x8xWmdlyMzuvK6+Bcy4F3AHsmrWdKWb2YmYEZbmZXZ/pSDHftZn9NJjZW2Y2MbOsIvO8LMqMwP3ezKq6kqPD8xLJPBdfNbO5wGwzG2dmrsN6z2Wem92B64FDMs/JmqzV+pr/FfHGzGMas7V5RLpKfermZTtMn2r+KOwY4I/OuUTm53nn3KbtH25mS8z/xm2N+d9UfD7r9zvtN83sFDN7I/P6zTOz48zsJ8AhwPWZzNdn1nVm9jUz+wD4INP2WzNbnPn9mWZ2SL7HkcfZwEjgU8659zLv3VXOuauccw/neR7+YmZXZ91vN/XFzC4xs6WZ13GOmR1lZscBPwDOyDyONzPr9jKzmzPP+9LMezqc9do9b/7n0DrgyuzXM+t5uMjMPsi8p27I/H/c9N78Tea1+NDMplnWlKKeTsVzD2BmI4CpwOvATsAI4O4t/NqngSb80ZTH8P+DdceZwBeBYcBY4EXgFqAvMAu4osP6nwIm44/wnAJ8KfNYTsX/T3gaMAD4D/D3zLL+wD34o0H9gXnAQVnb/DJwIrB3ZtunbyHzfsCczLZ+Cdy86T8m8DfgFfwR3Sszj62QZ4DDM7cPBeYDh2Xdf6bjLzjnDs3c3DMz4nRn5v5goBf+83g+cIOZ9dnC49g0uvB54KWs5jTwrczjOwA4CvhqZtknM9kmAL2BM/BHIgB+kWnfCxiXyXL5ljJ04mT8UbxOp5Q4594GpgH/yTwn2V8pfg74b/z30yLgqm7kEemU+tTNdqQ+dS0wF7jNzE41s0F51hmceWzD8L9duMnMdsosK9hvmtkU/G8tvovf3x4KLHDOXYb/ekzLZJ6Wta9T8Z/PTQMir2a23Rf/ubzLzCrzZOzoaOBR51xTF9btVOaxTgP2dc7VAcdmHsejwE+BOzOPY8/Mr/wVSOE/H3vjf+5kT1HZD/+1HQj8pMBuT8T//NgT+Gxmn+C/N4/Hf072wX++Pj6cc/oJ4AdYgN9RbwAWAv8LVOF3fA6o3MLvPwlcl7l9FrAaiBZY93DAy+wr+6cms/xp4LKs9X8DPJJ1/yTgjaz7Djgu6/5XgemZ248A52ctCwEtwCj8D6OXspYZsAS4IHP/KeCirOWfzOwrkpVz07rnAnOz1q3OrDsY/6/0FFCdtfw24LYCz89RwFuZ24/idw4vZe4/A5yWtc/nOjwP4zo8z62b8mbaVgH7F9jv05nnZgOQABqAozp5zb8J/DNz+0jgfWB/INThOW0Gxma1HQB8uIX3U7vHlmmLZB7joVlt4wDXYb3ngHMzty8Anu6w/Dbg91n3TwbeCfr/oH7K6wf1qZv+/+/Ifepw/G+/5mVen2eB8VnbSm16jTJt/8D/o77TfhP4A3BtgX1ufg47PI4jt/B+W4//hwL4f4wUei6fAH7ehff+0ZnbfwGu7vAcLsncHpd5/o6mw3u7Ywb8qS9xoCqr7Szg31mv3aIO28j3eh7c4fn+ftZ78ytZy44m673Z03808hysU51zvZ1zo5xzX3XOtfLR6OGQQr+UGVU5Arg90/QvoJIOX8l1sCyzr+yf5qzlK7Nut+a5X0t7i7NuLwSGZm6PAn5r/lSDDfjzBw3/r/ih2b/n/P8x2dsZSu52O7Mia1stmZu1me2sy2rrmLejF4EJmZGKvfBHGEZkRnWm4HfAXbXW+VMwNmkh97nL9nXnXG/81+9E4G4z2wPAzCaY2YNmtsLMGvFHBvoDOOeewv+QuAFYaWY3mVk9/shUNTAz6zV4NNO+rTp77rpqRdbtLT0nIttKfeoO3Kc655Y456Y558biP2/NmX1vsr7Da7Tped5SvzkCvyDfGu2en8zUk1nmT6/ZgD+a3pUD/tbSyXt3azjn5uIPwlwJrDKzO8xsaIHVRwFRYHnWc/IH/FHmTbry2VCo7+/43izG50zJqHjueebgv4k+3ck6X8R/7R4wsxX4X5tU0v2vGbfGiKzbI4FlmduL8f+azP5AqXLOvQAsz/69zNeB2dtZTu52t8Vy/Dm21QXytpP5QJgJfAN/RDQBvAB8G5jnnFtT6HeLxfnz2P6D/7XjpqOVbwRm44+c1ON/dWtZv/M759wk/ANJJuB/pbgG/4N5t6znv5fzD6La5nhZt5th8/zQTQYXWFekJ1Cf2n672+Lj2Kcuxh9cmJjV3MfMarLub3qet9RvLsafepN3V1tqz8xvvgR/2kKfzIBJA1n9eSeeBI7tkLszzfh/CGyS3T/jnPubc+5g/OLY4U9XaZc3YzH+yHP/rOek3rU/mLw7/f1y/G8KNin4fuqJVDz3MJmRg28D/21m55l/ipqQ+Qdd3JRZ7WzgR/h/0W/6+TRwgpn1K1HU75pZn8yIzTeATfPTfg9cama7weYDDj6TWfYQsJuZnZY5KODrtP+P/Q/g62Y2PDOn7fvbEsw5txCYgX8AQ8z800d1PKq+o2fw54Jtmov3dIf7+awEPrEtGfPJ5NwV2HSauDr8I8WbzGxn4OKsdfc1s/3MLIrfWbYBaeecB/wRuNbMBmbWHWZmx1IcKzI/X8gc8HEhfie8yUpgeCaXSODUp+4YfWrmufuR+Qc1hzKj3F+i/XEkAD/KPIZD8L/tu6sL/ebNwHnmH1wXyizbeSsy1+FPGVkNRMzscvyzZnTFrfiF7D1mtnNm//3MP/Ax32n43sA/+LSvmQ3GH2km85h2MrMjzawC/zOjFf/Ymk2PY7RlDlR1zi0HHgd+k/V/ZqyZHUZx/AP4Rua57I3/x8XHhornHsg5dzf+AWBfwv+reCVwNfAvM9sfGA3c4JxbkfVzP/6o5VkFNjvUcs9J2tlIzJb8C39k4Q38DvzmTPZ/4v8le0dmqsE7+AcFkBlt+Azwc/yvosYDz2dt84/4B+q8CbwG3NuNfJ/Hn7O2Fv+5uxP/r+hCnsHv4J4tcD+fK4G/Zr7S+uw25tx0lHYTfif5Q+fcI5ll/4V/oN1G/Ofmzqzfq8+0rcf/6nEt8OvMskvw3wsvZV6DJ/EPmuq2TCHyZfxR8DX4c+hezlrlCfyjy1dmRvBEAqc+dYfoUxP4r+OT+IMO72TynZu1zgr8PnMZ/hSdi5xzszPLCvabzrlXgPOAa/FHjJ/ho0GD3wKnm382id8VyPYY/tz19/H76za6OE3BORfHnw88G79/bcQ/cLM/7fveTW7Ff70X4Be/2Z8bFfjvlTWZ52Igfl8O/oGyAGvN7LXM7bOBGPAe/vN2N0WaQoL/3nwceAv/wN6H8f/ASHf2Sz2F+Z+FIuXNzO4EZjvnOh7hLiIiW+nj1qea2eH4B8QN39K6Unrmnxbx9865UVtcuQfQyLOUpcy0hrGZr5qOwz/1031B5xIR+ThSnyrFZGZVZjbV/OsJDMM/deM/g87VVR+Lk1GLbIPB+F9R9sM/ddPFzrnXg40kIvKxpT5VisnwjzO4E3/u9UN073oEJVWUaRuZv0J/C4SBPznnft5h+bnAr4ClmabrnXN/6vaORURkq6nPFhHZdt0eeTb/Uo03AMfg/zX6qpnd75x7r8Oqd7r2V98REZESU58tItI9xZjzPAX/qkTzM+dyvAN/LpSIiPQ86rNFRLqhGHOeh9H+lCtL8K933tGnzexQ/FO1fCtzAvOC+vfv70aPHl2EeCIipTdz5sw1zrnuXNlxe9kufTao3xaRj6+t6bOLUTznu0JOx4nUDwB/d87Fzewi4K/AkTkb8i+4cCHAyJEjmTFjRhHiiYiUnplt6VLIQSlanw3qt0WkPGxNn12MaRtLaH9ZxeF8dFlRAJxzazMn+gb/xNiT8m3IOXeTc26yc27ygAE9ccBGRORjr2h9dmZd9dsiskMpRvH8KjDezMaYWQw4E7g/ewUzy74izcnArCLsV0REtp76bBGRbuj2tA3nXMrMpuFffjIM/Nk5966Z/RiYkbnE6dfN7GT8Sy+uo/3lMkVEpETUZ4uIdE+PvTz35MmTnebOicjHlZnNdM5NDjpHKanfFpGPq63ps3V5bhERERGRLlLxLCIiIiLSRSqeRURERES6SMWziIiIiEgXqXgWEREREekiFc8iIiIiIl2k4llEREREpItUPIuIiIiIdJGKZxERERGRLlLxLCIiIiLSRSqeRURERES6SMWziIiIiEgXqXgWEREREekiFc8iIiIiIl0UCTrA9uDSayHxAlglVByCWWXQkUREpICUF2dh8wySXhsja/ahOtIn6EgiIgWVXfHsNf8FNv4GCIMZYNDnD1hs34CTiYhIR0tb3uL+JT8EwDmHR5qDBpzP3n0/HXAyEZH8ymrahku+BxuvAeJAC7hmcE249V/Bubag44mISJaUF+f+JT8k4bWQ8FpIulbSLsELq//MqrYPgo4nIpJXeRXPLfcAiXxLIP6fUscREZFOLGyekbc97ZK8t+HREqcREemasiqeoRnwcpsd4FpLHUZERDqR9NpwzuW0OzwSnvpsEemZyqp4topjwarzLElB7KCS5xERkcJG1uyDRzqnPWqVjKs/NIBEIiJbVlbFMxWHQezArAI6BFRC3bexcL8gk4mISAfVkT4cNOB8IlaBZT6OolbJiJp9GFMzJeB0IiL5ldXZNsxC0Pt6iD+Diz8K1GDVn8aiuwUdTURE8ti776cZVr0H7214lITXyrj6QxlTM8Xvz0VEeqCyKp4hU0BXHoFVHhF0FBER6YKBleMZOHh80DFERLpEf9qLiIiIiHSRimcRERERkS5S8SwiIiIi0kUqnkVEREREukjFs4iIiIhIF6l4FhERERHpIhXPIiIiIiJdpOJZRERERKSLVDyLiIiIiHSRimcRERERkS5S8SwiIiIi0kUqnkVEREREuigSdIDtZc3Stfz+O//Hyw/NJBwJc8zZh/Gln36OqprKoKOJiEgHKS/BS2v+yrsbHiHl4oysmcShAy+mV2xI0NFERNopy+K5tamVr025lA2rGvDSHgAP3fQkc2bM47fPXY2ZBZxQRESyPbT0Sha3vEHaJQD4sOkllrW8w9ljb6Eq3CvgdCIiHynLaRvTb3+OlsaWzYUzQDKe5MO3FjLrpfcDTCYiIh2tjS9kScubmwtnAIdH0rXxzvqHA0wmIpKrLIvn92fOo605ntPunGP+W4sCSCQiIoWsjX+IEc5pT7sEK9pmB5BIRKSwsiyex0wcQbQid0ZKKBRi+ATNnxMR6Ul6x4bhuVROe4gI/SvGBJBIRKSwsiyeh44bQjKe2xHX9q1lz8N3CyCRiIgUUhPph0c6p90Bu/c+sfSBREQ6UZbF818vvzNvezhkOlhQRKSHeWPdvRi5fbNhhC0aQCIRkcLKsnj+8O3885pXLlxNKpk7Ii0iIsFZ0vImHrl9cyQUY238wwASiYgUVpbFc+9B9Xnbq3tVE47kHpQiIiLB6R0bmnfk2XNJaqMDAkgkIlJYWRbPZ116GhXVFTntfQf3obWpLYBEIiJSyD59P0PYYnmWhHC4kucREelMWRbPJ130SQ7/7IE57cvnr+Bnn/9tAIlERKSQAZXjOHLwt3LaU66NuxZ+k7RLBpBKRCS/siyezYw1S9fmtKcSaV578i3WrVgfQCoRESmkNbWeELkHB6Zcgg+bXg4gkYhIfmVZPAOsWrQmb3skFmHdig0lTiMiIp3ZmFqNR+4Is+dSNKfy9+ciIkEo2+J5zyN2y3twoJd2jNhpaACJRESkkKFVE4laVU67YQyu3CWARCIi+ZVt8XzWpadRVVdJOPLRQ6yoruC8q8+goir3YEIREQnOJ+oOpHdsaLsDByNWwfDqvRhUtVOAyURE2su9hnWZGDiiP79/7VfcdvXdvDH9HfoO6c0Zl5zKgSfvG3Q0ERHpIGwRPjPqOl5bdxezG6YTtigTe09l9z4nBx1NRKSdsi2eAQaNGsB3/nhx0DFERKQLoqEq9ut/Nvv1PzvoKCIiBZXttA0A5zxcahEunXvmDRER6XmakmtoTK7EOZ3fWUR6prIdeXbxZ3ENl4LXBKRxsX2wXtdi4X5BRxMRkQ7WJ5bw8NKrWJ9YhGHURPpz/NDLNN9ZRHqcshx5dqm5uPXTwFsNtAIJSMzArf+SRjNERHqYlJfg7oXfZE18PmmXJOUSNCSXcc/i79Kabgg6nohIO0Upns3sODObY2Zzzez7nax3upk5M5tcjP0W4ppvhZzzhaYgvQBSs7bnrkVEPhZ6Ur/9YdNLJF0cOlyK27k0sxumb6/diohsk24Xz2YWBm4Ajgd2Bc4ys13zrFcHfB3Y/peKSi8G0nkWhCG9YrvvXkSkJ+tp/XZzai2el8ppT7k4G5Mrt+euRUS2WjFGnqcAc51z851zCeAO4JQ8610F/BJoK8I+OxfbH6jMbXdJiO623XcvItLD9ah+e3DVzvj1fHtRq2Jo9cTtuWsRka1WjOJ5GLA46/6STNtmZrY3MMI592AR9rdFVn0mhHoB0azGKqg6HQsPKkUEEZGerEf124Mqd2Zo1UQi9tEFrMIWo1dsCJ+oPXB7715EZKsU42wblqdt88Q1MwsB1wLnbnFDZhcCFwKMHDly2wOF6qHffbjmG6FtOoTqsOpzoOq0bd6miEgZ6VH9tplx8oireGPdfbzb8DCeS7NT/ZFM6ncGoTwj0iIiQSpG8bwEGJF1fziwLOt+HTAReNrMAAYD95vZyc65Gdkbcs7dBNwEMHny5G6dFsPC/bD6H0L9D7uzGRGRctTj+u2wRZnU7zNM6veZbd2EiEhJFGPaxqvAeDMbY2Yx4Ezg/k0LnXMNzrn+zrnRzrnRwEtATgcsIiIlo35bRGQbdbt4ds6lgGnAY8As4B/OuXfN7MdmdnJ3ty8iIsWlfltEZNsV5QqDzrmHgYc7tF1eYN3Di7FPERHZduq3RUS2TVleYVBEREREZHtQ8SwiIiIi0kUqnkVEREREuqgoc557ondfmMPTdz6PhYyjPncIO+07LuhIIiJSwIbEMt7b8CitXiNjaqYwunY/neNZRHqksiyeb/zWLTz0x+kkWhMAPPzHJzn9Oydx7o/ODDiZiIh09EHjszy+/Bd4LoVHmjkN0xlUNYFTR/yCsJXlx5SIfIyV3bSNua9/yEN/fJJ4SxznHM454i0J7vrVAyz5YHnQ8UREJEvKS/DEil+RcnE80gAkXSsrWucwp2F6wOlERHKVXfH84gMzSLYlc9qd83jpAZ3fX0SkJ1ne+h6W52rhKdfG7EYVzyLS85Rd8RyrjBGK5M6TC4VCxCpjASQSEZFCwhbFufxX9Y5YRYnTiIhsWdkVz4d99gBCodxRDAcc8un9Sh9IREQKGly1M9FQZU57xCqZ2OeEABKJiHSu7IrnwaMH8v9uuIBYZZTK2kqqaiuJVcX43l+m0WdQ76DjiYhIlpCFOXnET6gI1RINVROxSsIWY2LvqYyp0YCHiPQ8ZXkY83HnHckBJ03mlYdfx0LGfifsQ12f2qBjiYhIHoMqJ3DBuH+woPll2tIbGVG9F71iQ4OOJSKSV1kWzwC9+tdzzNmHBR1DRES6IBKKMa7ukKBjiIhsUdlN2xARERER2V5UPIuIiIiIdJGKZxERERGRLlLxLCIiIiLSRSqeRURERES6SMWziIiIiEgXqXgWEREREekiFc8iIiIiIl2k4llEREREpItUPIuIiIiIdJGKZxERERGRLooEHWB7cl4TrvWfkHwHIhOw6k9jod5BxxIRkTw8l2bexudZ0PwK1eHe7Nb7eHrHhgUdS0SknbItnl16OW7taeBawLUClbjmG6HfnVhkbNDxREQkS8pLcO+i77ImPo+kayNEhDfW/5Njh3yfcfWHBB1PRGSzsp224Rp/Ct76TOEM0AZuI67h8kBziYhIrlkNj7M6PpekawPAI0XKxXl8xa9IeYmA04mIfKRsi2fizwBeh0YHyZk4lwoikYiIFDCn8SlSLp7TbsDKtjmlDyQiUkD5Fs8W3Xxz4fsVzHymloa1YSBMOT9sEZGPo0ioIm+7c46IxUqcRkSksPKd81x5Kvf97nFu+dkA4vEQkYjDDE79Si++fJ0FHU9ERLLs0fskFjW/jqP9N4OeS9EnNjygVCIiucp2CPaGHwzgDz8aTLwtDM5IJUMkEyHuvznOv//+XNDxREQkS8pL4kjntDs8nl55QwCJRETyK8viec3StTxy8zO4jlOegXhLgruvfbD0oUREJC/nHM+u+l/A5SzzSPP+xn+T1rEqItJDlGXx/P7M+USi4YLLN65rKmEaERHpTNxrojXdUHC55zzSLlnCRCIihZVl8dx/WF+cyx3BALCQsf+Jk0qcSEREComGqghZ4QGPPrHhxEJVJUwkIlJYWRbP4/f5BMPGDSEUyn149X1rOevSTwWQSkRE8glbhIm9pxImmmdZlCMHfzOAVCIi+ZVl8Wxm/Pzx/2anKe2vJBiOhrngF1+g7+A+ASUTEZF8Dh74FXbqdTT+mZ03MXauP5qhVRODiiUikqMsi2eA3gPqaVjdiIU+6ojTyTTX/7+b+fDthQEmExGRjsIWATzClj367JjT+BTvbHgoqFgiIjnKtnie9dL7rF/ZgPPaz31OxlP8638fCyiViIjkk/ISzGl8irRrfynulIszc91dAaUSEclVtsXz+pUN7UadN/HSHmsWrw0gkYiIFJJ0rbg8p6oDaOvkTBwiIqVWtsXzLvuPJxnPPS9oRXWMKVP3CSCRiIgUUhmqpybcN88SY1j1HiXPIyJSSNkWz30H9+G0b0ylsqZic1usMsqA4f045pzDAkwmIiIdmRlHDP4GEatg00GDRphoqIqDBlwQbDgRkSyRoANsT+f/7PPsNGU8/7r+EZo2NHPo6ftzyrTjqaqpDDqaiIh0MKZ2P04feQ0z1t7B+uQShlTtxuS+Z9ArNiToaCIim5V18WxmHHLafhz8qSlAAohhljsPWkREeoZBVTtxwvArSLskRqjTi6eIiAShrItnAK/1Qdj4S/BWgdXhar6C1ZyvIlpEpAda2TqH6SuuY018HiELs1P9URw+6GtEdYVBEekhyrp4dm3ToeEHQFumoQGa/gdHGqv9SqDZRESkvcbkSu5Z9F8kXSsAaecxp3E6G5MrOW3krwJOJyLiK9sDBgFc03VsLpw3a4XmP+BcOohIIiJSwBvr/knaJdu1pV2S5a3vsS6+KKBUIiLtlXXxTHpJ/nYXB9dc2iwiItKpNfH5eOSeYjRkETYklgaQSEQkV3kXz5Gx+dutBqy2tFlERKRTgyt3Jkw0pz3tkvSrGBVAIhGRXGVdPFvtd4COp6WrgtpvYlbWD11E5GNnz76nEgl9dJ5ngLDFGFOzH71iQ4MLJiKSpawrSKs4AOtzI0R2ASogPBLqryJU87mgo4mISAc1kb6cMfp6RtdMIWIVVIZ7ManvZzlu2GVBRxMR2aysz7YBYBUHYRX/CjqGiIh0QZ/YcE4Z8ZOgY4iIFFTWI88iIiIiIsWk4llEREREpItUPIuIiIiIdJGKZxERERGRLlLxLCIiIiLSRUUpns3sODObY2Zzzez7eZZfZGZvm9kbZvacme1ajP2KiMi2Ub8tIrJtul08m1kYuAE4HtgVOCtPJ/s359zuzrm9gF8C13R3vyIism3Ub4uIbLtijDxPAeY65+Y75xLAHcAp2Ss45xqz7tYArgj7FRGRbaN+W0RkGxXjIinDgMVZ95cA+3Vcycy+BnwbiAFHFmG/IiKybdRvi4hso2KMPFuetpwRCufcDc65scAlwA/zbsjsQjObYWYzVq9eXYRoIiKSh/ptEZFtVIzieQkwIuv+cGBZJ+vfAZyab4Fz7ibn3GTn3OQBAwYUIZqIiOShfltEZBsVo3h+FRhvZmPMLAacCdyfvYKZjc+6ewLwQRH2KyIi20b9tojINur2nGfnXMrMpgGPAWHgz865d83sx8AM59z9wDQzOxpIAuuBc7q7XxER2Tbqt0VEtl0xDhjEOfcw8HCHtsuzbn+jGPsREZHiUL8tIrJtdIVBEREREZFDqduiAAAgAElEQVQuUvEsIiIiItJFKp5FRERERLqoKHOee6rmxhZuv/pu/v335wmFQhxzzmGcdemnqKiqCDqaiIh04JzH2xse5PV195LwmhlZM5kDB5xHXXRg0NFERDYr2+I5nUrzrUP+myXvLycZTwJw16/v5/Xpb3Pdc1djlu8aASIiEpR/r/wfZjU8Qcq1ATCncToLml/mi2P+THWkd8DpRER8ZTtt46UHZ7J8/srNhTNAoi3J/LcX8ebT7waYTEREOmpOreXdDY9uLpwBHB5Jr5U3198XYDIRkfbKtnh+5/nZtDXHc9oTrQnenzEvgEQiIlLIqta55LlCOGmXZGnLW6UPJCJSQNkWz0vfz3+lWed5DBzZv8RpRESkM23pBjxSeZfVRXTZbxHpOcq2eF44a2nedudgzB6jSpxGREQ6s6JtTsFlfSvUZ4tIz1G2xXO0IlqwvapGZ9sQEelJwhYBcg/kjliMuuig0gcSESmgbIvnE758FBXVsXZtZsbwCUMYOFJfAYqI9CQ79zqaiMXyLDHG1O5X8jwiIoWUbfF80sXHstcRE6msqSBWGaW6rpJeA+q5/K7vBB1NREQ6GFg5nin9vkDYYkSsgohVErEKjht6GRXh2qDjiYhsVrbneY5EI1z9wKXMmTGPWS++T79hfdn/xH2IxvJP5xARkWDt2/8sdup1JAuaXiZsMcbWHUhluD7oWCIi7ZRt8bzJTpPHstPksUHHEBGRLqiPDmKPPicHHUNEpKCynbYhIiIiIlJsO0Tx7JzDpZfh0quCjiIiIl3QktpAQ2I5zuVeOEVEJEhlP23DJd5k8Wvf5ZUn4lRUpjnopEH0GXcdFhkRdDQREemgObWOh5b8iBVtszGgIlzHcUMvZWTNpKCjiYgAZT7y7NJrueWSb3PxkXX8+Sf9+cOVA/niPvD87efhXDLoeCIiksU5xx0LprG87V0caTzStKY3cN/iS1mfWBJ0PBERoMyL5/eeuYV7b+pNoi1EMhEi3hom0Rbi51/tRfPq6UHHExGRLEta3qQplTu9zuHxwqpbAkgkIpKrrIvn6X9/n0Rb7hWrQmHHq4++FUAiEREpZEHzKwWXrWibVcIkIiKFlXXxvHp5DXmPNXHgmeY8i4j0JLFQdcFlNZE+JUwiIlJY2RbP61asZ+ZTK4HckedUKsy+J5xY+lAiIpKXc45ZGx4ruHzffp8rYRoRkcLKtnh+4tZnyT/sDMd96Rjq+9aVOJGIiBSyom0WLekNeZf1ig7lE7UHljiRiEh+ZVs8r1u2jmQ8ldMejUUYubOmbIiI9CStqQ2Y5X5TCH7xXGiZiEiplW3xvNeRu1NVW5nTHoqE2OOwXQNIJCIihQyq2oV0nlOIRqyCMbX7B5BIRCS/si2ep0zdm0/sMYqK6tjmtsrqCg44aTKf2GNUgMlERKSjmkgf9u5zOhH7aNAjbDFqIv3ZtfexASYTEWmvbK8wGA6H+eX0K3joD4/zxP89RSTSwNRzYhz9hSE4rwXr5KhuEREpvQMHfIlBVTvx+rp7aEyupCpUz6jaKSS9FmKhqqDjiYgAZVw8A0RjjlPOmckpn30QyBw82Pw8rvVG6HcPFu4faD4REfmImRFPN7GidRYeKZpYxdp1C3lz/b18auQvGVKlKXciEryynbbhnMOt/wq03s7mwhmAVvBW4ZquDSqaiIjk8eqaO3hqxbV4fHSwt0eSpGvj0WU/wxU4g5KISCmVbfFM8i1IzKR94bxJGtqeKHUiEREpIOm18craW/FI513enFrLxjyX7hYRKbXyLZ5T75C/cM6wWOFlIiJSUhsSywhZuJM1HBH12yLSA5Rv8RwaClZoSncYqj5T0jgiIlJYTaRv3lPVbTKgYjzVukS3iPQA5Vs8VxwCVkfehxidhNVeXPJIIiKSX3WkN5+oPYBwntHlqnAfjh92WQCpRERylW3xbBbB+v0donsBUf8nNBB6/ZpQv9swff0nItKjfHLIJYyvO5SwRQlblKhVManvGVww7g7qo4OCjiciApT5qeosPAzrdwfOWw8uhYUHBB1JREQKiIQqOHbo9znC+wbx9EZqIv22MA9aRKT0yrp4BkglU/ztJ49z/42P09bUxp5H7MbF15zL8AlDg44mIiJ5zN/4Ai+t+StNqdX0iY3k4IEXMqpmUtCxRESAMp62scnPv/g7/vGr+2lY3Ui8NcGrj7zBtP0uZc2ydUFHExGRDt5e/xDTV1xDQ3IZaZdkTXweDy65nEXNrwUdTUQEKPPiecWCVbx4/wzirYnNbc45Em0J/nX9IwEmExGRjpzzeHHNn0m5eLv2lIvz/Ko/BZRKRKS9si6eF763hGhFNKc9GU8x++W5ASQSEZFCEl4r8XRT3mXrEotKnEZEJL+yLp6HjRtMMpHKaY9Ew4zZfWQAiUREpJBYqIpIqDLvsvro4BKnERHJr6yL5+EThrL7IbsQrWw/+hyJRfjUN6YGlEpERPIxC7Fvv7OIWPsCOmIVHDjgvIBSiYi0V9bFM8AV9/wXR3/+UKIVUUIhY9w+Y/j1U1cyZIzOGSoi0tNM6nsGBww4h8pQHUaImkh/jhr8bcbWHRR0NBERAMw5F3SGvCZPnuxmzJhRtO15nkc6lSYay50DLSJSbGY20zk3OegcpVTMfts5R9olCVsUMyvKNkVECtmaPrvsz/O8SSgUIhQr+4F2EZGyYGZEdCVYEemBVE2KiIiIiHSRimcRERERkS5S8SwiIiIi0kUqnkVEREREukjFs4iIiIhIF5Xt2TZWNzRx25Ov8eqcxdSkPAYsbmRkr1qOPfcIxkzU1QVFRHqSlBfn7Q0P8n7jv3E4KsO9qA33Z3yvQxlZPUmnqxORHqMsi+fl6xr53E9vp6UtQTLtgefAedTd9zYP3vg4F117LideeEzQMUVEBEi7JHct/CbrEotIuXi7Ze9vfIoxtQdw3NAfqIAWkR6hLKdt/P6BF2lqjfuFM0DIIBqm6aAxtLUmuH7azbz1n/eCDSkiIgB80Pgs6xOLcwpngKRrY+7G/zBz3Z045wWQTkSkvbIsnl+atZC0l3vlRBcN49VWkE6lueSYq/jN+f9LT73CoojIjmJB88skXVvB5R4pXlh9C39fcDHxdHMJk4mI5CrL4rl3bVX+BSHDEikAUokUT//jBV68f+suJbvwvcW8+ujrrFuxvrsxRUQEqIn0wwh3uo4jzdrEIl5cfctWbbspuYYFTa+wLr6oOxFFRDYryznPZx8zmZ/87UnaMoUyACmP6OL1hBLpzU1tzXEeveUpDjxl3y1us3HdRn544s+Z/9YCItEIibYkUy84iq/97kuahyci0g0Te5/AW+vvJ+XSna7nuSRzNj7F4YOnbXGbznlMX3EdsxufIGwxPJdiUOUEThp+FRXh2mJFF5EdUFmOPE+dsjOfP3IfKqIhKkhCKk10eQO1/56bs66X7tocul+ecz0fvDafeEuC5oYWkvEkj97ybx750/RixxcR2aH0iQ3n2CGXEgvVbHHdrs57fmP9fcxpnE7aJUl4zaRcnBWts3ly+TXdjSsiO7iiFM9mdpyZzTGzuWb2/TzLv21m75nZW2Y23cxGFWO/neTha6ccxGM/PZMbznuMP51xF/2nv0so2X5Uo7KmgmO+eNgWt9e0oZnXnnyLVPZINhBviXPvbx8qanYRke2tp/XZAOPqD+bC8XdTHe5XcJ0QEcbXbbnPBnhz/T9zDkBMk+TD5hdJeq3dyioiO7ZuF89mFgZuAI4HdgXOMrNdO6z2OjDZObcHcDfwy+7utyvqawex14Td2XV8K5/92ioiUY9IzANzVFZH2Pe4vTnk9P23uJ2Wja1YKP9T1bRBB6+IyMdHT+6zwxblgAHnECaWsyxEhProIA4aeH6XthX38vfNDkfSyz2rh4hIVxVjzvMUYK5zbj6Amd0BnAJsPhecc+7fWeu/BHyhCPvtktmzzueyqcvw0ikiUUgmjL0OG8g5V32d3Q7cqUvzlQcM70d931rWLF3Xrj0cCTFl6j7bK7qIyPbQo/vs0TX7EQ3fTDqd2NwWC9VwxKBvML7+UMLWtY+tkTWT+aDxaRztp3nURQZQFe5V1MwismMpxrSNYcDirPtLMm2FnA88km+BmV1oZjPMbMbq1au7HSyZSHLZCb+hqcHR0hSmrSVEOmXMeqWBSDTc5QP9zIxv//EiKqpjhML+UxarjFLXp5azr/hMt3OKiJRQ0fpsKH6//fjynxNPN7Vr81yK5tSaLhfOAAcNOJ+KUC1hi/o5CRGxCo4a/G0d5C0i3VKMked8vVDekyeb2ReAyUDeSWvOuZuAmwAmT57c7RMwvz79HTwv9+CSRFuSR25+ip2njO/ytvY9bm+uf/nn3Pvbh1g2dwV7Hr4bJ3/1WHr1r+9uTBGRUipanw3F7bfj6SaWtryNo/3xKSkX5+0NDzCp32e7vK366CC++ImbeXP9v1jW8jZ9KkawV5/T6FsxsjsRRUSKUjwvAUZk3R8OLOu4kpkdDVwGHOZcnstIbQdtzW15PxKc52hp3PoDRkbvNoJv33RREZKJiASmx/bZKZfwR4Xz9NvbMk+5OtKHAwac2/1gIiJZijFt41VgvJmNMbMYcCZwf/YKZrY38AfgZOfcqiLss0v2OmIiqWQqp72yppJDu3CgoIhIGeqxfXZ1uA91kYE57SEijK07qFQxREQ61e3i2TmXAqYBjwGzgH845941sx+b2cmZ1X4F1AJ3mdkbZnZ/gc0VVePq2Uw6rJFwxGEhfyijssaYePBOHHjqli+MIiJSbnpyn72haR0fPl+PlzK8lD+7JJ0IUxXqzf79zylFBBGRLSrKFQadcw8DD3douzzr9tHF2M/WuPe3D3LzpX/BS1fgnMPM+MTEFr7wrfXsf/rphMOdXwpWRKRc9cQ+e2nL29y98HvU7ZnGQg4vDa3rY3z4yChGhA+j+orepY4kIpJXWV5hcMWCVdx86e0k2oxUMoSXDuGljcVzKxk5oZFQ/I6gI4qISIbn0jy09EcQSRKp8LAQhKMQrUqRaAvx3PT5eF63jyEXESmKsiyen//nK7g8/Ww6Bc893BtKc+yLiIh0weq2uTlXAwSIVHqMOHg5znM4Fc8i0kOUZfFsZnnP42kGRhiqTgwglYiI5FXgDBsAOGPiniMIR8ry40pEPobKsjc6+LQpedtDYTj4lF5YtQ48ERHpKQZWjCMaqsxpT8VDrJ4xkm9dckIAqURE8ivL4nngyAF85ddnE6uMEq2IEI0ZsUrj3CumMHzfu7FQddARRUQkwyzEicN/RDRURcQqwYVwqQi1rTvx8/+6iuEj+wUdUURks6KcbaMnOvmrx7HfCZN47t6Xcc5x0KlTGPKJQUHHEhGRPIZU7cr5Y//OBxv/Q2t6A8Or92Rw5S66lLaI9DhlWTwn4kn+esWdPPzHJ4m3JNjr8N3Y/8RJQccSEZE8nHO81/AYL6+5lebUWvpWjGRg5QQVziLSI5XltI2rz7iG+/7nEZrWN5OMJ5nx+JtM2/8S1i55K+hoIiLSwZvr7+PpldezMbUSjxRr4vN5YPF/89Kcp0invaDjiYi0U3bF85IPljPzibdItCY2tznnSLS08MC138BbfQIutSjAhCIisonn0ry05q+kXFu79jQJHp9/A5898VpmvDwvoHQiIrnKrnhe+O5iItHcqwcmEyFmvx6D9Dzcui/inEYzRESCFvea8p7jGaB6UBONDa1ceeldrFy+ocTJRETyK7viefhOQ0kn0znt0ZjH2N1aAQ9cIyReKX04ERFppyJUS9iieZc1r6oCwEs7Hn3ozVLGEhEpqOyK51G7DGfXAycQrcjujB2RmOOQcxpo80L+yfi9NUFFFBGRjJCFmdT3TCJW0a49FQ+x8OkhVPZpI5lMs2ZVY0AJRUTaK8uzbfzovkv4/bf+whO3/ptUIs3wveKM/WEbT9X2w2s0xkXbOKrf7uQf6xARkVLat99ZREJRXl51O3HXRLwxSjiSZrcz52Gfm0fT0lp2rzwi6JgiIkAZjjwDVNVU8s0bT+WB5btwzQfzmXR7M7W7OJKESGPMTVbxzJp7go4pIiKAmbFr9an0nfk9Xrr6QCKVaaI1HpFKj3DMo27kRpaNuAnnCl3DW0SkdMqyeHbppbg1J2Jtd/J6qpZUh4eZBmY3Pk7Ky3+QioiIlE4ymeabX/kL//enZxm0/0JC4fYHdIfCjsb4apa3vhdQQhGRj5Rn8dz0O3BNQJoWl3vmDX8lj7jXXNJcIiKS69mn3mPpknUk4imq+7URyjOhMJXyaE6tLX04EZEOyrJ4Jv4i4I9cDAnH8Y8QbK8iFKU63Lu0uUREJMeMl+fR1poEYPV7fUjFcz+aLOwxuGrnUkcTEclRnsVzqN/mmwdWbSCKw7IK6AiOwwach1l5PnwRkY+T/gPqiUT8/njJ84OJN8RIJz+6NHcqHqZfy/7URQcGFVFEZLOyrB6t5gIw//ygfcMpzqhbwfhoM71CKUZE4pw8YCoT+pwecEoREQE4/qS9CGeK53Q8wnM/nsT8x4bTtKKKhgV1pF8/nM9PujLYkCIiGWV5qjoqp0J6ATTdCKToE05xbM16iB4Cva4mFBkcdEIREckYOrwv/3316fzsyn/S2pIg2RJlzr1jSbyxP1//3lQmHjci6IgiIpuV58izGVZxOGBsmvsMDpKvQssfgwsmIiJ57b7nCCKRULvT0S1etJbf/ephnaJORHqUsiyeAVzTDUCiQ2srtPwD520MIpKIiBQw/bG3ibelyK6TUymPFcs28OZrC4MLJiLSQXlO2wBIvU/Hs2wsmF3JQ7cNYn3DbzjgpCM47IwDiVXoOoMiIkGbP28VbW3Jdm0V9XFGHbWIV9PXkVyzHxP7TKUq3CughCIivvItniO7QHoxmwrop/7Zm2u/M4JU0vDS7/Lqo3P55+8e4tr/XEVFVUWwWUVEdnBjxw+isjK6uYCuG9bEgZe+TijiaIl5vLL2A15bdxdnjr6eXrGhAacVkR1Z2U7bsNqvApUAxFuNa741gkRbCC/tn/6orTnOotlLefTPTwWYUkREAI765O5UVccIhfw+evez3ydSmSYc849bSbk4cW8jz6y6MciYIiJlXDxHd8H6/hkiE/nrL4aQTFjOOvGWBE/f+UIA6UREJFtVdYzr//Ql9jtoPJEY9BnbSMdT8Tsci5pnBhNQRCSjbItnAItNwvrdw2P/GIF/5o1cNb2qSxtKRETyGji4Fz/+xWeZtO/Yzd8SdhQmVuJUIiLtlXXxDNC29jaaG1oLLj/p4mNLmEZERDoz692lvPLCPJa9MrDdVQYB0okQIzgsoGQiIr6yLp5dehmx5C+p75POu7y+fx37Td2nxKlERCQfz3Nc+f27cA7evX08DQvrSLWFSLaGScVDrJ/bm0OGXRB0TBHZwZXv2TYA1/o4Zim++N3l/PHHQ4m3hjcvi1Ua3/nTxQGmExGRbPM+WMHGRv+bwlRbhBd+ug/1IzdSO7iFpmW1HHXA4fSqrQs4pYjs6Mq6eCbxMpDmpHPWEYnArb8ZxLqVUQaPTHD+Ff058OR9g04oIiIZ8+euIpls/01h46I6GhfVUVEZ4cvTjg4omYjIR8q7eE7O2Hzz+M+v4/jPr8taGMElZ2PRnUufS0REcrz4n/cLLou3pXj0gTc46bRJJUwkIpKrrOc84xo6WZjCtd5dsigiItK5D+ev6nT5fXe9UqIkIiKFlW3x7JxHodPTbeZtKEkWERHZsnTK63R5S0uiRElERAor32kb3logDKRoWBvm9usG8fzDvaio8jjp3LWcfN5GIpXHBJ1SREQyWlv94thCHqOPWsqoI5YRjnosn9GfDx4czQEHTwg4oYhIORfPoTogTGuzx7TjJ7BuZYRU0h9ov+VnQ5j12gguu+eoYDOKiMhmAwbW09jQyj4Xv8eAieuIVPgj0aOOXMaQSes5a5evB5xQRKSMp22YVULVKUy/ewCN68KbC2eAeGuIFx8Ns/SDzufXiYhI6XzunIPpNybernAGCEcddQM8VoVfDTCdiIivbItnAKu/nDdfHktbSzhnWTgSZs6r8wJIJSIi+Rx65C4cf87gvMtSro2lLW+WOJGISK7yLp4txvBdjyUayz87ZeDI/iVOJCIinTnioP2pqqzIaQ9blF6xYQEkEhFpr6yLZ4ATLjyacLT9yHMoHKLf0L5MPFjneBYR6UlGVO9NdaQ31uHjKUSE3XodF1AqEZGPlH3xPHDkAH72yGX0Gdx7c5tzDi+dZsn7ywJMJiIiHZmF+PTIaxhQMS67FcNYFZ8bWC4RkU3KvngG6D+8H80NzZvvO8+xfP4qvnPElaSSqQCTiYhIRzWRvrSk12e1OBKuhUeWXsX6xJLAcomIwA5QPHuex63/cxWpVPuT6zvniDfHmfnEWwElExGRfP4+8042tK3Lafdcinc2PBRAIhGRj5R98fzM0q8we8ECvGTu1QY9z2P9ys4u4S0iIqX0p/9M576Nz+BCLmeZR5qNSZ1iVESCVdbFc2N8Ae82z6ffgSnC1Xk64rSngwZFRHoIz3n8efXTrEnVEiK3z45aJaNrpgSQTETkI2VdPK9seY4QjiHHJKkZmSZU8VFnHKlyHPm5Qxg+fkiACUVEZJP1iWZS0TTN6QrmNg0g6X30EZXyjLroICbUHxFgQhGRcr48N1AdHQ1AKAYH3b6RD/9WwbKHYoQrHYecE2ba1y8KNqCIiGxWE6nEDBwwY90oVsfrmFC3koh5LF7dn/935PVEQrGgY4rIDq6si+ehNQdRHa6gMZ0gXGWMOz/OuPPjRHCcOfwSzHLnQYuISDAqw1H2cKN4M7kQosbC5n4sbO4HCTg+sRexUFXQEUVEynvahpnxqZG/p3+kmjCOKB6V5jhuwOn0q/1k0PFERKSD/zn+PCakhkAKLA4kYUrbOK447fSgo4mIAGU+8uy1Pkjdxl9wZu1q3nyzL7Pemcyo8Z9j9Nh9g44mIiIdzJm1jOt++TCrP2hkyPAwe50MUw+fwl6DjiQUKuuxHhH5GCnb4tm1PQUNPyCVbOPHF4zmjedqwS3FwtdS3auea575McPG6mBBEZGeYOXyDXx32m20tiYYO3UhE05eiHPwwppXeLXx95ww/HLG1O4XdEwRkfKdtuGargPauO/m/rzxXC3x1jDxthBtzR7rlm3gSzt/k3t/9xDO5Z4OSURESuu+u2eQTKbpNbqR8SctJBzziFR4hGJp0sS5f8llPLH816S8xJY3JiKyHZVt8Uzav4TrI7f3I94azlnspT3+fOnfePhPT5Y6mYiIdPDhvFWkUmmGH7QCi3g0LKzFS7c/qHt2w5M8vvznASUUEfGVb/EcGQvA6uXRgqvEWxPc9uO7S5VIREQK2GXiMGKxCFX92njxZ3vTvLoSrP03gx4p5je9SHNqbUApRUTKuHi2uu8y791exFs6f4jrVmwoUSIRESnk5NMmUVEdYvU7fWhcVEvtoFbyHSMYtpgu0S0igSpK8Wxmx5nZHDOba2bfz7P8UDN7zcxSZlaS8w1ZbArPPLrlXQ0dO6gEaUREeo6e2Gf36VvLFTcezLKXB+OlwqyfV4+Xyl0v7ZL0jg0vRSQRkby6XTybWRi4ATge2BU4y8x27bDaIuBc4G/d3d/WiFaPBgpfCKWiKsaFvzq7ZHlERILWk/vsoYMG4jy/z573yEjSyTDO+2h5xCrYo/dJVIbrShlLRKSdYow8TwHmOufmO+cSwB3AKdkrOOcWOOfeArx8G9heTrr4kwVr514D6rninv/igJMmlzKSiEjQemyf3b9iLCMmNWNhj5bVVTz/k31Y/U5fkq1hEg1VHDTgAg4ZeFEpI4mI5ChG8TwMWJx1f0mmbauZ2YVmNsPMZqxevbrbwfoO7sMFP/tCTvugUQO4dd717Hvc3t3eh4jIx0zR+mwobr9tZvzg21+mqneKcEWapmU1zPzf3Xj2kkM5Y9At7NX3U5gV/jZRRKQUinGRlHw92TadPNk5dxNwE8DkyZOLcgLmM753Cod+Zn9u/dFdbFjVwBFnHcwRZx5EJFq214cREelM0fpsKH6/PXrwBO646wfc98jjzJ29ip3GjuOEqftTU1vZ3U2LiBRFMSrIJcCIrPvDgWVF2G7RDBkziO/9ZVrQMUREeoIe32dXVVZy1qdODjqGiEhexSieXwXGm9kY4P+3d9/xcVV3+sc/505VL7Yk23KVu7EpbjG9mJ6ASULbbCgJ2SSEkJ5sfrvJhpCyS0jfTYGEJJACCZAAoZeY0IvBmGJwN+5dvUy59/z+GMmWrBlJIGlGvnreeg3WzFzNfY5ljr46c+45W4GLgQ8NwOsOGM/zeOL2p3n45n8ChjMuP5njz1+Ek24dJBERfxvyfTbA1uY3WFH7N1rdWiYVLmJ22fsIO3m5jiUi0v/i2VqbNMZ8GngICAC/sda+YYy5Flhmrb3HGLMA+BtQBpxjjPmmtfaw/p67L7zE23znoi/z7ANJErHUToPLHl7BzJ9M5Qf//CbBoKZviMjwMdT77NaWOD//2/cxc5bihCzGwJbWFby491Y+NOkGikIV2YghIpKRsXZAphYPuPnz59tly5b16zWs18TKB8/iC0sq8dzuo8zTF0zhJ898m0Cg+/bdIiL9YYx5yVo7rJbzGYh++0ufuYkx/3YrgTSbw4bI47Ipt1AQLOvXOUREDvZO+mxfz1uwrXfz8G1RPDf91dlrl6/nmbv719GLiMjA2LBuF7vcN3EyvCGYoJXn99yS3VAiIgfxdfFMcg2b1mQeVXaTHk/e+VwWA4mISCZbNu8jv6Klx2PWNz2TpTQiIun5u3gOzsT00sLC0vzsZBERkR5NnFSB6/a8L0vIaMk6Eckt3xbP1lpou5/qSbGMxwSChrOuWJzFVCIikslT/3yLlj2Zi2PrweFlSzI+LyKSDb4snl2b5Okd13LD7s3sXhTJcD5lydEAACAASURBVJTlzCsWM3VuTVaziYhId39d+zw/id7PxveH8dLs4+JaaHVrOELFs4jkmC+L58e2/4AV9U8TxyFe62RsZVlVcXaDiYhINw9sXc4P196HV2xJeEGW7xtH0nPoWAwq6Rnq4/m8tmcGjtHqSCKSW75b5LglWcvqxsdxSc2bs0mDCYE9ePaGgURbY/YDiohIF79Y8whxkvvvr2ocxb54AVOLdhIJJNncXMb65gqqbLKHVxERyQ7fjTzXJ7YTMOH99ytPSKR5AxBCYcux7z8+e8FERCStnW113R5rTEZoc0MUBOJMKNjHqEADp5TPzkE6EZGufDfyXBIag2vj++/nV3uMXRJj01865j4bwOK5Buv57ncHEZFDzqhoKds7FdDRQJz3jnmdkJMkYKCENkaObmZGYGcOU4qIpPiueswPljK9eDFBc+BCwdpXgqSK5o4xaIObNPz887/LQUIREensU9NOJ+oc2FJwZvF2QsYl0Oltw2DAYw13kPDacpBQROQA3xXPAItHfZ555RcScYrAQuPq9BeYrH15fZaTiYjIwc4YcyRfm/0BxuSVYSyMCTcQcGy34xwnwL7Y2zlIKCJygO+mbQA4JsCiistYVHEZyfhuHsm/kliaTatCkRCe5+E4vvwdQkTkkHH6mCM4fcwRJJMu33/6I5DX2u2YpNdGsNM1LSIiueDrqnH7hp1cMuVruMn0vyMkYkn+7+qbspxKRETSaXMTfPql3/CiKSOZ5poUi8cD276DZ90cpBMRSfF18XztBT9g77ZakvH0Ha2bdHnwt0up212f5WQiInKwG9c8yuv1m9ncUsCyfeNJeA4Jz8G1ByY/NyR28nbzshymFJHhzpfTNgB2bd7DppVbsF73eXOdhSMhtqzaRmlFSZaSiYhIOvdufYm4l1rLeV1TJRuaRlISbiXmBlgy9jUcY0naGHti65hU+J4cpxWR4cq3xXO8LYEx6VZ47ioRS1A5oSILiUREpCeJg6ZjeDjUxgswWDwLjoGACVESGpOjhCIiPp62UT1lFMFwz6POABMOG0fluJFZSCQiIj05LG8kpn132AMs5eEmgo7dv113TeExWc8mItLBt8WzMYZRE3o/btH75g1+GBER6dWRXpyokyRoUiPQAeMSMi6LRmzAetC6N0J5eDxBRytuiEju+HbaBkAoHAHikHaDbghHQ8xYODWrmUREJL2yYsvZxa+xKVbO3lghJeEWJhfuJeh6rPxLDY5jOO6TR+Y6pogMc74tnh/4zWOse72NTIUzWEZPtMw7/fBsxhIRkTT2xDawPfIoYddlWmQ3sHv/c8mEw/ZlFcz+0Hpm5Z+bu5AiIvh42sYt1/yFRFsyw7OWSbNa+fE9K3HshqzmEhGR7pbtvQ2PJE4AYo0BvCRY235zDaMX7GLtXdN49tFtuY4qIsOcb4vnvdtqAUiW5tE6ZzRtMyrxIgcG2ndtieC5Bhtbxq7Ne2iqa85VVBGRYW9fbCMWj1hDiO3LKtnwj2padkcxBkIFLtPPexsTbeGlF9bTkqylObkv15FFZJjy7bSNUZMqWVOZR2xWVeoBC83HTKLokVWEN9eRTMDN3yvnmYcepGHfPVhrOWrxHP79lqspLi/KbXgRkWGmMjqdV59qZPmvp6dm23mw6s4appy1iWnnvU0g5DH9vA2UVW/nN+t+A0BpqJozx/wHI6M1uQ0vIsOKb0eeT/zKOcRnVkEwkLqFUrfGU6djgw6x1gD33lLOnm3NxNsSJGJJXn7kVf7zvf+d6+giIsPOjPD7eeWm6XiJAF48gJcM4CUCrHtwPHUbijAOlE+vw5bsxLUJXJtgb3wjd2z6AjG3KdfxRWQY8W3x/HbIYkOBbo8ba0lUl2AD4HldLyZMJlw2vLaJjW9szlZMEREBVi1rIhSIdHvcTThsfa4Sz0Jql+6u6/e7NsmqhqXZCSkigo+LZzfDttzWgA0abNAc3AcDEAwF2LVpzyCnExGRzjzPYg/eH6XjOdfg2fQrJyVtGw2JnYOYTESkK98Wz2cvnEFeunX0HYM31xA/rxQb7t4Zx9sSTD5y4qDnExGRAxYsmpx20MMJeZjDYzy3ZxJemh9ZIZPHqLwZ2YgoIgL4uHg+5rCJLD6ikmgoAVhwUjdnURPxj5QTf38Z5DsEggf+CiL5Ec684hRGjC7LXXARkWGotKyAT1x9KgC24yNkaT3CY0XRaN5uGcG+tgIMof1fEzAhSsKjqSk8OlexRWQY8u1qG8YYrrn8X/jgy+/ln6tG8MCOapr3tpKoDeCNjhApsXzwnvfReksTz9/3MoWl+Xzgs+/l7H87NdfRRUSGpfPOX8DzT69m2fMbSIx3SU5x8So9SALWsHH78Zy7cCRvNjyMtR7Ti09hwYgP4Zju17eIiAwW3xbPAIYWZldv5plft2BubqUwaHCtIVQCn/tjM6ce9z04LtcpRUSkw57dqZUzgpscgpsdaK+L3bEev7zxSooL81hUcWkOE4rIcOfbaRsAeDt58t587vvDCBJllSSmzcKbNZtY8UR++aliVtY/jGsz7UIoIiLZtvntvQAYDMYaTDJ1C2xyuHnV49THW3KcUESGO18Xz9Z63P2bEcQqx8O4URAJp9Z8Li2mvmgmf3nwBv666ct41s11VBERAVw305Ib8Nc/vsBFT/2YPW0N2Q0lItKJb4tn6zVD3dU0NkWgshwCnebEGQPGsP7JSna3rWFd41O5CyoiIlhr+dF193Xc6fa8weC8ZqhPtPCrtf/IcjoRkQP8WzzvvRDc9UyYb8FLM5LhONRvKCZh21jX9Ez2A4qIyH43/O8j3H/38rSFcwen1eBal6d2v5XFZCIiXfmyePaafg3uGgBi+Ulw0jTT82hZE6d1a4C8QEmWE4qISId1a3Zy520vpO40NWc8Lhh2GZ+/j8Jg950IRUSyxXfFs7VJaPrR/vtvPhiCukY4eB6dtbB9N1vuinJY6VlZTikiIh2+9617Dtwx6XcSBMivbGFWyU4unnBsFlKJiKTnu+KZ5EogVSjX1wdo3ByENW/DntrU9A3PQmsbrNoAzTFGts5mZGRSbjOLiAxT1lo2rNt14IFIOEMBbRmzcDdFjuG8cQuylk9E5GA+XOc5CARIWo8vvrQIG2yCpIUNW2DjFjBOpznQlmOOm5vLsCIiw14g6JBMuLB5B4ypyHhc1YJdJPdOxfQwOi0iMtj8N/IcnAlOOUvrqllXPAJLp07W0uXiQWOg5sg52c8oIiJAajfYE06eiYknYNsuaGnLeOye9aXkh7SzlYjklu+KZ2MMpuyXLK2fSGswTPy4QjJdu20dQ8X4N7KaT0REurrqC2dQFrSpEY1tu7Bpe23Da7dNY9vDjVnPJyLSme+KZwATmkVx4dmYNo/wE41keoPPAt+46+/ZjCYiIgcpLs7jk58/g3A0CGMq6Vw7ByJJZl64ljN/8QSn/Psydrz6d+rqMq/IISIy2HxZPAOcN24RecvaemyhceH1q5p49UWtGSoikkuL3jcPJxKGgrz9c5rLp9Zx2o+fYfKZWwhGPApHtXL41xu475lf5jitiAxnvi2eZ5WM5ZTSWYCBwnyYMAbGj4aCvK4Hxi23XPOXnGQUEZGUvIIoX7jpU10eCxUm2PtmGbbTSqPBfNg3bimedbOcUEQkxbfFM8CnPnQ+zphqmFkDo0bC6AqYNQVqxkIotdCIsbDl1a05TioiIictmU9hSf7++zuXV/DyDTN5/oeH47mdJuCFksS8phwkFBHx5VJ1B9Q1xjFVI7rs9urmGxpmVuLmVxJd30jBo2sZO3V07kKKiAgA8XiSWFuiy2OOZzhq9F7mNbSSqEqwOl6AdUJEnMIcpRSR4c7XxfOtNz6G59n9C+63lRu2nRwBAzZoaJhcTmj+UXz7lLNznFRERJ74x5skEgemY5QUt/GT6+6npLiN/Lwkcc9wdLSe1aGP45hADpOKyHDm62kbK5a+jmcsrSODNEyMUDsjSiBmwKaKaRsyJEaEeC5fV26LiOTaM0+uwgKx0gANEyKULIzxxtYK9uzNx/Mg7FjCxuMwnsp1VBEZxnw98ty0fTd1Z8/EBgwEDKFmS6gZMOCGoWWsxQsb7l71Jp9bdEyu44qIDGv1dS001ERI5gdwXMv6p0v53mPHEwy6FBfE+Mpnn2buEdux7ptY24oxeb2/qIjIAPPtyHPCdWk6dSo2lCqcAUzHhzUEYlD4tgELIce3fw0iIoeMhrBHMj8ADhSvbSPQajEW3ESA2rp8rvmfk9ixqwAw+PjHl4gMcb7tfb722wdpKQzvn+98MIMBD/JaHS6efXiW04mISGf3372clbv2QsAQbPJwXNttgyvXdbjv4WmY8DEYE8lJThERXxbP2/c18M9X1/V6nLEwIWi45PAjs5BKRETSsdby2xuXQtwDawnEPPC6H5dMBtiytQxT8t3shxQRaefL4nnjjto+TcUwHpifvohJbMpCKhERSSceT9JQ30pkXxIsBFrTVM4AWJ55fgwP3b8tq/lERDrz5QWDE6pKiSf7tvtUcmcbL939aRZeeO/+LWFFZPhafMY3SZogJfkx7vnrtbmOMyyEw0EKC6M0NLSSvzVOqMXtNmUjxQCGH33vPo45YRrFnTZUEZHhadpPf0Cy/fP1n/liVs7py5HnMSNKeM+M8eDZng+04FWMoG5XPbi9T/MQEf/6+Oe+zYJLv0/thCIax+expaKUoy+8jtffXJvraL5njOGSK44HIG9fkkBb7333E0vfzEIyERmqHnxrJTWdCmeAmp/+gGN//ctBP7cvi2eA6z9xDuWFbqqAthk6YgNN76nhsIVx0FavIsPail0R3DwntTpPwIBjiJWG+ORXbs91tGFhyfkLmHPkOIAMo84pFovnWVqa49kJJiJD0qcefiDt49tbBn/vjgEpno0xZxpjVhlj1hpjvprm+Ygx5s/tzz9vjJk4EOftSTgY4IJPvMqJ81/KfJAxxEaGsUURCM0a7EgiMkR9+Rs/IFEUAOegsi1gaBsZzk2oQTbU+m1jDGPHjejTsY5jWLBo8mDGEZFD2Hl/vGVQX7/fxbMxJgD8DDgLmAX8izHm4Er0CqDWWjsF+BFwXX/P25s9sQ0kvH1UHlXX4zCGsZb1LZ/GGH/+gBSR3m3aWJfxOevDXaCHYr/teZbHHnmN+BFJrJN52obBMHZ8OZMmVw5mHBE5hK2vz9ynD4SBuGBwIbDWWrsewBhzG7AEWNnpmCXANe2f3wH8nzHGWJtpPkX/NSV24RKntaUILBkLaGvBCdfw52f/l+VP7SYYDLDwpGqOn3UGZZFxgxVPRIaQH3/3k5zz9dvwIgd1FJ4l3JBM/0WHtiHXb7uuS+25zSRrPEKv9rxz4NTZo/jZiw/y7GNrcFs8Jh9VxXnHL2DeiBpd+C0iPLDkg4P6+gNRPFcDmzvd3wK8J9Mx1tqkMaYeGAHsGYDzd9Mcj/OHV1fw5OYjsLt6aKK1BGpb+Nl3fkvtziq8RAhjLC/etZWl51/Dv37odOaPuGgwIorIEDK6uprCza00TMpv37zOgGcxrmVM0JfXQwypfttal20Nt/Nfpz3L9h0F/Mku6PH4h595DefB1A6xuLD7nw088/fVHHHleK6fdwlBx4dvF4hIF2EMcdL/Ll9dXT2o5x6IOc/pfs0/uDV9OQZjzMeNMcuMMct27979rsLUtrZy1p9u5g+vNLBx/Wj2vFnZfZdBm7qIMNjoUvxmPfu2VOHFA2AN1nPwEgFW3jGepatupS6u9URFhoPHH/kaFdvriexNEGxIkrcjzrzCVu7807dyHW0wDJl+21oPW3cVo+LfZS57+Ms35/X6Nc4+g0kajGswGEzC4KwzLH/6bR7avuIdZxCRQ89bn/kCo/MLuj2ejeXqBmLkeQvQeX7DWODgirPjmC3GmCBQAuw7+IWstTcCNwLMnz//Xb01+PNlz7OzqZlE0qFkS6pj7XYeA27IpXxDGyYYwab7HcLA9hWlrJ/+NHNHXPBuoojIIeahv38z1xGyZej027F/Qvw5QibGN/77bJJJh3R1e1VFIy0tQRqaoxAADlrK3yQMvGK575iXeW/13HccQ0QOPU9/7JM5Oe9AjDy/CEw1xkwyqavuLgbuOeiYe4DL2j8/H/jHYM2be2jdGhKeS6iJNGMkByQLHTafFoFIOO1SdgZwAmCMb1fzE5Hha8j02zb2KNgW9u7LY+OmMjJdoBIKu/zx13dijE27dTeAdcBRny0ig6zfvYy1Ngl8GngIeBP4i7X2DWPMtcaYc9sPuwkYYYxZC3wB6LYs0kDJD4WA1NbbPfFCkCh3SEZN2iLbWhh9VC2Ti44bhJQiIrkzpPptUwgEaIv19EaopaggRjjscfZpq9P32SGLmQdLxs4flJgiIh0GZHtua+39wP0HPfZfnT5vA7Iy9+GSw4/iu08+TltBkp6u147uNoSaoWVMmOJ1rTj57b2xASwc+ZF1nDblYxSHqrIRW0Qkq4ZKv23yPohtuZXRVY2EQi7JZLqL/QxrN4zgS187g6mTU9cr2qBN9dceYMA9wuO4Y2eweNTswY4sIsPcgBTPQ8m/zD6cV3Zs597Vb+GFXJwE3eY9GwzGgtNsSJSEaR0R431ff4Gdr4ykNFrF4pMXcMS4KykKaR1REZHBZELTsEX/gdP4Hc49czV//tthpJu6kUgEef3NKt5aPRKAhqtaCW0MUJCMcO4J8zn18DnMKBncK+xFRMCH23M7xnD9aWfy0IcvZ8ScWI/bvLZ/AS3jCwkUukxcvJWppzRyXM0FKpxFRLLEKbgYU/kUZYHe19ZPuqmR6cAuh8QRLg3zWvn08WeqcBaRrPHdyHOH8SWltMYTYKM97jAIEAi47N1Zwujx+ygN994B79ndwNJHVtLc1Mb8RZM5bM5YLcwvItIPxinh8bvyetrTqovgNgd3qkd5pLDXY1tb4jz+2Bts21rLtOmjOfr4aQSDWgtaRN4d3xbPAHmPNBOrKIRgz12xxRDJixM04V43RXn6ibf472/chedZEkmXO297nkXHTeX/XfN+HEcFtIjIuzZqJGyp79OhNh+igRCXTTqxx+O2bNrLZz/xO+KxJG1tCfLywoysLOKnN36EwqLoQKQWkWHGd9M2OrTF4zTtjUGgt8LZYupjjB0V5qzqr1MZnZr5NdsS/M837yYWS5JIuGBTjz339BqefXL1QDdBRGTYsNYyYvxOgoG+rYYXHOdwec1JXDjh6B6Pu+7au2lsaKWtLQFAa2ucHdvquPlXj/czsYgMV74snnc3N3PyT39NYs4ErEkVyB1s+4dnLNZYvDB4tXv5yMQ/UlO4qMfXXfHy22lHl9taEzz60KsD3g4RkeHA2gTrXr6Kt96yGOPReS06a9p77ZDFhi1evqX1tDg/mH0JH518co9T5pqbY6xZtaPbUv6JhMvjj64cpNaIiN/5snj+4t/vo219G8a2b92K2V80t42w1E+1NI+3NFaniuhkTTXf+c3fen3dnqZlBBxf/lWKiAy6RMNf+Op/lrN3Xx7JZJCOWc8WS2K6S+PVbTRfGqP5khjJMQkiTwX54Xf+3uvrmv3/SfOcptmJyLvku4rPs5aXXt/SbRH9juXqovsMkX0QrIWiLYZgzOB4Dne9spm5n/hBj699xNwJpOuJo3khTn/vEQPVBBGRYWX5cw+TSKTflju0KkD0kRBuvUfBTRFCa0M4rQ47dzZy2jHf5rlnVmV83fyCCLNmj+028BEKBzj1rMMHuhkiMkz4rngGcJLd13aGjvWdDdG9hmiDOfBYx+i0Y1h45fczvm44HOQb3/0gkWiIaDREKBQgHAly2pmHs2DR5EFrj4iInzU2OVjbMdp8YOyjo88OvRWg+I68A4+1fwB8/Uu39/ja//71cykrLyA/P0ww6JCXF2ZSTSWXfPT4wWqOiPic71bbcIxh6qQKNqzYjbHp35brmMaRrsBO9PL7xNwFNdx612d46p9v0dwUY/57JjOxpmJAsouIDEdz5h1L0t1KMuqQyDNEa90uz6frq/uqanQpv7/zap57ajU7ttcxedoojpo3UcuLisi75rviGeAnF53Dkjd/i42lL5Ah88i0PXi+RxpFxXmcdc5R/c4pIiJQOeFSlpzzZX61ajyFG/uwudU7FAoFOP7kmQP8qiIyXPly2sa40lIuXDin27zn3lgsxnuHXyQiIv1iTJjDz/4cTtCSyHfI8KahiMiQ4MviGcDD9Pq2XMcKHB2fA/z7eQsGPZuIiHTV0pYkGLC0VYTalxg9IO2QRvv6cyMqe99hUERkIPm2eD5mzgSs0/MocsfUjY4i+vPnzuWis07IRjwREelk7tRqktZgQw71U/NIFAX2XzyYaRikemwpt931uSymFBHxcfG8cMY4KEr2OofZYHAj0DgLTjlW85hFRHKhrCifRdENOMbDixgaa6Ikoz28e+h5XPivx2QvoIhIO98Wz/mhCJMXh/p0bCAOhcEQ1UXFg5xKREQyWVj2HoKOC8YQ2Zsg2GYzXzwYcFh03LRsxhMRAXy62kaHgpcqgF29HucFYcaboUFfuijhtbGm8QlqY5sYGa1hcuFxBJ3woJ5TRORQYY+aTuLePWAt+Tvivay6YcgviAxqHs96PLdnDcv3baQiWszpo4+gNJw/qOcUkaHPt8Xzxh37eO3t3b2uD2qxBJs8alfvIx5PEg4Pzl9JQ2Inf974aRJeKwnbRsjk8UzwJi6a8H/kB8sG5ZwiIoeKpOvxy4eexXqAZzHJTEempuIFAoaVr29h7vxJg5In5ia46sWbWNO4g1Y3TsQJ8fPVD/G/Cz7KnNLxg3JOETk0+Hbaxor12/t0nMFgww4YizOII8+Pbf8hrW49CdsGQMK20pTYw5O7bhi0c4qIHCq27a0n6VkwEIh52ECmI1P9tGshEB68Pvv2Tc+xqmE7rW4cgJiXoMWN8x+v3Iq1WtJUZDjzbfE8orjvb61ZA5Vzd/HU3p8PShbPumxuWY7F6/o4LusanxqUc4qIHEpKCvJIeoAxhOtcYmWBHtd79oIe36u/h6TnZj6oH+7f+jIxL9Ht8fpEC2837x6Uc4rIocG3xfOimRN4JwPJk0/fxOv199OQ2DlIiTKEMb79FoiI9FlJQZSKkgIAYiOCeEFDvDiAF0hN1PAO6kIT5U1sat7D0p1vDEoeJ1PfbHt4TkSGBf/2ACbJYbOa6Ms2gxbLjpcqMQTZ3rpywKM4JsCkwoUYur4P6RBietHJA34+EZFD0bxkkLxtMUpXt5G/M0m4wcV40DQuRN2s/C69ebAxSowkz+1ZMyhZzh07n2ig+4pNFdFixuWPGJRzisihwbfF8wt7/0gytKvX0jl1wWCSfWtKcJMezcl9eHbg3wY8ZdTnKQpVEnLycAgSMnmUhas5rvLfBvxcIiKHmpcffZWXf/c0eXuSGEuXW+GWBFhL7WH5NI8OEQrFcGIhAjGHmJugNRkf8DwfGLeQuWU1RAMhgiZAfiBMUTCP647610FfmUlEhjbfrrbxRt0DrFm5oNfVNgBs0FC3uogHrl7Axg/ey0un/5kPjP8eIyITByxPQbCcy2p+x8am56mNb2FkZBLjC+Zh9PafiAj33/QYBRPD1GUY8Qg3usTKQ7SNDBGt8Yg9DQXXR3imZjVnXPQdrpl3AaeMmj1geYJOgB/Nu5Q36rewonYjIyJFnFQ1i2hAy4uKDHe+LZ49myDpOb2WzgaDjQSxxiHZ5rDy9olEyt7kb3yFK6bcNqDFrWMC1BRpRywRkYPFWmJMnN7GK+u7P9flwkHHsK21jGh+glCLh7MWwtcFueaK2znswrFU5ZUOWCZjDLNLxzG7dNyAvaaIHPp8O+xZU3QMpiTR6/bcAFhwQ+3LH8UDrLl3PC1u3aDMfxYRke5OvuhYxpbtxKTps42FeNGBsZ6k6+BGUz++DAaTNIR/H+S+LcuzlldEhi/fFs8Lyz9KJNH3uctO4kCH3VYbweLR6tYPRjQRETnIiRcewzMvTMMzBmtSl3pbUqPOzdVhbOjA8LP1DIF416U/ScCmlXuymllEhiffTtu494UNxBrzet9h0FgidUmcjn7YeJRNSRXNlZGpg5xSREQAdu1soNaUkywK0FIRItzkYh2Ilwbxwp3HeSwm6RFsOqh4NjAtOjqrmUVkePLtyPM/lq+lxxX2Sa20ES+yhBrbF8I3HoGwx/T3b6QgOJKicGUWkoqIyLIX1mMdg0la3DyH1lFh2irDBwpna8FaDJbjR2zsNiziWIdzj52f9dwiMvz4tniuKiqmL2s8x8otbZUOOB6j5+/h+P96mbJql3PGXjv4IUVEBID8/NQqFsEWD+OlCuUuLOBZoqEk55/6+v7rWSyWYNjhs188i8LCaJZTi8hw5NtpGxedeCQPvriqx/rZOuBFIIDHnEtXM+GEHZSFxvGB8ddTGBqZvbAiIsPc0cdNIxh0iIUMNmDoskWstalNWh3DN5b8g4ZYGHesh1dqCRcE+d8rP8qUaaNyll1EhhffjjwHmlxCZU7G1TYslkRhqkOePW0j44/fQdBEOLP6P1Q4i4hkWd2OWj58zk7aKoJ0m5NhDBioKGzi2Blvc/fyqTR/OIa9wPLxzy1W4SwiWeXL4vm23z/NVz/3J/aWJjIeYzA4cZg9ehNzPrCBEZEJnDv221RGdZGgiEg2vbL0dT5+xJeYVPUsI8a0dB113s9Q3xrlDxtm8PK8EsoKC/jMtLM4f/yirOcVkeHNd9M2avc1cctNT5CIuwQbM/9uYLE4LpwyZyWfmPJXosHiLKYUEREAay3Xf+RnxFpibF4bpmVLAMpt2gI67ga487uH852fvZ/TJh2Vg7QiIj4ceV6+bCPBYACAktXJjMcZDMYzgCESKMxSOhER6WzP1n3U7UotD3rv70bibAPj0v2CwfaLCAsjLtMrteOfiOSO70aeCwoi+6fLhZot8bL0x1kswdoxzwAAEWBJREFUbhSSLScO6BbcIiLSd5H8MNZLFco7t0YwY6FkZQtugUOw1cM6hrayAIE2j8TUEJde+gbjC3Rdiojkju+qxrkLawgEU82KlfT8u0HbSI9te47MRiwREUmjuLyIOSfMIhAMkByZj8UQsBBu8nBcCCQs+buSRBo8Lj1mOQUlvvuxJSKHGN/1QqFQgP/+0YcoKsnDzQ9m3mHQgBuFt3ZqO1cRkVz66h8+w4RZY2k5ehLJqOm2RpIBHMfjpAnreXKVNq8SkdzyXfEMMH3mGL73q0v3j0BnYpIQqu99IxURERk8ZZUl/HL59Zixpdhw+iGPaDjJho3lJBK+/LElIocQ3/ZCFWVFOIGem2cDkHyhLkuJREQkE2MMI0sLSUYdbJrq2WIYU9XAS89V0dDQmv2AIiLtfFs8lxREOX7OpLRbpFhjiZdacKClro09uxuynk9ERLr66BkLiY8IddskxTiWUVVN3LtuOgvG7eKl59fnJqCICD4ungEWLppIsthy8EesBFqrUsfsOjJAIBDIbVAREWH+9LGMHtVEy9QwyYghGTHEix2qahppnBzh3hUzGHdELcFQrpOKyHDmu6XqOntw4xqaqy2BcijYbFJrhwKROvDCEB9hiJcHKCsvyG1QERHhpdWbueHSu7l3xVR+sXQhHfM33jCjYS+ApTEWZv4Crc0vIrnj6+I57rpgDxTOnS9DydsFXshSMUqdsIjIUFBR2ERJfhsvvz0GMN12GXSMx3umbCOaNyI3AUVE8Pm0jVMnTibQSrfCuUPeTsMlR2qdZxGRoWDBzGkkXXhxw1i6TXwGPOtQXVGFcTToISK54+viuTw/ikmkf85gcJLw0SPnZTeUiIik1Zoo4IUNY/HSLbfRbnXth7OYSESkO18Xzze8tAzj9XzM3158IzthRESkR3c8+SrfuvukHo95ec2G7IQREcnA18VzYzxGuMlk3mUQeH3zjiwmEhGRTHbVNtEYy+vhCMOGHY1ZyyMiko6vi+fjxk8g7ULPnbSZZHbCiIhIjxYfNaWXIyxb94WzkkVEJBNfF89fOfp44qWptZ0zeWWfRp5FRIaChTPGM29ST7sHGnbVe7heL/PxREQGka+L59K8PApGRvCCpC+gDdTbWPaDiYhIN8YYLl9cQGl+C5nfNrR4tpe3FEVEBpGvi2eAFjdJ89juHa01lngJFBdEc5BKRETSWbExws0f+yuhgEv3Atpy+LjdhLQrrIjkkO+L52gwiJcH8ZKuW3R7AWgbaTmmenyuI4qISLvR5ZbygjauXvxs+yN2/5+RQJKvv39trqKJiADDoHgeXVhEqAHCDalVNzo+nCQUbDNcftTcXEcUEZF2J8502FFfyM+XLiK1UUrHakmGUNBjTLXWeRaR3PJ98XzhzNlE9hjMQYvuGwzhNociR1dui4gMFaUjz+beV6aRSHb/8eTZAE+tnZmDVCIiB/SreDbGlBtjHjHGrGn/syzDcQ8aY+qMMff253zv1Ir12/jtrc8TyHBNYDQUpKG5LZuRRERyaij327vrmvj4Tx7k5mfm4tru85pdG6GuqafVOEREBl9/R56/CjxmrZ0KPNZ+P53rgUv6ea53ZPu+Bj7x4ztoaopl3CQl4DiMr0r7c0NExK+GZL/teh4f/f6fWL62ATL02dbCvGnjshVJRCSt/hbPS4Cb2z+/GTgv3UHW2seArG4LdccTrxJPuj0e8+ULTtJV2yIy3AzJfvvFVZupbWrG9vBjaeGM8UwaVZ6tSCIiafW3eK6y1m4HaP+zsv+RBsbGnft63V3w7Pdo7pyIDDtDst/etrcBr5fNTz68WBd4i0ju9Vo8G2MeNca8nua2ZKDDGGM+boxZZoxZtnv37n691lFTqjE9tc7zuHr+ZezdtrVf5xERGWoOxX57xrhKoKd3Ai1f+8ytLH3k9Xd9DhGRgdBr8WytPdVaOzvN7W5gpzFmNED7n7v6E8Zae6O1dr61dn5FRUV/XorzjplNUV4k49bcbsBj9SstXDb1syQS8X6dS0RkKDkU++1ZE6o4auo4IMPoswW3OcF3v3EXy5dteNfnERHpr/5O27gHuKz988uAu/v5egOmMC/C9Z8+Fy/fdCmgOzZJiRXFAUOsFe68/he5Cyoikl1Dtt/+8ZXnEQoaUnPuOg98pD437Q9dd+2QiSwiw1B/i+f/AU4zxqwBTmu/jzFmvjHm1x0HGWOeBG4HFhtjthhjzujneftk3oRqojPzaB5nSUYtXsCSzIemCR7xqjDxiggArz35VjbiiIgMBUO23w4FA5w4ZxpdN0dhfx0dK01N66jd1zzYUUREMgr254uttXuBxWkeXwZ8rNP94/tznnfLMYYfnn4Wl9x1B02FnUcxDFjYdtUsJl7zMuOmj8pFPBGRrBvq/fYXLziRx19dS9Lt1GebVCHdPD5CuKGFwnxtbiUiueP7HQYnlpalb6QxuCVhElVRLvnW57MdS0RE0qgqK+pxCdFYWZCPXdWt9hcRyRrfF8+F4XDmFessXP6riygoKs5mJBER6UEomKF4NoaJh1Xx3iVask5Ecsf3xXMIh4r6CEVrDUVrDZE97J8/FwwGOPWEE3KaT0REDrDWMqV6ZNrnAo7hXz9wdJYTiYh01a85z0NdwnX5yPf/jN3lEUim5sxF90CwGdomwOk1UxhbXJLjlCIi0uG625ay8u2daZ+rKClk8VFTspxIRKQrXxfPT6xYz5Y99SQ6bdNtrCHUBudMnMk3zzgth+lERKSz3XVN3PXM68Q79dkdpo0dyc+v/iDhkK9/bInIIcDX0zZe3bCN1lii2+Mhx2F2YWWPF6WIiEh2vbl5V8b5zmWF+ZQX52c5kYhId74unkeXFxMNdx+lCAUCVJUW5SCRiIhkUlVaiOt132Ew4BjGVWiKnYgMDb4uns9aOJOA07WJjjHkRUKccHhNjlKJiEg608dVMmlUOcFA1347FAxw8clH5SiViEhXvi6eSwqi3Pj585lQVUYkFCAcDDB9XAU3ffGizEshiYhIzvzf1R9g3tSxhIIBouEgFSUFfP8T51AzekSuo4mIAD6/YBBg5vgq/nbN5eysbSTgOIwsKch1JBERyaCsMI9ffPaD1Da10hqLM6qsGMcxvX+hiEiW+L54Bnhy9Qb+608PU7enBSdoOGnhFK6/+Gwcx9cD7yIih6Rte+v53M/vZt22vWBgxthKfnLVEkaWFOY6moiIv6dtACx/eyuf+cld1O1sxrhgY5Z/PLWGi3/6p1xHExGRg7TFk7z/G79j7ba9WMDa1Coc53z9tyST3S8mFBHJNt8Xz9fe/ih4YDjwtp+xsGb1bjbtrc1hMhEROdgN9z1Lwu1eJMcSSX7/2LIcJBIR6cr3xfPWHfWknS1n4Pl1m7MdR0REerBi3baMz728ZmsWk4iIpOf74rmsNA+L7f6EhZnVldkPJCIiGY2vLMv43MSq8iwmERFJz/fF8+fOPR5M17Fna6CsMp/Z1aNylEpERNL5zHnHpX230DGGK885Out5REQO5vvi+dipE6kYU4jt9FE5ppA7v3xJrqOJiMhBSgujLJo5vstj0XCQX3/hAvKj4RylEhE5wNdL1Xmex1nX3kRLfXz/BYMW2LunmaCjTVJERIaaT/zkDl5a3XVuc1s8SVsimaNEIiJd+Xrk+c5lr9PSEMd0mvJsADdh+fFDT+Usl4iIdLevoaVb4dzhO396NMtpRETS83Xx/NyaTaS7VtB48Mo6XbUtIjKUvLRmS8bndtY2ZTGJiEhmvi6eJ1WVk+7KE2tgXGVp9gOJiEhGU8eOzPhcYV4ki0lERDLzdfH8sRMX4gRNl6XqLBYMfOHs43OYTEREDjaxqpxRZUVpn9NKGyIyVPi6eI6GgvzuSxdTUBZJbfMKhAuDXHflexk/IvNaoiIikhu3fe3DTKg60D87xnDZafO44IQjcphKROQAX6+2ATBn3Cie+u5V7KhvJOl6jC0vyXUkERHJoDg/yt+uuZyGljZ21zUzaVQZjuPrcR4ROcT4vnjuMKok/VuBIiIy9BTnRynOj+Y6hohIN/p1XkRERESkj1Q8i4iIiIj0kYpnEREREZE+UvEsIiIiItJHKp5FRERERPpIxbOIiIiISB+peBYRERER6SMVzyIiIiIifaTiWURERESkj1Q8i4iIiIj0kYpnEREREZE+UvEsIiIiItJHKp5FRERERPpIxbOIiIiISB+peBYRERER6SMVzyIiIiIifaTiWURERESkj1Q8i4iIiIj0kYpnEREREZE+MtbaXGdIyxizG3g71zkOMhLYk+sQA0xtOjT4rU1+aw90b9MEa21FrsLkgvrtrPBbe0BtOhT4rT3Qjz57yBbPQ5ExZpm1dn6ucwwktenQ4Lc2+a094M82+YHfvi9+aw+oTYcCv7UH+tcmTdsQEREREekjFc8iIiIiIn2k4vmduTHXAQaB2nRo8Fub/NYe8Geb/MBv3xe/tQfUpkOB39oD/WiT5jyLiIiIiPSRRp5FRERERPpIxXMaxpgzjTGrjDFrjTFfTfN8xBjz5/bnnzfGTMx+ynemD236gjFmpTHmVWPMY8aYCbnI+U701qZOx51vjLHGmCF9pXBf2mOMubD9+/SGMeZP2c74TvXh3914Y8xSY8zy9n97Z+ciZ18ZY35jjNlljHk9w/PGGPPT9va+aoyZm+2Mw5H6bPXZuaA+exj32dZa3TrdgACwDqgBwsAKYNZBx3wK+GX75xcDf8517gFo08lAfvvnV/qhTe3HFQFPAM8B83Odu5/fo6nAcqCs/X5lrnMPQJtuBK5s/3wWsDHXuXtp0wnAXOD1DM+fDTwAGGAR8HyuM/v9pj5bffZQbY/67NzfBqvP1shzdwuBtdba9dbaOHAbsOSgY5YAN7d/fgew2Bhjspjxneq1Tdbapdbalva7zwFjs5zxnerL9wngW8D3gLZshnsX+tKefwN+Zq2tBbDW7spyxneqL22yQHH75yXAtizme8estU8A+3o4ZAlwi015Dig1xozOTrphS322+uxcUJ89jPtsFc/dVQObO93f0v5Y2mOstUmgHhiRlXTvTl/a1NkVpH4TG8p6bZMx5ihgnLX23mwGe5f68j2aBkwzxjxtjHnOGHNm1tK9O31p0zXAh40xW4D7gauzE23QvNP/16T/1Gerz84F9dnDuM8ODlqcQ1e60YiDlyTpyzFDSZ/zGmM+DMwHThzURP3XY5uMMQ7wI+DybAXqp758j4Kk3gY8idQo05PGmNnW2rpBzvZu9aVN/wL8zlr7A2PM0cDv29vkDX68QXGo9Q1+oD5bfXYuqM8exn22Rp672wKM63R/LN3flth/jDEmSOqti57eFsi1vrQJY8ypwH8C51prY1nK9m711qYiYDbwuDFmI6m5TPcM4QtQ+vrv7m5rbcJauwFYRapjHqr60qYrgL8AWGufBaLAyKykGxx9+n9NBpT6bPXZuaA+exj32Sqeu3sRmGqMmWSMCZO6uOSeg465B7is/fPzgX/Y9pnnQ1SvbWp/u+wGUp3wUJ+XBb20yVpbb60daa2daK2dSGpO4LnW2mW5idurvvy7u4vURUIYY0aSektwfVZTvjN9adMmYDGAMWYmqY54d1ZTDqx7gEvbr+BeBNRba7fnOpTPqc9Wn50L6rOHc5+d6yshh+KN1NWXq0lddfqf7Y9dS+p/ZEj9Y7kdWAu8ANTkOvMAtOlRYCfwSvvtnlxn7m+bDjr2cYbwldt9/B4Z4IfASuA14OJcZx6ANs0CniZ1VfcrwOm5ztxLe24FtgMJUiMWVwCfBD7Z6Xv0s/b2vjbU/8355aY+W332UGyP+uzc3warz9YOgyIiIiIifaRpGyIiIiIifaTiWURERESkj1Q8i4iIiIj0kYpnEREREZE+UvEsIiIiItJHKp5FRERERPpIxbOIiIiISB+peBYRERER6aP/DzD+XkoLO8DVAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 864x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, ax = plt.subplots(1, 2, figsize=(12, 8))\n",
    "ax[0].scatter(p_embedding[:, 0], \n",
    "              p_embedding[:, 1],\n",
    "              c=df['tag'])\n",
    "ax[0].set_title(\"PCA Embedding with Base Truth\")\n",
    "ax[1].scatter(p_embedding[:, 0], \n",
    "              p_embedding[:, 1],\n",
    "              c=base_predictions)\n",
    "ax[1].set_title(\"PCA Embedding with Spectral Clustering\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "tsne = TSNE(n_components=2, perplexity=20.0, early_exaggeration=12.0, \n",
    "            learning_rate=300.0, n_iter=1000, n_iter_without_progress=300,\n",
    "            verbose=0)\n",
    "t_embedding = tsne.fit_transform(dense)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, ax = plt.subplots(1, 2, figsize=(12, 8))\n",
    "ax[0].scatter(t_embedding[:, 0], \n",
    "              t_embedding[:, 1], \n",
    "              c=df['tag'])\n",
    "ax[0].set_title(\"T-SNE Embedding with Base Truth\")\n",
    "ax[1].scatter(t_embedding[:, 0], \n",
    "              t_embedding[:, 1], \n",
    "              c=base_predictions)\n",
    "ax[1].set_title(\"T-SNE Embedding with Spectral Clustering\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "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>tag</th>\n",
       "      <th>text</th>\n",
       "      <th>title</th>\n",
       "      <th>predictions</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6</td>\n",
       "      <td>U.S. To Lift Tariffs On Steel And Aluminum Fro...</td>\n",
       "      <td>U.S. To Lift Tariffs On Steel And Aluminum Fro...</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6</td>\n",
       "      <td>Rep. Tulsi Gabbard said she looks forward to a...</td>\n",
       "      <td>Rep. Tulsi Gabbard warns war with Iran would m...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>6</td>\n",
       "      <td>Get YouTube without the ads.</td>\n",
       "      <td>Brexit talks between Labour and the Conservati...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>6</td>\n",
       "      <td>LONDON/OSLO (Reuters) - Iran’s elite Revolutio...</td>\n",
       "      <td>Exclusive: Insurer says Iran's Guards likely t...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6</td>\n",
       "      <td>Thousands of marriage-equality advocates celeb...</td>\n",
       "      <td>Taiwan becomes first in Asia to legalize same-...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   tag                                               text  \\\n",
       "0    6  U.S. To Lift Tariffs On Steel And Aluminum Fro...   \n",
       "1    6  Rep. Tulsi Gabbard said she looks forward to a...   \n",
       "2    6                       Get YouTube without the ads.   \n",
       "3    6  LONDON/OSLO (Reuters) - Iran’s elite Revolutio...   \n",
       "4    6  Thousands of marriage-equality advocates celeb...   \n",
       "\n",
       "                                               title  predictions  \n",
       "0  U.S. To Lift Tariffs On Steel And Aluminum Fro...            5  \n",
       "1  Rep. Tulsi Gabbard warns war with Iran would m...            1  \n",
       "2  Brexit talks between Labour and the Conservati...            3  \n",
       "3  Exclusive: Insurer says Iran's Guards likely t...            1  \n",
       "4  Taiwan becomes first in Asia to legalize same-...            1  "
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['predictions'] = base_predictions\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### A Last Attempt to Quantitatively Evaluate Our Clustering\n",
    "\n",
    "Since we have base truths, we can do an external evaluation by looking at the variance within groups.  If our clustering alogithm has done well there will be low variance (ideally 0 because they are all the same class), otherwise there will be high variance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "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>tag</th>\n",
       "      <th>predictions</th>\n",
       "      <th>labels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>3.653209</td>\n",
       "      <td>BUSINESS</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>1.198551</td>\n",
       "      <td>ENTERTAINMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>0.455694</td>\n",
       "      <td>HEALTH</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>4.294203</td>\n",
       "      <td>SCIENCE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>0.754658</td>\n",
       "      <td>SPORTS</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>5</td>\n",
       "      <td>2.424224</td>\n",
       "      <td>TECHNOLOGY</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>6</td>\n",
       "      <td>2.030021</td>\n",
       "      <td>WORLD</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   tag  predictions         labels\n",
       "0    0     3.653209       BUSINESS\n",
       "1    1     1.198551  ENTERTAINMENT\n",
       "2    2     0.455694         HEALTH\n",
       "3    3     4.294203        SCIENCE\n",
       "4    4     0.754658         SPORTS\n",
       "5    5     2.424224     TECHNOLOGY\n",
       "6    6     2.030021          WORLD"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "error = df.groupby('tag').agg({'predictions':'var'}).reset_index()\n",
    "error['labels'] = encoder.inverse_transform(error['tag'])\n",
    "error"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Interesting we see a range of results.  Sports apepars to have been relatively consistently identified with the lowest variance.  However, Health seems to have been extremely hard to distinguish from other classes.  These findings may not be consistent run to run of course. "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}