3-UN-speeches-topic-modeling-sklearn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "d1359715",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd \n",
    "import numpy as np \n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "sns.set_theme()\n",
    "import regex as re \n",
    "from collections import Counter\n",
    "import pprint as pp\n",
    "from see import see \n",
    "import nltk\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.decomposition import NMF\n",
    "from tqdm.auto import tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "804b9b64",
   "metadata": {},
   "outputs": [],
   "source": [
    "# read pickled DF from notebook #1 (initial exploratory data visuals)\n",
    "df = pd.read_pickle('df_un_general_debates.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "99f29d19",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "6a8c470e",
   "metadata": {},
   "source": [
    "## Topic Modeling P1\n",
    "* Using sklearn modeling libraries \n",
    "\n",
    "#### GENERAL ITEMS OUTLINE: \n",
    "* Split text from documents to paragraphs within documents\n",
    "* Tf-idf vectoriser & data preparation\n",
    "* Nonnegative Matrix Factorisation (NMF)\n",
    "* Topic prevalence over time \n",
    "* NMF with bigrams\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "754617b2",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "18d31011",
   "metadata": {},
   "source": [
    "### Introducing the challenges of the UN-speeches data set \n",
    "* Due to the nature of the content, each country is allowed to only deliver a single speech each year\n",
    "    * Thus, it is quite likely that different paragraphs within the speeches are separated by paragraphs\n",
    "    * Accordingly, these paragraphs quite likely have different topics\n",
    "    \n",
    "* Also raises the issue of formatting issues with how do we define beginning, ends etc. of paragraphs "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "79eb3d97",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'33: May I first convey to our President the congratulations of the Albanian delegation on his election to the Presidency of the twenty-fifth session of the General Assembly?\\n34.\\tIn taking up the work on the agenda of the twenty- fifth session of the General Assembly, which is being held on the eve of the twenty-fifth anniversary of the coming into force of the Charter of the United Nations, the peace-loving Member States would have wished to be in a position to present on this occasion some pict'"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's recall the same example from notebook #2\n",
    "# example 1 \n",
    "df.iloc[0]['text'][0:500]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "ac23a7a8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\\ufeffOn behalf of the State of Kuwait, it\\ngives me pleasure to congratulate Mr. Han Seung-soo,\\nand his friendly country, the Republic of Korea, on his\\nelection as President of the fifty-sixth session of t'"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# e.g. #2\n",
    "df.iloc[4729][\"text\"][0:200]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4cc9130c",
   "metadata": {},
   "source": [
    "* There are newline characters being used to separate prargraphs in the first example\n",
    "* But in the 2nd example below, newline characters used to separate lines\n",
    "\n",
    "#### Thus, we need another method to discern the paragraphs \n",
    "* E.G. try splitting at stops, exclamation marks, question marks etc. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "bdbf9a8f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# \\s* = zero or more whitespace chars \n",
    "\n",
    "# define the number of paragraphs of the text, splitting on the punctuation and ignoring the whitespace after them\n",
    "df[\"paragraphs\"] = df[\"text\"].map(lambda text: re.split('[.?!]\\s*\\n', text))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "b0d6efdf",
   "metadata": {},
   "outputs": [],
   "source": [
    "# define number of paragraphs by country by mapping the len function\n",
    "df[\"number_of_paragraphs\"] = df[\"paragraphs\"].map(len)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "aa0372bc",
   "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>session</th>\n",
       "      <th>year</th>\n",
       "      <th>country</th>\n",
       "      <th>country_name</th>\n",
       "      <th>speaker</th>\n",
       "      <th>position</th>\n",
       "      <th>text</th>\n",
       "      <th>length</th>\n",
       "      <th>tokens</th>\n",
       "      <th>len_tokens</th>\n",
       "      <th>paragraphs</th>\n",
       "      <th>number_of_paragraphs</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>25</td>\n",
       "      <td>1970</td>\n",
       "      <td>ALB</td>\n",
       "      <td>Albania</td>\n",
       "      <td>Mr. NAS</td>\n",
       "      <td>NaN</td>\n",
       "      <td>33: May I first convey to our President the co...</td>\n",
       "      <td>51419</td>\n",
       "      <td>[may, first, convey, president, congratulation...</td>\n",
       "      <td>4125</td>\n",
       "      <td>[33: May I first convey to our President the c...</td>\n",
       "      <td>60</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>25</td>\n",
       "      <td>1970</td>\n",
       "      <td>ARG</td>\n",
       "      <td>Argentina</td>\n",
       "      <td>Mr. DE PABLO PARDO</td>\n",
       "      <td>NaN</td>\n",
       "      <td>177.\\t : It is a fortunate coincidence that pr...</td>\n",
       "      <td>29286</td>\n",
       "      <td>[fortunate, coincidence, precisely, time, unit...</td>\n",
       "      <td>2327</td>\n",
       "      <td>[177.\\t : It is a fortunate coincidence that p...</td>\n",
       "      <td>50</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   session  year country country_name             speaker position  \\\n",
       "0       25  1970     ALB      Albania             Mr. NAS      NaN   \n",
       "1       25  1970     ARG    Argentina  Mr. DE PABLO PARDO      NaN   \n",
       "\n",
       "                                                text  length  \\\n",
       "0  33: May I first convey to our President the co...   51419   \n",
       "1  177.\\t : It is a fortunate coincidence that pr...   29286   \n",
       "\n",
       "                                              tokens  len_tokens  \\\n",
       "0  [may, first, convey, president, congratulation...        4125   \n",
       "1  [fortunate, coincidence, precisely, time, unit...        2327   \n",
       "\n",
       "                                          paragraphs  number_of_paragraphs  \n",
       "0  [33: May I first convey to our President the c...                    60  \n",
       "1  [177.\\t : It is a fortunate coincidence that p...                    50  "
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# E.G. \n",
    "df.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "c5c8e941",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# for curiosity, does the number of paragraphs change over time?\n",
    "## as we noted before, avg words had decreased as num_countries went up -- so we somewhat expect the same\n",
    "df.groupby('year')['number_of_paragraphs'].mean().plot.bar(figsize=(14,8));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d94944d8",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "db6555a3",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "f1e68eed",
   "metadata": {},
   "source": [
    "## TOPIC MODELING PREPARATIONS\n",
    "* Start with document-term matrix\n",
    "    * matrix with elements as word frequencies (often tf-idf scaled weights)\n",
    "        * of the words (columns) in the corresponding documents (rows)\n",
    "    * The matrix is sparse, as most docs contain only a fraction of the total vocabulary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "a90dfbd3",
   "metadata": {},
   "outputs": [],
   "source": [
    "# re-construct stopwords with additional corpus-specific elements (from notebook #2) to pass into tfidf_vectoriser \n",
    "stopwords = set(nltk.corpus.stopwords.words('english'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "07d61d6d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# make manual adjustments \n",
    "additional_stopwords = {'dear','regards','also','would','must'}\n",
    "exclude_stopwords = {'against'}\n",
    "\n",
    "# user set operators to add or remove these elements from our base collection\n",
    "stopwords |= additional_stopwords\n",
    "\n",
    "## use the set difference to remove these words \n",
    "stopwords -= exclude_stopwords "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d6ae12a5",
   "metadata": {},
   "source": [
    "### TF-IDF VECTORISER\n",
    "* `min_df` = when building vocab, ignore terms with a document fequency lower than this threshold (AKA cut-off value)\n",
    "    * this value represents proportion of documents if float between (0.0, 1.0)\n",
    "    * Else, absolute counts if integer \n",
    "* `max_df` = the inverse, ignore terms with highr than this threshold (akin to: corpus specific stop-words)\n",
    "    * same rules apply for float/integer \n",
    "* `smooth_idf` = adding 1 to document term frequencies (same as notebook #2) \n",
    "* `ngram_range` = different n-grams extraction:\n",
    "    * (1, 1) = only unigrams\n",
    "    * (1, 2) = unigrams and bigrams\n",
    "    * (2, 2) = only bigrams etc. \n",
    "    \n",
    "    \n",
    "    \n",
    "#### First iteration: start with just uni-grams"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "8571c6cb",
   "metadata": {},
   "outputs": [],
   "source": [
    "## tf-idf vectoriser for speech's text: ##\n",
    "\n",
    "## start with just unigrams \n",
    "tfidf_text_vectorizer_unigrams = TfidfVectorizer(stop_words=stopwords, min_df=5, max_df=0.7, smooth_idf = True, \n",
    "                                       ngram_range=(1,1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "5a32a492",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(7507, 340767)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf_text_vectors.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "b9e989d0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 12.7 s\n"
     ]
    }
   ],
   "source": [
    "%%time \n",
    "\n",
    "tfidf_text_vectors = tfidf_text_vectorizer_unigrams.fit_transform(df['text'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "ac591dc1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(7507, 24747)"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf_text_vectors.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fd556180",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "9c087713",
   "metadata": {},
   "source": [
    "#### Repeat for the paragraphs \n",
    "* slightly more challenging, must \"flatten\" each praragraph to it's own row + keep the right reference to its respective year \n",
    "* empty paragraphs are omitted "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "5821ceb9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# if we want to keep the country variable too\n",
    "## perhaps test in subsequent iteration \n",
    "\"\"\" \n",
    "paragraph_df_by_country = pd.DataFrame([{ \"text\": paragraph, \"year\": year, \"country\": country} \n",
    "                               for paragraphs, year, country in zip(df[\"paragraphs\"], df[\"year\"], df[\"country\"]) \n",
    "                                    for paragraph in paragraphs if paragraph])\n",
    "                        \n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "eafa2789",
   "metadata": {},
   "outputs": [],
   "source": [
    "paragraph_df = pd.DataFrame([{ \"text\": paragraph, \"year\": year } \n",
    "                               for paragraphs, year in zip(df[\"paragraphs\"], df[\"year\"]) \n",
    "                                    for paragraph in paragraphs if paragraph])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "f076e057",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "287346"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['number_of_paragraphs'].sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "1b3134d4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(282210, 2)"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# E.G. can see the length of this DF is roughly equivalent to number_of_paragraphs total (less a few for removing empty ones)\n",
    "paragraph_df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "652e51f0",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "5113ebdf",
   "metadata": {},
   "outputs": [],
   "source": [
    "# apply tf-idf vectoriser to paragraphs; same min and max params\n",
    "tfidf_para_vectorizer = TfidfVectorizer(stop_words=stopwords, min_df=5, max_df=0.7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "9d4101f1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 12.5 s\n"
     ]
    }
   ],
   "source": [
    "%%time \n",
    "\n",
    "tfidf_para_vectors = tfidf_para_vectorizer.fit_transform(paragraph_df[\"text\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "827c22bf",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(282210, 25324)"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf_para_vectors.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cb1c24f7",
   "metadata": {},
   "source": [
    "* NOTE: the number of columns has changed as min_df and max_df have had their scaling effect in selecting features changed slightly as the number of documents is different for paragraphs vs. text "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7bee5c22",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "af90b816",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "71ea85ee",
   "metadata": {},
   "source": [
    "## \n",
    "\n",
    "## Nonnegative Matrix Factorisation (NMF)\n",
    "* the factorisation of the document-term matrix (DTM) \n",
    "    * DTM has only positive elements (can't have a negative count for a word obviously) \n",
    "* Conventional notation:\n",
    "    * original matrix = V\n",
    "    * Factors: W and H\n",
    "        * such that: V $\\approx$ W * H\n",
    "* Matrix W has the same number of rows as DTM and maps documents to topics\n",
    "* H has the same number of columns as features (shows how topics are part of features)\n",
    "* The number of topics (columns of W and the rows of H) can be chosen arbitarily to start with\n",
    "    * The smaller this number, the less exact the factorisation\n",
    "    * The larger the number, the more the factorisation is exact + computationally expensive\n",
    "        * In reality, an approximate factorisation is sufficient "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "29fdcc9d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# start with an arbitrary 10 topics (can be tuned later)\n",
    "nmf_text_model = NMF(n_components=10, random_state=20) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "0976bc9d",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\jason\\anaconda3\\lib\\site-packages\\sklearn\\decomposition\\_nmf.py:312: FutureWarning: The 'init' value, when 'init=None' and n_components is less than n_samples and n_features, will be changed from 'nndsvd' to 'nndsvda' in 1.1 (renaming of 0.26).\n",
      "  warnings.warn((\"The 'init' value, when 'init=None' and \"\n",
      "C:\\Users\\jason\\anaconda3\\lib\\site-packages\\sklearn\\decomposition\\_nmf.py:1090: ConvergenceWarning: Maximum number of iterations 200 reached. Increase it to improve convergence.\n",
      "  warnings.warn(\"Maximum number of iterations %d reached. Increase it to\"\n"
     ]
    }
   ],
   "source": [
    "# specify (W) matrix by applying nmf model to tfidf of text vectors \n",
    "W_text_matrix = nmf_text_model.fit_transform(tfidf_text_vectors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "5d036d60",
   "metadata": {},
   "outputs": [],
   "source": [
    "# specify (H) using the components_ function form the NMF class\n",
    "H_text_matrix = nmf_text_model.components_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5a16c54c",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "c5f673d6",
   "metadata": {},
   "source": [
    "### Specify a function for outputing a clean summary\n",
    "* Consider matrix (H) and find the largest values in each topic (row) \n",
    "    * Then, use these as the lookup index \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "c1255698",
   "metadata": {},
   "outputs": [],
   "source": [
    "def display_topics(model, features, top_words=5):\n",
    "    for topic, word_vector in enumerate(model.components_):\n",
    "\n",
    "        #calculate the sum of the vector so we can later get the % contribution of individual words \n",
    "        total = word_vector.sum()\n",
    "\n",
    "        # invert sort order (put index of largest value first)\n",
    "        largest = word_vector.argsort()[::-1] \n",
    "\n",
    "        # print header \n",
    "        print(\"\\nTopic %02d\" % topic)\n",
    "\n",
    "        # loop through top words (# specified number of {})\n",
    "        for i in range(0, top_words):\n",
    "            print(\"  %s (%2.2f)\" % (features[largest[i]],\n",
    "                  word_vector[largest[i]]*100.0/total)) # print contribution/total as %"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "id": "299a8128",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Topic 00\n",
      "  co (0.79)\n",
      "  operation (0.65)\n",
      "  disarmament (0.33)\n",
      "  nuclear (0.26)\n",
      "  relations (0.25)\n",
      "  negotiations (0.23)\n",
      "  problem (0.23)\n",
      "  south (0.21)\n",
      "\n",
      "Topic 01\n",
      "  terrorism (0.39)\n",
      "  sustainable (0.35)\n",
      "  challenges (0.34)\n",
      "  climate (0.33)\n",
      "  millennium (0.32)\n",
      "  goals (0.30)\n",
      "  2015 (0.30)\n",
      "  change (0.29)\n",
      "\n",
      "Topic 02\n",
      "  africa (1.14)\n",
      "  african (0.81)\n",
      "  south (0.62)\n",
      "  namibia (0.35)\n",
      "  delegation (0.30)\n",
      "  apartheid (0.28)\n",
      "  oau (0.27)\n",
      "  angola (0.27)\n",
      "\n",
      "Topic 03\n",
      "  arab (1.02)\n",
      "  israel (0.87)\n",
      "  palestinian (0.60)\n",
      "  israeli (0.54)\n",
      "  lebanon (0.54)\n",
      "  iraq (0.46)\n",
      "  resolutions (0.40)\n",
      "  occupation (0.32)\n",
      "\n",
      "Topic 04\n",
      "  american (0.29)\n",
      "  america (0.28)\n",
      "  latin (0.27)\n",
      "  democracy (0.19)\n",
      "  panama (0.18)\n",
      "  bolivia (0.18)\n",
      "  central (0.16)\n",
      "  guatemala (0.15)\n",
      "\n",
      "Topic 05\n",
      "  pacific (1.50)\n",
      "  islands (1.17)\n",
      "  solomon (0.82)\n",
      "  island (0.78)\n",
      "  zealand (0.69)\n",
      "  fiji (0.69)\n",
      "  nuclear (0.53)\n",
      "  small (0.52)\n",
      "\n",
      "Topic 06\n",
      "  soviet (0.81)\n",
      "  republic (0.77)\n",
      "  nuclear (0.69)\n",
      "  socialist (0.62)\n",
      "  viet (0.62)\n",
      "  nam (0.58)\n",
      "  korea (0.52)\n",
      "  struggle (0.45)\n",
      "\n",
      "Topic 07\n",
      "  guinea (4.18)\n",
      "  equatorial (1.72)\n",
      "  bissau (1.50)\n",
      "  papua (1.44)\n",
      "  republic (0.58)\n",
      "  portugal (0.54)\n",
      "  portuguese (0.43)\n",
      "  delegation (0.39)\n",
      "\n",
      "Topic 08\n",
      "  european (0.42)\n",
      "  cooperation (0.40)\n",
      "  europe (0.30)\n",
      "  bosnia (0.25)\n",
      "  nuclear (0.25)\n",
      "  regional (0.22)\n",
      "  herzegovina (0.22)\n",
      "  reform (0.22)\n",
      "\n",
      "Topic 09\n",
      "  caribbean (0.99)\n",
      "  small (0.66)\n",
      "  bahamas (0.65)\n",
      "  saint (0.63)\n",
      "  barbados (0.62)\n",
      "  haiti (0.56)\n",
      "  tobago (0.52)\n",
      "  trinidad (0.50)\n"
     ]
    }
   ],
   "source": [
    "# pass in nmf model to output function \n",
    "## + the tf-idf vectoriser (from text) to get word feature names (aka the word itself) \n",
    "display_topics(nmf_text_model, tfidf_text_vectorizer_unigrams.get_feature_names(),\n",
    "              top_words=8)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1c7de3e1",
   "metadata": {},
   "source": [
    "### Brief Overview of Insights:\n",
    "* Topic (0) is interesting, possibly talking about nuclear disarment\n",
    "    * the existence of south, possibly meaning south korea might mean the topic is NK nuclear disarment \n",
    "* Other topics can be quite generalised, representing regions of the world\n",
    "    * E.G topic (3) is clearly focused on the middle-east, including the Israel & Palestinian situation \n",
    "    \n",
    "    \n",
    "    \n",
    "* Important to note that the individual contributions of words is quite small, generally across the board\n",
    "    * Primarily due to the large amount of words\n",
    "* Generally, a topic is well define if the % within a topic is rapidly decreasing\n",
    "    * We can try to tune this later "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "344c52d6",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fb837daf",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "ebf8b8e4",
   "metadata": {},
   "source": [
    "### we can do a generalised-test to see how many documents are assigned to each topic \n",
    "* a measure of how \"big\" the topics are across the board\n",
    "\n",
    "* can be calculated by: using the DTM & summing the individual topic contributions over all documents\n",
    "    * normalising them with the total sum & conver to percentage \n",
    "    \n",
    "    \n",
    "#### General Indications: \n",
    "* having an even distribution (not having a small amount of massive topics) is a good indicator of good number of topics\n",
    "    * Otherwise, consider adjusting topics up or down accordingly "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "id": "b3da877f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Topic 0 = 10.68%\n",
      "Topic 1 = 15.16%\n",
      "Topic 2 = 13.35%\n",
      "Topic 3 = 9.90%\n",
      "Topic 4 = 13.34%\n",
      "Topic 5 = 5.84%\n",
      "Topic 6 = 7.75%\n",
      "Topic 7 = 4.18%\n",
      "Topic 8 = 13.31%\n",
      "Topic 9 = 6.50%\n"
     ]
    }
   ],
   "source": [
    "for i, val in enumerate(W_text_matrix.sum(axis=0) / W_text_matrix.sum() * 100.0 ):\n",
    "    print(\"Topic {} = {:.2f}%\".format(i, val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "54166220",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8b82b1fd",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "82f73174",
   "metadata": {},
   "source": [
    "## Re-create the topic models but using paragraphs instead of aggregate speeches\n",
    "* Given the UN speeches have lots of intertwined topics, it can be hard for a top-level modeling algo to find a common topic of an individual speech- particularly in long texts as they perhaps evenly cover a vast amount of topics\n",
    "    * We can account for this by breaking down the speeches into smaller entities (paragraphs) - which makes comparing paragraphs to paragraphs across speeches more cohesive \n",
    "    * It is also logical to assume that 1 paragraph = roughly 1 topic; as is the natural way to conduct a formal speech so we assume this method will be more robust \n",
    "        * We already created this paragraph DF above "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "4a3d7fc4",
   "metadata": {},
   "outputs": [],
   "source": [
    "# from: from sklearn.decomposition import NMF\n",
    "# repeat steps for text model \n",
    "nmf_para_model = NMF(n_components=10, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "2a91e8d8",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\jason\\anaconda3\\lib\\site-packages\\sklearn\\decomposition\\_nmf.py:312: FutureWarning: The 'init' value, when 'init=None' and n_components is less than n_samples and n_features, will be changed from 'nndsvd' to 'nndsvda' in 1.1 (renaming of 0.26).\n",
      "  warnings.warn((\"The 'init' value, when 'init=None' and \"\n"
     ]
    }
   ],
   "source": [
    "# define matrixes from para model + the afore-defined para vectors variable \n",
    "W_para_matrix = nmf_para_model.fit_transform(tfidf_para_vectors)\n",
    "H_para_matrix = nmf_para_model.components_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fac60b27",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "id": "763ecb21",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Topic 00\n",
      "  international (1.45)\n",
      "  world (1.33)\n",
      "  community (0.67)\n",
      "  new (0.64)\n",
      "  us (0.58)\n",
      "  peace (0.53)\n",
      "  order (0.44)\n",
      "  problems (0.43)\n",
      "\n",
      "Topic 01\n",
      "  general (2.70)\n",
      "  session (2.69)\n",
      "  assembly (2.65)\n",
      "  mr (1.90)\n",
      "  president (1.72)\n",
      "  election (1.52)\n",
      "  like (1.00)\n",
      "  delegation (0.91)\n",
      "\n",
      "Topic 02\n",
      "  countries (4.22)\n",
      "  developing (2.38)\n",
      "  economic (1.46)\n",
      "  developed (1.29)\n",
      "  trade (0.89)\n",
      "  debt (0.62)\n",
      "  least (0.46)\n",
      "  industrialized (0.46)\n",
      "\n",
      "Topic 03\n",
      "  nations (5.81)\n",
      "  united (5.70)\n",
      "  organization (1.27)\n",
      "  states (1.06)\n",
      "  charter (0.97)\n",
      "  member (0.90)\n",
      "  role (0.85)\n",
      "  secretary (0.55)\n",
      "\n",
      "Topic 04\n",
      "  nuclear (4.83)\n",
      "  weapons (3.20)\n",
      "  disarmament (1.96)\n",
      "  treaty (1.68)\n",
      "  proliferation (1.43)\n",
      "  non (1.20)\n",
      "  arms (1.19)\n",
      "  states (0.91)\n",
      "\n",
      "Topic 05\n",
      "  rights (6.46)\n",
      "  human (6.13)\n",
      "  respect (1.15)\n",
      "  fundamental (0.85)\n",
      "  universal (0.81)\n",
      "  protection (0.79)\n",
      "  democracy (0.66)\n",
      "  declaration (0.64)\n",
      "\n",
      "Topic 06\n",
      "  africa (3.63)\n",
      "  south (3.17)\n",
      "  african (1.62)\n",
      "  namibia (1.33)\n",
      "  apartheid (1.15)\n",
      "  regime (0.83)\n",
      "  people (0.83)\n",
      "  southern (0.79)\n",
      "\n",
      "Topic 07\n",
      "  security (5.87)\n",
      "  council (5.55)\n",
      "  permanent (1.41)\n",
      "  reform (1.37)\n",
      "  peace (1.37)\n",
      "  resolution (0.99)\n",
      "  members (0.87)\n",
      "  membership (0.73)\n",
      "\n",
      "Topic 08\n",
      "  people (1.32)\n",
      "  peace (1.27)\n",
      "  east (1.24)\n",
      "  middle (1.13)\n",
      "  palestinian (1.12)\n",
      "  israel (0.97)\n",
      "  arab (0.81)\n",
      "  region (0.69)\n",
      "\n",
      "Topic 09\n",
      "  development (4.40)\n",
      "  sustainable (1.17)\n",
      "  economic (1.11)\n",
      "  social (0.99)\n",
      "  goals (0.91)\n",
      "  agenda (0.73)\n",
      "  global (0.70)\n",
      "  millennium (0.70)\n"
     ]
    }
   ],
   "source": [
    "# display topics for paragraph model, set top words = 8 \n",
    "# change to para vectorizer \n",
    "display_topics(nmf_para_model, tfidf_para_vectorizer.get_feature_names(),\n",
    "              top_words=8)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "75fd2972",
   "metadata": {},
   "source": [
    "* We see some sharper dropping in topics E.G. topic 5 seems to resolutely be about human rights \n",
    "* Most regions have dropped out, except in those prevalent due to extra-ordinary events like israel/palestine/middle-east"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "id": "020e1820",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Topic 0 = 13.06%\n",
      "Topic 1 = 10.13%\n",
      "Topic 2 = 10.24%\n",
      "Topic 3 = 14.04%\n",
      "Topic 4 = 6.66%\n",
      "Topic 5 = 7.01%\n",
      "Topic 6 = 8.69%\n",
      "Topic 7 = 8.25%\n",
      "Topic 8 = 11.31%\n",
      "Topic 9 = 10.61%\n"
     ]
    }
   ],
   "source": [
    "# visualise what % of documents could be assigned mainly to each topic \n",
    "for i, val in enumerate(W_para_matrix.sum(axis=0) / W_para_matrix.sum() * 100.0 ):\n",
    "    print(\"Topic {} = {:.2f}%\".format(i, val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "01467aa0",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "208a1078",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f455d61a",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "b24a8b41",
   "metadata": {},
   "source": [
    "## Key topic distributions over time\n",
    "* make use of sklearns topic model's `transform` method - which calculates the topic distribution of existing documents, keeping the already fitted topic model \n",
    "    * rather than re-train the model on each year, for example \n",
    "    \n",
    "* filter to see the most important topics over time (rather than all of them and clog up the visuals)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "61d2faef",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>year</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>33: May I first convey to our President the co...</td>\n",
       "      <td>1970</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                text  year\n",
       "0  33: May I first convey to our President the co...  1970"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "paragraph_df.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "dcc39c68",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "739fb02e58a6438e81d374e9d8d19714",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/46 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# initialise list to append to\n",
    "year_data = []\n",
    "\n",
    "# specify unique years value (recall it's an included column in the paragraph-df)\n",
    "for year in tqdm(np.unique(paragraph_df[\"year\"])):\n",
    "    \n",
    "    # use existing fited nmf model \n",
    "    W_year = nmf_para_model.transform(tfidf_para_vectors[np.array(paragraph_df[\"year\"] == year)])\n",
    "    \n",
    "    # get contribution of topic \n",
    "    year_data.append([year] + list(W_year.sum(axis=0)/W_year.sum()*100.0))\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "e14252cb",
   "metadata": {},
   "outputs": [],
   "source": [
    "# filter the list of topics and highlight their most important words: \n",
    "\n",
    "## initialise list \n",
    "topic_names = []\n",
    "\n",
    "# get feature names (topics)\n",
    "voc = tfidf_para_vectorizer.get_feature_names()\n",
    "\n",
    "# loop through what is equivalent to the (H) matrix \n",
    "for topic in nmf_para_model.components_:\n",
    "    important = topic.argsort()\n",
    "    top_word = voc[important[-1]] + \" \" + voc[important[-2]]\n",
    "    topic_names.append(\"Topic \" + top_word)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "8338e931",
   "metadata": {},
   "outputs": [],
   "source": [
    "## initialise DF with index as year \n",
    "df_year = pd.DataFrame(year_data, columns=[\"year\"] + topic_names).set_index(\"year\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "41512beb",
   "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>Topic international world</th>\n",
       "      <th>Topic general session</th>\n",
       "      <th>Topic countries developing</th>\n",
       "      <th>Topic nations united</th>\n",
       "      <th>Topic nuclear weapons</th>\n",
       "      <th>Topic rights human</th>\n",
       "      <th>Topic africa south</th>\n",
       "      <th>Topic security council</th>\n",
       "      <th>Topic people peace</th>\n",
       "      <th>Topic development sustainable</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>year</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1970</th>\n",
       "      <td>14.086841</td>\n",
       "      <td>9.649841</td>\n",
       "      <td>8.795623</td>\n",
       "      <td>19.898217</td>\n",
       "      <td>6.867280</td>\n",
       "      <td>6.173577</td>\n",
       "      <td>8.715151</td>\n",
       "      <td>6.623932</td>\n",
       "      <td>11.955804</td>\n",
       "      <td>7.233734</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1971</th>\n",
       "      <td>13.589432</td>\n",
       "      <td>10.963462</td>\n",
       "      <td>10.381826</td>\n",
       "      <td>16.640036</td>\n",
       "      <td>6.216328</td>\n",
       "      <td>4.867889</td>\n",
       "      <td>10.776778</td>\n",
       "      <td>6.966861</td>\n",
       "      <td>13.020982</td>\n",
       "      <td>6.576407</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1972</th>\n",
       "      <td>15.062996</td>\n",
       "      <td>11.678255</td>\n",
       "      <td>11.473567</td>\n",
       "      <td>15.038008</td>\n",
       "      <td>7.140651</td>\n",
       "      <td>5.426078</td>\n",
       "      <td>9.230937</td>\n",
       "      <td>6.219866</td>\n",
       "      <td>12.291309</td>\n",
       "      <td>6.438333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1973</th>\n",
       "      <td>15.224113</td>\n",
       "      <td>9.964191</td>\n",
       "      <td>11.938750</td>\n",
       "      <td>14.333491</td>\n",
       "      <td>6.589634</td>\n",
       "      <td>5.859560</td>\n",
       "      <td>10.208296</td>\n",
       "      <td>6.449393</td>\n",
       "      <td>13.197995</td>\n",
       "      <td>6.234578</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1974</th>\n",
       "      <td>14.695502</td>\n",
       "      <td>12.219276</td>\n",
       "      <td>15.142427</td>\n",
       "      <td>11.906834</td>\n",
       "      <td>6.756947</td>\n",
       "      <td>4.777968</td>\n",
       "      <td>9.786046</td>\n",
       "      <td>4.814750</td>\n",
       "      <td>13.516960</td>\n",
       "      <td>6.383292</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      Topic international world  Topic general session  \\\n",
       "year                                                     \n",
       "1970                  14.086841               9.649841   \n",
       "1971                  13.589432              10.963462   \n",
       "1972                  15.062996              11.678255   \n",
       "1973                  15.224113               9.964191   \n",
       "1974                  14.695502              12.219276   \n",
       "\n",
       "      Topic countries developing  Topic nations united  Topic nuclear weapons  \\\n",
       "year                                                                            \n",
       "1970                    8.795623             19.898217               6.867280   \n",
       "1971                   10.381826             16.640036               6.216328   \n",
       "1972                   11.473567             15.038008               7.140651   \n",
       "1973                   11.938750             14.333491               6.589634   \n",
       "1974                   15.142427             11.906834               6.756947   \n",
       "\n",
       "      Topic rights human  Topic africa south  Topic security council  \\\n",
       "year                                                                   \n",
       "1970            6.173577            8.715151                6.623932   \n",
       "1971            4.867889           10.776778                6.966861   \n",
       "1972            5.426078            9.230937                6.219866   \n",
       "1973            5.859560           10.208296                6.449393   \n",
       "1974            4.777968            9.786046                4.814750   \n",
       "\n",
       "      Topic people peace  Topic development sustainable  \n",
       "year                                                     \n",
       "1970           11.955804                       7.233734  \n",
       "1971           13.020982                       6.576407  \n",
       "1972           12.291309                       6.438333  \n",
       "1973           13.197995                       6.234578  \n",
       "1974           13.516960                       6.383292  "
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# E.G. \n",
    "df_year.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2f09e206",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "0e75ee83",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA6cAAAIZCAYAAABaq0nBAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAAEAAElEQVR4nOz9WWxkS37f+34j1pRzkkzOVay5as9Dz2q1uls6tmW1rcY9bh/AkB/k+6YnwRBgw4AtWIABQQZsQC+GHwz4ScCFjwzfA9myT+vouuWW1K1W997dex5q2DWRVZzJnHNNEfdhJbPImgeSSbL+nw3uHEhmBskscv3WP+IfylprEUIIIYQQQgghhkgPewBCCCGEEEIIIYSEUyGEEEIIIYQQQyfhVAghhBBCCCHE0Ek4FUIIIYQQQggxdBJOhRBCCCGEEEIMnYRTIYQQQgghhBBDJ+FUCCGEEEIIIcTQucMewN02NtoYc3C3Xq3VSqyttYY9DCF2hbyexVEir2dx1MhrWhwl8noWAForRkeLD3z/gQunxtgDHU6BAz8+IZ6EvJ7FUSKvZ3HUyGtaHCXyehaPItN6hRBCCCGEEEIMnYRTIYQQQgghhBBDJ+FUCCGEEEIIIcTQSTgVQgghhBBCCDF0Ek6FEEIIIYQQQgydhFMhhBBCCCGEEEMn4VQIIYQQQgghxNBJOBVCCCGEEEIIMXQSToUQQgghhBBCDJ2EUyGEEEIIIYQQQyfhVAghhBBCCCHE0Ek4FUIIIYQQQggxdBJOhRBCCCGEEEIMnYRTIYQQQgghhBBDJ+FUCCGEEEIIIcTQSTgVQgghhBBCCDF0jxVOW60Wv/qrv8r8/DwAP/zhD/n2t7/NL//yL/P7v//7g4/7+OOP+c53vsPf/tt/m3/xL/4FSZLszaiFEEIIIYQQQhwpjwyn7777Lr/2a7/GtWvXAOj1evzzf/7P+ff//t/zP/7H/+CDDz7g+9//PgD/9J/+U/7lv/yX/Mmf/AnWWv7wD/9wTwcvhBBCCCGEEOJoeGQ4/cM//EN+53d+h8nJSQDee+89Tp48ydzcHK7r8u1vf5vvfve7LCws0Ov1ePPNNwH4zne+w3e/+909HbwQQgghhBBCiKPBfdQH/O7v/u6O28vLy0xMTAxuT05OsrS0dM/9ExMTLC0t7eJQh++nqw3eu3KbOE6HPRQhdoV3xZHXszgy5PUsjhp5TYujRF7P++ML4xU+P14Z9jCe2iPD6d2MMSilBrettSilHnj/k6rVSk/8OfvFNNosNLvDHoYQu6c37AEIsYvk9SyOGnlNi6NEXs/74sdrDb5wcpzxQjDsoTyVJw6n09PTrKysDG6vrKwwOTl5z/2rq6uDqcBPYm2thTH2iT9vP3ixIUoN8QEdnxBCCCGEEOL5FSU9Pl7Y4MWR4rCHcl9aq4cWI594K5k33niDq1evcv36ddI05Y//+I/5xje+wbFjxwiCgLfffhuAP/qjP+Ib3/jG04/8gHKeohoshBBCCCGEEHtNH/Ks8sSV0yAI+Nf/+l/zm7/5m4RhyDe/+U1+5Vd+BYB/+2//Lb/9279Nq9XilVde4dd//dd3fcBCCCGEEEIIIY4eZa09UHNUD/K03nfXmvzR9WV6qRn2UIQQQgghhBBih0Br/sHZ6ednWq8QQgghhBBCCLHbJJwKIYQQQgghhBg6CadCCCGEEEIIIYZOwqkQQgghhBBCiKGTcCqEEEIIIYQQYugknAohhBBCCCGEGDoJp0IIIYQQQgghhk7CqRBCCCGEEEKIoZNwKoQQQgghhBBi6CScCiGEEEIIIYQYOgmnQgghhBBCCCGGTsKpEEIIIYQQQoihk3AqhBBCCCGEEGLoJJwKIYQQQgghhBg6CadCCCGEEEIIIYZOwqkQQgghhBBCiKGTcCqEEEIIIYQQYugknAohhBBCCCGEGDoJp0IIIYQQQgghhk7CqRBCCCGEEEKIoXOHPQAh9oJDSo4QiyLCI8EB1LCHJR7IMs4GORVSt2VaFLBy7kwIIYQQ4rki4VQceC4JASE5QnIqyi4Ht0NyRNuuZ7c9lex4jNQqYjyiwZtPaHfejnbc9vq3/cHtGBcJuLsnT5c5tcicus1xtUhehYP3JVbToMymLbNJmbqtsGnL1KnQIxjiqIdDYfCJMWhivGEPRwghhBBiT0g4FUOjSSnRoaQ6lGhTokO5fz2veuSICAjxVPrAxwitR4+AHj4dm2eDCgkuqXEwSuOSorBoDBqDQ4pHgqcSKqqFT4xPjEeMfkShzlp2hNd4EGDvDrQ73+JByHUJCTDPaUVQkzKjVgZhdFxtAtC1AfN2mg1bpWybOMqggICIUVXnJAs42g4ep2d9NqlQt1vhNQuuDcqkOMP54h6DQ4pPRLD1pmKC/isjux31b8fbbsfZ+1UMZCdZrto5PjDnWWQCOVkihBBCiKNEwulzRGPwBmEsyS5Vdt2iSHBIrJtdsnV55/qTTbO05OntDJ+q0w+gWRAtqN49n9WxAW0KtGyBdarEeKTGwSoF1uKQ4qgUlzQ7iO9XUqu0mFTrqMc8VrcWUhwiPLrk2KREbLLpvylOFmWtwigHiwJUP+Sm/YCb4pKQUyEVmoOQ62rzyOfuWp8ueTo2R5ccHXJ0bH7b9Rwd8oT4h3xqq6VKkzl1mzm1yKxawlMpqdUs2Ro/NS9jUYyrdU6pec6r6/c8Qmo1DUq0TKE/PdvFIaVAj+NqkRf01TvPZqFJkU1bpkmJLjm6Nuhf5ugS0CNHD5/dDHUuMUW6FFWXIp3B9cLgvi55ergPOckCEFsnq+jjEeLTsgU28IhxMDZ7TZboMKduc869wZqt8qE5z0V7ikSqqUIIIYQ4AiScHjhZAHP7b9uvuyq5531bQdNTdyqA/o7byeDyUQfHj5Jaff/gau/cFxBRVm2KdHDVzqAWW4c2BZq2yAZVQuOTKgdtU3yVVYhKqkuJNuNq47GD5tNQisH3sUA/JD/F822fLtwhT2RdUpt9bwwOBkVqNbYfchVmW7AOmWKVAj1cfe/PxlhFlyxc3R1eI/zBzyDtn1RI0YOfQzr4+Tj9Su3+VNh8Io6ppUEgLas2AHVb4qI9RdsUqagms2qZz+uPHvl4jjKM0mRUNe95X2g9lu0YHRsQ42HReMSUVIdJ1gmIUPfJ9sYqegTZ93ZbaN26vj3UGnQ/bHZ2hM1C/74iXfy7ppADRNalQ562zbNIjQgPYxyMUtmpDmuzn44y/X/TCYFKBpXUMm1cld7330BkHRbMBHkV8g3nLb5i3+WiPc2H5hybVB/5PRVCCCGEOKgknO4Dj5gx6oyqOmOqzih1fBXjZhNQB4HTIX3oFNaHSaxDhDsISjEuLVsYBEiDvlOBBLDZOrasEpnFKIPCAsZqrNL9imEmqx0a9F1TZB0Mjkop0MUlJcRj1Y5yk2kS64JVuET4KqZAj5LqMqNWnjkoHxSOsjhE5Iju3PkE1dsIjyYFuibon1rIwq1RzrYgG5FTIWNskqe3Y4rr4z7P9sB696WxelAtTvs/3XTwU86up4OPu//7K7SY07eZYg2tLJF1uWUn+dScxiVhSq/yovrsicf+MIGKmWT9nu+3tdCiQJ0ikfF3BHaLRmFxSfBICFTIJG3y9PD1vSHzbqlVWegkz7qtcptJYuNilUJbM5iJUFA9inSZVGtP/W/6QXyVckytYC2s2BEi6/Oyusxr7kXmzRQf2vNcs8cOedVdCCGEEM8jCae7yCVhdFsI3QqkZdUZfExsHTap0LUBnX5lZvuBvrVbEXDriNui7FYwNGhlUdj+FFPbD4kJvkoGVdIS7ayu9hhTTHeDtWDIqqp5ekyptT2teh4VSkFATEB838rgdsYqQrK1lj0TZJU4dP+EgsLY/itCqW2hZGtCsoVtJxV0v6bbn7yMq7IA7NznfXr79YcES2thlVHetxeIjMeoqnNMLXFK39q179fjUgrKdCjTeeSJgq0TBB1yrNmA2GZTh1OrSVV2ssdYBrMTcoQUVY8RmkN/nSsFE2yCgrYN2DAjjKgmf1v/JS2b5yNzjk/sWTrkhzdIIYQQQognIOH0KTikjNDYUQkdU3UqtAYHq4nVbFJh0Y5zxRQwaAJCSqrDmKrv+bTV/aQUg0Aj9oZWljwhecJdn6FrrML0T3+kg5q4zqq42yulVpHabHqyGZwiUf31yi4TaoPX1af3nUp7UG0/QTBC69D2FyqqkKJaIrWwaGooBV923ucL9gNpoCSEEEKIQ0PC6WOy1pJf/x7/O1epOE20yqpIxirqlFm1o1yzx0hx8elRVF3GVIOz6uZDK05CDJtWWRWexzm5INnmQHMUTKs1ADZtkaYtSwMlIYQQQhwaEk6fgI43WLcVbjBDYh08G1Poh9BTan5X19MJIcSzGFFtRlS730BpctBA6efsO1yzxwnx7qw1tlsVcmdHtXzn7Tvrjrcq7D2yztNCCCGEELtBwuljUkoxM/NF/Bv/A8z8sIcjhBCPJWugtLyjgdIxtTiYtO08Yj3xw1gL1+0x3rcXWLBTSGldCCGEEM9CwukT0srKykohxKGzvYFSz/q0yQ/WF6f3rCfWg+tsX2NsudOOTSlyhMyqZU7pBdZslQ/MBS7ZUyTyp0UIIYQQT0GOIIQQ4jmTU3dtfwRPXfRMrOaWmSCnQr7p/ISv2Hf52J7lA3OeNsVnH6wQQgghnhsSToUQQjw1Vxlmt/ZdNSPEyuMN9QlvOJ9w1R7nffMCi4wjU36FEEII8SgSToUQQjwzpWBCbQLQtHk2bYXjapGz7k2W7RgfmAtcticwOMMdqBBCCCEOLAmnQgghdlVZdSmrLrF1WDATFFWP/835ET9n3+Eje44PzTnp8iuEEEKIe0g4FUIIsSc8lXKsP+V32YxhlOaL+gM+pz7iij3B++YCK9Qe89EsPjEBETlCAtW/JMruUxEawy07ybydJsLf069NCCGEELtPwqkQQog9pRRMqnUA6rZI05Y4pea54F5j0Y5z0ZxCYckREaiwH0Cz67l++AyI0OrBW95ENvtz9qq+hLGKJca5YWa4YWdZYwRZ8yqEEEIcfBJOhRBC7JuqalNVbULrsmCmKKsW33DeGrw/si4hPj0CQuuzRoEYlwQXYzUGhbYWhwRHpTgYPGICFRMQ0jV5InzKqsVXnPf4Cu/Rtnlu2mlu2FmpqgohhBAHmIRTIYQQ+y5QCcfUEsbCgpmkSw63HzJzRJToMKY2cR5SLb0v1QTAWlizFRq2hIvhtJrnRX0VYxWLjHNTqqpCCCHEgSPhVAghxNBoBcfU8q4/rlJQo0FNNQD6ldrJ+1ZVb9gZbtoZqaoKIYQQQybhVAghxJGXVWqzELxVVW3aEg6GM+omL+nPBlXVG2aWK/YETUpDHrUQQgjxfJFwKoQQ4rnyoKpqiE9Ftfg5511+jndZtDUum1NcsXOy9Y0QQgixDyScCiGEeK7dXVVdt1WatkBZdfgF521+3v6UBTvFZXuSq/a4TP0VQggh9oiEUyGEEKJPKRijzpiqYy2s2iptW2RU1fkl/dd83f6E6/YYl+1JbthZUpxhD1kIIYQ4MiScCiGEEPehFIxTZ1zVMRaWzBg9AmbUMmf1TULrcdUe57I9yYKdwqKHPWTxFBSGaVY5rhfp2DyLdpx1qvLzFEKIIZBwKoQQQjyCVjCl1gFIrOaWmSDF4Yy6yYv6Kh2b44o9wSVzkmVqPO72NC4JRboU6FJQ3ez61mX/ukfCJ/YM75kXZUrxLnFIOaYWOa3mOaUWyKsQa0H182hsXZZsjSXGWbLjLNkaIcFwBy2EEM8BCadCCCHEE3CVYVatAHeaKaHgJXWZ19yL1G2Jy/Yk180xHNIHhs4iXXyV3PP4idV0yNMhz7odwSfii/pDXlWX+Jl5mQ/seVL58/3EfCJOqFucVvPMqdv4KiGybrbfrRmholpUbYMeAaCpqCaf4yO0zvba3bAVFu1WWB1ngwqyR64QQuwuZa19wh3O99baWgtjDtSQBtobH7Bx879j0nDYQxFCCHHAtG3Auh3BI2VSraHVzr9l20Nn2+YICYiti1UabVM8leARkyOkoHpZeCVGKWjYIj18JtUGbZvnLfMqn9ozGJl6+lB5upxSC5xW8xxTSzjK0LEB1+0xGrbMqNrkpLpFoOJ7Ptda2KRM05aI8MgRMq42yKkIgNB6LNnxLLAyzrKtEePt95cohBA7BFrzD85O8+JIcdhDuS+tFbXag7dqk1OvQgghxC4oqpCiWgKgYQssmxodW8BTMS4x+X7oHKXOlFpFPUHRraLaVGizYcsYNN90fsKb9mN+Yl7jsj2JVPDuKNPitLrJaT3PNKuDcP+hPUfP5JhQq5xT1/F0+tDHUQpGaTKqmoP7Qutyy0zQJUABI6rJl9T7KAXGKtapsmTHuWJPcMtOIj8XIYR4MhJOhRBCiF1WUR0qqrPrj7sVlNZMFUcZ/qbzV7xpP+bH5nVu2FmezzBkqbHJaT3PKTXPuNoEYM1Wece8RIrDjFrmFXUJRz/bzKxAJYMp3ZAF0g2qNEyRBJc8Pc6ra7yiL7NuK3xoznPRnpaKqhBCPCYJp0IIIcQhU9PZVjcrZoScivg7zp9z247z4/QNbjM57OE9I4vfn94cEJFT/csd10MCQnIqokCXkupiLSwxzlvpqygMc3qRN/XHT1ShflJa2cHWQ1u61mfBTJBTMV933uYr9l0u2VN8YC6wQXXvBiOEEEeAhFMhhBDiEFIKJtTmYJubqmrx/3L/JzfMDD82r7PK2K48jyZllDrjapMxtckoDRQWC1h0/1I94O3B7wPwiQchMyAkR0RAdM963e1C6xHi0yOgZwPqlNgwVQIiTujbfNH5YFe+7qeVVxHH1Mpgn9zQ5nhBfcYr7mVu2Uk+MOe5Zo/LemEhhLgPCadCCCHEIba1zU1qFYumxqRa4/9w/4QrZo4fm9epU3nMR7IU6DGmNhlng5rapKY2GaExCIuJddikTIIziJlZPAX9kIi6FUfvvi/CG4TMNlViPBJcjM0eTVmLQ4KjDC4JHgmByqqoZVqMq42HBtlh2tonF1WnYwMWzQhV1eSXnR/Qtnk+Mmf52J6jQ37YQxVCiANDwqkQQghxBDjKMq3WiK3mthnPtk1x5vnUnuYt8ypt7nRuzKqhDWr9auhWGM2rO93oWzbPmh1hnili6+HbiJJqU1N1RmkAYPqR0+yolO6sppp7Yuudj9EYCvQYVQ0cZfbz27WvCiqkoJYwFhZNDRR8yfmAz9sPuWrn+NCc5zYTHNU1w5qUSdaYVctMq1VaFLhoTrPIOEf1axZCPB0Jp0IIIcQR4inDjFoltC6rtsYFdY0LzjU+tadxSQdTc7fCYGI1G1S5YWdpmxwKyKseI6rBrFrhpLr9wOdy9ulrOiq0gmm1BkDdFmnYMnPqNufcG6zZar+B0imSJ2yg5JBSoEueXn8P3R75rUt6dMmxakdYs6OsMbLnDZpcEibVKrNqhRmWmVJruCrrjrxhK8ywwsvuFRq2yEV7movmFA3KezomIcThIPucPgHZ51QIIcRh07UBm7bMlFqjS8CaHWWTMjEeno0oqw5jqk6F1p42DxL3F1uHZVsjr0LGVJ3IunxqT/OhOU+HPAW6FFSXAr3+9f5lP4gW6A32Xt3OWOiRo0tAke6Oj6nbEqt2lDU7whqjrNoR2hR42iqmS8y0WmVWLTOjlplkHUeZwfY6i3aCrs1RVB0m1RpF2qzaMTzSwbZKt+04F81prtgTRPhP++0U4rl32Pc5lXD6BCScCiGEOKyaNk+LIjU28VUy7OGIu1gLq4wQWZ9ptfrAac6J1XTI0yFHx+YJ8Ymti1EOyhpcYlyVkiOioLLKqUPKhq3QIyDFxSemqhqMqNbgcXvWZ82OsMpoVmG1I2xQwdynPu4TMa1W+mF0hQnW0cpirGKVURbtOKH1Kak2U2ota6L1gNxbt0XqtkJZtRhVTVKruWaPcdGe5qadkcZRQjyhwx5OZVqvEEII8Rwoqy5lusMehngApWCCTVDQtjlWzRh1yhg0jk1xVdYuKkdIQfUYpcGUWnvsandere64nVrNmh2h1Q+4DoayavMKl3F1OviYDSqDsFpUHWbVMjU20cqSWs0KY7xvLxAbj5JqMa1WeU1dRD1mpqyqNlXVzrZGsiN0bZ5ZtcRZfZOuDbhkT3LRnGaVUWR9qhBHn4RTIYQQQogDpKh6FNWtPX0ORxlqZB2Zt1gLTQo0TImQAIMmT485dZsX9FUSq1m2Nd7jBRLjUFYtZtQqr6tPHzuMPsidcL5JbDW3zARWKV5Rl3ndvci6rXDRnOaSPdWfgiyEOIoknAohhBBCCJSCCh0qqrPj/tB6LNoabZtnSq0zq1b2dByeMoPnaNuAdTNCTkX8nPMuX7HvsmCn+NSe5qqdI5FD2eeQRaroR5f8ixZCCCGEEA8UqJhp1oaSB4oqpKiWAFi3ZZq2xJiq8zf0j2jbd/mReYNL9hQSVo6+Ghv8gvMWNTa5bSeZt9PctNNsUkF+/keHhFMhhBBCCHHgjakmY6qJsbBkxvBVwt9wfsSr9hI/SD/PMuPDHqLYAx4xX9Lv86q6SIjPNXuMKbXGSZ1NfW/ZfD+ozrBgp+iRG/KIxbOQcCqEEEIIIQ4NrWBKrWcdjs0IZdXiO+6fcsmc5EfmTVmTemRYzqob/Lz+GQW6fGzP0rMBr+tP0FhW7CgtW8Al5ZRa4EV9FYAVOzqoqi7aift2nBYHl4RTIYQQQghx6CgF42qT1CoWTY3T6iannXl+Zl7iXfuSrEc9xKo0+Lp+i+N6iRU7ylvmVd7UH1PVd7Y/mmCDCbUBQGhdFswkIT4F1eUN9Qmf0x8TW2fHFOANqsgU4INN/tUKIYQQQohDy1GWabVGz/o0bIkvOR/wov2MvzZvcNmeZC/DiE/ESbXAaTWPAubtFAt2mk3Ke/q8R5VDwuf1R7ypPibB4a/SN5hSq3xD/+Sh2yYFKuGYWh7cbtgCm6ZCisuIqvPz+mcAtG2eeTvFvJ3mhp0lJNjrL0k8IQmnQgghhBDi0MupiJxap25LWBR/0/mrwXrUFWq79jw+EafUAmfUDebUIo4ytG2eFM1pPQ9AyxYGQXXeTtElv2vPf1SdULf4Bf0WFdXmkjnBhh3hi/pDPJU88WNV1J2u08YqVuwILVvAwXJS3eIFfY2OzfF/pn9HAuoBI+FUCCGEEEIcGVXVwlpYMSNUVZO/7/4/fGpO8dfmDTpPuR41C6TznFU3Od4PpC2b5yN7lq7JM6VWmFUrLNsxWjZ/zzrINVtl3k6zYKe5ZSdI8HbzS94T0yzzBf0hRdVhoV9tvGWniHd57CXa/Lz+KWf0PBu2zPfTL/Gavsh5fWNXHl8rywSbTPT39I2sw6KpManW+bJ+j78wX9qV5xG7Q1lr7bAHsd3aWgtjDtSQBtobH7Bx879j0nDYQxFCCCGEEI+QWsWKHWNCbWDQ/My8zLv2BdLHqM/4RJxW85zZEUgLXLXH6docU2qZ42oZR5n7fn5oXdbsKD0C8oRMqDVcZUitZoka8yYLq8uMYdG7/aU/tWlW+KJ+n+N6ia4NWLMjTKlVPJWSWsUy48ybbA3nyjOMXZPyuvqUL+gPUMDPzEtUaHJBX3/oFN7dElkXl4T/kv5t1hjb+yfcJ4HW/IOz07w4Uhz2UO5La0WtVnrg+yWcPgEJp0IIIYQQh0/X+jQpManWadoCf2U+x2d2jrvXhQaEnFILnFU3OKaWcJSh2Q+kPRswrVY49pBA+jAtm2PDjhDjUVFNamyiFITW41a/ac8w16veHUrfNxco0uZFfY0Qj7X+2MuqzTgbg7FvVVXn7TQNyo/1XDMs83XnJ4ypBtfMLLfsFF/QHxCoeI+/yp1i67DOCP9X+rc4KmuED3s4lWm9QgghhBDiSMuriDxb61Hhl50fcMtO8IP087Qo9iukW4HU0rQFPrTn6RmfGbXMK+oSjn624klJ9SipRQCshU3K1NMyKEVNbXJaLwDQsEWu2BNcNidYY5S9Dk13h9Ifp69RpM0X9IeDEF4gpKCWto29RN2UAc2EWudMf61t3ZYGQXXBThHh73iuPF1+Tr/DC/oaTVvkz9Mv8qL+bNCwaL85GKbUGhfUNS7a00MZg9hJKqdPQCqnQgghhBCHm7WwakcpqzYBEQaFoywNW+SqPUZoA2bUMrNqGUftzzFpajVrjNC2eQIiptUqWlk2bZkr9gRXzAnWd3kblClW+KL+gDm9OKiUFujwkr76RJXh1CpW7Sgd8ngkTKo1fJVkjYgYy7ZxMdOMqTpf0e/ikvKefQHPxrysr6D36Xv8IKH1SNH8p/RX7wnTh9Fhr5xKOH0CEk6FEEIIIY6GxCqW7ThrjNK1AbNqiVm1MvSwBNC0BTZsBZ+ESbWKVrBhK4OK6ibVp37su0PpB+YCeTq8qK/iPsV05bt1rceaHSXCp6Q6jLOO7mfqBTPJZ3aOL+gPKajeMz/XbrEW3rcX+KH5wrCH8swOeziVab1CCCGEEOK54yrLrFphlpVhD+UeZdWh3N8KpWELbJgqARFfUB/wRfcD1myVK+YEV+wJ6lQe6zGnWOWL+v1BKP1J+io5enxOf7groXRLXsUc7+85ai00KLFpyqzZMabUMl933t6159otMS6vqkt8zFk2GBn2cJ5rEk6FEEIIIYQ4oLbv2Vm3RTZNhbwK+bLzPl/mfVbtCFfMCS7bkzS5tyI1xSpf0O9zoh9Kf5y+SoEen9Mf7WoovR+loEqLqmpxktt7+lzPwlcJsXX4un6L/2r+BkelOdJhJOFUCCGEEEKIQ6Cq2lRVG4ANW6JhyxRUj6847/EV3mPZjg0qqgW6fHFbKP1J+ip5enx+H0LpYaQxzOoVztobXLEnhz2c55aEUyGEEEIIIQ6ZUdViVLWwNluPWrclSqrDV513+CrvANCz/iCUfk5/jKvS4Q76AHOUJbQeP69/xvV0lgRv2EN6Lkk4FUIIIYQQ4pBSCkZpMKoaWAtrtkLTlli3I3jEEkqfgE9MoGK+oD/kr82bwx7Oc0nCqThyUqNodDzqbQ+toZSLKedjPHf43feEEEIIIfaKUlCjQU01OMWtYQ/n0FEKQuvyuvqUTzjz2M2mDgbLV/S7nFC3seb/PezBPDUJp+LQi1NFve1Tb3vU2z7NrodFAZbtC9oDL6Wcj/thNaGUjwm8/VtzYW32S088X6JE0ep55P2UnJfKa0AIIYQ4wAKVkFiHr+u3+WPzixyO5kiWX9Bv86q+xHvmAm+kS3CogvUdEk7FoRPGmnonC6KbbZ92zwUUSlnK+ZhjtQ6uY+nGmk7o4jsGR1tSo2n3XFYbAVu/aHw3pZTLgmo5H1PKJ08VIIyFKNGEsUMYZ5fRtuthrAkTh0o+5qUTm+T2MRSL4bAWbq/n+WyxTGI0AFoZCkFKIUgo5BIKQUIxSMn7CVoPecBCCCGEAEBhOa4XOWXnuWbnhj2ch1IYvql/zIv6Ku+aF5hQm4z7h3e2oITTQyxOFI62R/qg1lroxQ71tsdm26fe9ulG2ctWa0MlHzM33kYr6EYOjY7HfKfAg85yKWUpBgmBn+JqSC30Qof1VnHwOa5jBlOBS/mEcj7OpnnEO8PnjiCa6HueUyuL76UEXkqlEKN1xHI9x9uXxnn5xCajpWgPv3NimFo9l4sLFRodn2ohYqQU0Y0csIooUdQ7Psv1/LbPsOT9dFtgTQYh1nUO7x8YIYQQ4jBylCG0Hl/TP2U+nSE5oJFJY/gl/SPO6+u8nb7MMb3IjNMe9rCeycH8Tot7GAvtnku97dPoeDQ6Hr3YRWEp5BJKuYRiLs6qgLkE/5BW5qyFTugMqqL1jk8YO0AWGiuFiIlqD4BOz6XR9dhsB0/w+Ip26NEOt3dgy4JBzk/wHIuxil7kML9WxNr7h1xHG3zPkHNTCqUI1zFolU0kTlNIU0WUOESpQ6vrUd9WOdOO5d2ro5yZbjE33pZpnkdIahTXlorMrxZxHcu5mQarjYDryzv3nXN0dgLEd1Mcx6IVxKmmGzqsN4MdrzvfzUJqMZedKJmo9nCO8AkpIYQQ4iDYao70pv6It8zrwx7OPTQpf1P/kDN6np+kr3JKLzChNoDHPy4+iCScHlBRomh0siBa7/g0Oy7GZkekvptV4iaqPaIkq95ttHyWNu9UYrz+dNViLqHUD62F4OBNHbQ2q3hutLIwutn2iZMsjPpuSrUQURhNSU0/jLZ91pu5XR6Fohu5g4psf2QEbkohSHEdQ+BbwJKmkBhN3J/CW+/4mAcE2PsxVhPGWRj+bLFMo+Px4vG6VMeOgLVGwMVbFcLYYWqkQ87Lfsb3e32kRtPqabirTb1Slly/2u46BkdngTeMNYsbeRbWily5nXKs1uFYrSNNvoQQQog9kjVH8nhTfcynnKFJ6dGftE8cUn5Z/yUn9S1+lL7BeX2NmqoPe1i7QllrD9TRzdpaC2MO1JAAsNbyf/1//4Ao7BB4aTZd000JPJPddrPLwMvWNz5JNcxuVUU73iCQbgUlhaWYT6jkY3zXECWKds+l2fUGYRWySkw+SAhcg+tYUpNV/9qhO6jC3K/KWswl+94UqBs5WRDtB9JoWxgdKUYUgpQkhWbXu+frPEpyfkIvcsj7Ka+e3KSYS4Y9JPEUwlhz6VaF1UaOfJAwN95mfrVIJ9zNc38W3zXkvGwrgEbXRyvLzFiH4+Md8r5sESCEEELshcRqFuw0/7f55rCHAoBLwq/oP+eYWuKH5nO8rC8zqpqD9ysdMH7q75GvXhjiKB9Ma0Wt9uCg/0xHT3/0R3/Ef/gP/wGAb3zjG/yzf/bP+OEPf8jv/d7vEYYh3/rWt/it3/qtZ3mKA2VmaoLN1YvEqabXn3qapPcGJ60NwbawmgXZO7c919AJHRqdrMNss+uR9qd9eltTVys9lIZu6NDsetxaz/OwbmGp0bS6Pq0d91oCLyXvZ1UYrSF6QJV1q5Nozs+qNrnB7SxsP62tNaNbQXSjdSeMeoMwmpCmimbXY7WRe6JK5GHWi1w8JyVONW9frvHC8TpTI71hD0s8JmthYa3A1aUS1ipOTjTpRQ4XFyrsfme//jTx/r+dnBcT+JZb6wUW1gpMVHvMjbepFOQEhxBPwtpsjfhaM6De9qkUYmbHOvt60lYIcfCd1Lc4YRe4YY8NdRweMd9yvs80q/yF+QJv6E+pqtajP/EQeepw2u12+d3f/V2++93vUqlU+LVf+zW+973v8a/+1b/iD/7gD5iZmeE3fuM3+P73v883v3kwzjQ8C6UU08fOc+3mInEco1Q27TPvJ3j9brBaW5RSWAtJqohTnVUGY93f2uRulmIuYbLaxfcscQKdXtb4Z21Xpq4qwtgljHf+mB1tKOUjAjcbt7GKOMn2Bl2p5+4Zq+uYQXDNeSnBXUHWc3ZWiu+ujG6tGfUcw0gxpJhLSQy0uh5rjYAVm+d5FacOYMn5KR/fHKHZaXNmpol+PvL5odXsulxcqNLseowUQ8YrEdeXi8T3OVm1F3qxRy8Gz0kp5hLWmwEr9TzVYsTceJtaOZS1zEI8QJwqNlo+682AtWYwWEqS9xM2Wj43louMV3scq3WoFmL5tyTEc84dNEd6m4V0mhRnKOPwifi7zv9ignX+wnyRz+sPKavOUMayl546nKZpijGGbrdLoVAgSRJKpRInT55kbi5rufztb3+b7373u0cinN7NWkUvdunFD/0oXMeSDxJ8x+A6BqUtjgbfNfTirNp5e+PB3WX3wv2rrP2pwX6C76U4mqxRC1lH2yTVdEKH9ZaPMTsPwLU25PrTm7uhQ68fhl3HMFKMmB7tYoyi1XVZbwWsNI7mNN2np+hFLnk/Zn6tSKPr8cqJTTlzfwAlqeLqUomFtQKeazg322B5I8fl2+WhjCdOHTbbDloZqoWQbujywfVRCv3pxVMj3QO3zlyI/ba1dGa9FfQrpB6gcB3DaDGklE+yLcraPoGX9RrYOuFTymXbk02OdKURmRDPsa3mSK+rj/mZfXXfnz9Hj191/oxRGvy5+RJf0u9TVN19H8d+eOpwWiqV+Mf/+B/zrW99i3w+z5e+9CWWl5eZmJgYfMzk5CRLS0u7MtDDSZGkKgt2wx7KY0iNphNpOpF33/dvVYv9XIzXnybs6KyTcJIq4sShlE+YGu1hDLR6Hhstn9XGbjcwOpq6kUfgJbS6Hm9frvHyiU1Gig89+yH2ibWw2gi4dKtClGhmRru4juHK7fIDOzrvJ2M19U4AWCqFiCTVfLpQ5epSiWO1DrO1Dp403RLPkeSu6ujWlPhiLub4eBtXZzMgshOmO2fvhLE7OOETpQ6fLlS5slhmZjRrRJbz5cShEM+breZIX9AfcSk9TYvivj13gS6/6nyPCm3+3HyJn9PvkFfhvj3/fnvqcPrJJ5/wX/7Lf+HP/uzPKJfL/JN/8k+4du0aatv8F2vtjtuP42ELZIdtbe35nX4K26vFD3rZWNo9V8LoMwhjF0cZUPDOZ2OcnWlyvNaRaWVD1IuyhkdrzRzFXMypqRY3Vor0ooPY7Dzr8g2WUi5CKcXVpTI3VorMjHY5Pt6WA2txJG1tQ7YVRusdH2sVjjaMliJKuS5xqqi3feZX7+xr/SDbT/iUchGOhpurRW6uFhmvhByrdRgpRvK7WYjnSKBiUqv5ef1T/h/z9X15ziJtvu38GUW6/Ln5Ej+vf0pORQ/9HKUV1WqB6sRwZnU9q6c+uvrLv/xLvvrVr1Kr1QD4zne+w3/8j/8Rx7kzD3tlZYXJycknetyD2q0XoNE4muXz3aM4oD+6QyW1mjS25IOUK7crNDoeLxxryHYz+8wYmF8rcG2pBChOTTVpd909ani02xStng9Azo8JPMvCWoH5tQKT1R4zY10CL8XR2dIDrZ6sw7gQw2Yt9CJn0OV+vekPTpwWgqypkeda2l2HjXbwDCdN7/xbCtyEfJCy2c5mBBWCbMrv1EhPfj8L8ZwwKM7oeY7ZRRbs9J4+V5kW33a+R46Iv0i/yNectwnUo2fUWWOp1ztENB/5scOwZ916X3zxRf7Nv/k3dDod8vk83/ve93jjjTf4b//tv3H9+nWOHz/OH//xH/P3//7ff9qnEOI5puiGLnk/YaWeo93zeOXEBsWcbBmyV1KTbV+02fapt33qHQ9jNGPlkNFiyPXlEok5fIvOepFHL8q2aioECWvNgOX63bNAspCahdWsUZqrLY6TbU3laoujzT0fs7U+T4Kt2GtJv6N7o3PnLWsol/VLyPob9EiMot72WFjb/V4OYeISJi4KS7UQEqeaS7eqfLZYZma0y2ytQyGQ39FCHGWeSomsy9f1W/xh+i3MHjVHqtLg2873cEn5S/N5vu68haeej478Tx1Of+EXfoGPPvqI73znO3iex2uvvcZv/uZv8rWvfY3f/M3fJAxDvvnNb/Irv/IruzleIZ4r3f52M1Gi+emVGi8erzNRPbrrDPZTkmYdquttn83+/sJb60eLQczUSI+8n7KyGXBlsTLk0T67ra1otDKMlXs7Gm5Zm03bNxZSozAm6zbejRwSo0hT/cAtnoq5mJnRLlOjXVnX+piMgTjVxIm+61INbkfb3mesYqLS49h4h9KQ90NOUsV6y8dR4Pf3+PZds6snKLb2w250POr9f5vtnstW2Mz7CWPlkJxvSFPV/9jd6nL/GONDDab8FoMYz92amVBkrBRybLzNSDGSBkpCHFEuCSOqyWvqIu/al3b98UfZ5NvOn6Gw/JX5HN/Qb+Gq5+fEl7LWHqijiYM8rffq1cv86Ed/QRxLkxqxvxSWwE/pRS5z421OT8t2M09qa71ZvZ1VR1tdr79tkqWUT6jmI1zXEsWKRtffcTD8PFJYHMfiaoPjWBxl0f1LpSxaZdXmXuzS7nkoZZmoZFOG93MtXi/SrDZyrDRyNDsevpsS+Nn2V9v3bA52Yd/mB7E2C5tRnIXK7M0hireHTzUIoelDKvCuY/C23tzse2+MYr0ZYKyiWow4XmtTq4T79jvAWqh3PG6vF1ip328vajsIqVv7evuewXez8BpsXffMfcecpIpG16PR9rLLzp09xB1tqORjivkER1vCKJtm2+65D9iibTh8J6WQS2iH7mBrGt9NyQfZlnN5P3stbl333IN5nCOEeDyh9VBY/lP6d+lQ2LXHrbHOrzp/hkHzE/M6X9dvZb1InoDSAeOn/h756oVdG9du2rNpvUKI/WP7280Ugpibq0UWN/LkgyQ72Nk6AO8f+ATe7lYxDqso0YMgutm+EzaVspTzMcdqbRwnCzeNjs/C+v5u6XTQ2W3dxnnEllnFXIzvGtb7U4ZzfsLMaJfp0e6ebInUCR1W6jlWGwHNbrYesBDETI12B1XHbI9lzd0/0+37Ngfe9r2bs/u9fhXQ9qvIg7AZO9uC57237/fa0criuXeCZt6PB3tiawUWMEZhTBb0U6NJ0izAdkKXpKsGj+tqQ6UQ0Y1cPrwxSuClzI51mBnr4O9R0IkSxdJGntsbBTqhi6MNUyNdcr6h2XXQisFWRanJthyLYodm1yN+wPfEc8yg4uo5hnbPpR1unQiyFIKU8XKPwDekRtHpOTS7PhvtYE++xt0SpQ5RO9u3uloIyfuGpP/6Wd/WLXiLqw25bcE13//9nQuycC+/w4U42AIVY6ziq/od/qf5+V15zElW+bvO/yLC46fmFb6uf4Kjnr8TWVI5fQJSORUHQc5LKOUT4lTTizRh7LD9IFApOzj4zvUPenJ+Qs7LrrvOszW/yaaAgulPAzU2O8pWikFjna3L3TjAMgYSk4WkdMelus9tTZIqupFLJ8zOvWmVba9SzsdoDd0wmy744K7T4mlplVW5Uqv6odFSK4dMj3afqdK3tU/lSiPHaj2gHWbbXZXyMWOlbJr7Zsun0c32r4RsnWwWgrLXvNag1Z0ptWGc/du5u4qpVFYF3JpOezdFFji3qoRev7qs+g9jjOoHzey1GSUOcZK9Lnen0pedXIFsDaZSlqmRLsdqHcr5Z5/yay1stH1ur+dZbeSwVlHOR4xXQ7qhZrWef+ja663vj+dk1VJH28H+3pAF8SS9M305H6QUczGOgjDWNHsenQNWFX02Fs+xBF6C36+EOzp7fcSJohc7hJGz4+tVyg7C6kgxYqLaky7bQhxAsXXwVMofJX+D29xpAKswuKR4JHgkuP1LT8X9++M771PJ4Po5dZ0uOd41L/IL+m30UwbTw145lXD6BCScioPG0Vm1x/cMjjaDNU5Jmp2x70buYHrc9s/ZmmLmOiar3GwFTaMGaw+NVTveZ7ddf/wKY1Yh2poGqnaE13vvsyjSVA3WOSb98TzK3VNQAy/tr81TdMKsMnp35ULsrbwfk/cNrZ5LlDh4bsr0aJeZ0e5jNY2xFhpdj9V6wEoj19+6x1IpxIyWIoyB9Vbw1NOvlbL4TlY1zRo9GRyt+ideGDR/Uv2xbJ0MifuVzSjRD6wO7pe8HxN4hkbHw1hNpRBxvNZhvNp74hMBYaxZ3MhzeyNPL3JxHcNkNat8rzZyNLu7Pc3d4jmmX6F/fhdnamXJeQm+Z/qNxrL74zSbLbN1kq2cj5is9iSoCnHARDY7mRbi3wmiT7g+NLYuMQ6bVPjMzPE1/dNnOrkv4XSXSTgVYrdkFaCcl+K62Vo7R2+vHDkkqepPMbT9KXr2PtXPbSGSfjVUZYepavBM2QE8225jd77PklVYs9tqUIG1qEFDHgDHMdn6xv54lcrCJ/2pllmAzgKEtVlgyN52VlAf1MBH7Lescq1Q1DtZZbNajJgZ7TBR7e1oGmMs1Ns+q42AlXqOKHFQWEZKEdVCRJxmUyS7B3KP2eFxtKGcj+lFDr3YxXezKb+zY138h0yrNhbWmwG31/OsNQO2fja1cm+wZ/XD1seKvaWwFIKEwDeEsabdy2YMHOagaiys1nNobRkthdI0Shx6xsItO02XgBSN2fa2RVmLIs2ObQbvtTgkOBg8leD1K6pl2s8860zC6S6TcCrE/lDKErgplmepjO6mrfZEEiqPqsBNKORSumEWorI1jD1GihHrrWzvyCTVaJUduFYKMWGkWGvl+9PXxcNlU34V0Oj6KGWZrPY4VutQKdz5u9WNnKxKup4fVLW39upc2bwzbVocJJZikBD4KWHs7AiqE9Uekwc8qBoLy5s5ri2X+rMgspMqtUrIRKXHWFmCqhC75bCHUzn9LMRzylp1wNZdKg7maSmxW8LEJWxl03PLuQjHsSxu5Lm1XsDRhrFySCmX0Akd1lvBvm0NcnSoQYOonJeQD1JWGwFLm/lBtW29FbDRyj5mtBRxvNSh2XG5tVaQ2QYHmqIdev0TB9kWNltB9bPFCp8tVgZBdaIakvcPxrYTdlso7UYuxVzMuZkGGy0fa7Mu1Mub+SyolkMmqhJUhXjeHaQjUyGEEM8FRbOXBSRXp4yVI7TKppiu1PNDHtvR0IvdrDqtDCOFkDBxuLJYIfBS5sbbKA0rmzk+WywPe6jiie0MqgU/JhdsD6oMPahaCyv1LJR2QpdikIXStUbA5dtltmbn6P7rUzuw0cq6fettQbVWjvZk+ychxMEl4VQcOXnrc4IJjjOOg6ZDRJeQLuE915PnaFNjIQ6ixDisNiSQ7pXUajY7AdkWJxGuY5hfLcr0+SND0Yk8OtG2oOqnhMnOoDpeCalVQopBsqfb1FgLq42Aa0sl2qFHIUg4N9NgvenvCKVbzOD1mQXVaiGbUbHZ8lmp59HKMlYOmexXVF1HgqoQR52EU3EkaKuYYYwTTDBBFYtllQYhMQEeVQpMM4rDzrlCsU3oENIl6l/uDLAhiWx9KYQ4AhT1jj/sQYg9df+gGiUOV5fKXF0qE3gptXJIrdxjpBTt2vRZa2GtGXB1qUS755H3k8H03fuF0vsxVg9eo6rfSM11LPWOx2ojNwiqWUVVgqoQR5WEU3F4WahS7FdJa3i4dAi5yAIWyzFqTFAFICGlQ0hIRERCisVi0Sh8XPL41Cjj3fVPIsXQtSEtetRpU6dDnTZdIgmtQgghDqidQTUfxOS9FGMVi5s5bq0XBo3HauWQsUpI7iGdnR/E9js+X10u0ep65PyEs9MN6u3HD6X3fVwUjW1BtVoIcRxo9IOqUpaxUki1GBN4af/NELgpWtarCnGoSTgVh45vXY4zzgkmqFAgxXCbdep0GKfMOWbRd/1BdHEok6fM/acPGgw9Yhp0iElIMKRYNODhUiRgihFU/3EjEuo2C6uNfmht0UX6iQghhDhYFN3Qo9vvwuxqQ6WYNR1q99ys8dgtKOXiQVCt5OOHTv+1FjZaPleXSjS7PjkvC6WNjseVxacPpfd9LhT1/tTfrKIa4jrQ6nn3aZqWbaE2CKvbgmuuf933zBPvAyyE2D8STsXhYGGSKieYYJpRNJpNWnzMDQI8jjHOccaf+uE1mgIBBYL7vt9gaNKlS0hMikZRIOA0U4OpwgkpTdvdUWFt0MEomXokhBDiYEiMZrO99bfOUs5HBF62l+r1lSLXV0p4TtaorFbpMVaKBlNorYXNdhZKGx2fwEs5M92g2d39UHo/WUW1H1SVpZKPBvt4a509e2ogTjSd0GGj5d9nr96dATbnpRRzCaOl8EBvxzMMGy2fa8slklRxvNZhaqQrlWmx5yScigOtaAPmmGCOCfL4hMRcY5keMTOM8iJzg2rmXtJoKhSoUBjcZ7F0CGkTEhMDihwes9Q4xRQABkvLdmn0w+oKDRqqs+fjFUIIIR4t236o2c1u5byEQpAAirVmtg2RwlItRoyWItZbAfW2j+9mobTddbm6WB5Kgy1rFY3ug9dRa2XxvZSSG+M6Fq1t1vnXqizApppO6LLe8jH9AJsPEsZKIaOliJFi9Nyua210PD5bLLHZDvDdFM81fLpQ5dpyibnxNjNjHdnuR+wZCafiwHGsZpYx5phgnAoWyzJ1PmORCgVOMIGLM+xholAUyVFk57SikJhVGkTEGCw+HjUqg8ruhm1xnWUWWCNVcpb2qAusR4UCddpEKhn2cIQQ4oG2tiCCO9u8uK6lE7pcXSrjuymnp5p0Qz20UPq4jFX0Ipde9OBDXa0svpvi52J81xKnmtvrBRbWioClUogHYbVciI/8dOBW1+XqUom1Zg7PMYOf9eJmgVIQgVJcvl3h+nKJ4+NtZmsdvOc0wIu9I+FU7BttFQEeOXxyg8s717fe5/dflm16fMoCCjhGjSlGhjr+xxXgEeDtuC8hZZ0WCQl5At7kDK9ykgW7xnWW2aQtDZaOGGUVZ5jiBY4PTqa0bY8NWmzQYp0WDTpYmfYthDiAtm/zApZiLiLvp1xbLmGPSIMFY9WOQA7gOSnVXILjWHqRw7XlEteWFY42jPQryGPlbP/YvdyWZz91QodrSyWW6zkcbTk50SJO4fpyCdP/WbfCrEpd8GMcB64ulbmxUmS21uF4rUPwFA21hLgfCadid1gok6dIjhwewY4Aeid83s1gCInpEdMmO3BPMETE1Chzgdl9mba711wcxigB2XTgTVpEpByjxkkmqdPmul1hnlXZe/UIqNkyr3GKCgUW2aBBlyIBLs6OKnqKYdO2+4G1yQYteioe8uiFEOJuinbPp90b9jj2Xpw6bLS3ZmdZ8n5C3k9QWtHu9htI3YbASxntV1VHSyG+e/hONPYizbXlEosbebSCufEOFsv8WuE+a3UzWQdoyHkxgWe5uVJkfrXIzGiHuYkOeV+OYcSzkXB6iHnWIcEMrfKiraJGhWlGmGJ0RzMhgyUipkdEh4g6bWJSTH/7FgeNi8bvB9kCASMUj0QQfRSFYqQfVENi1mmRw+N1TvEyc9yy69xgmXVaUk09ZALr8QonOM44HULe5SonmOACo4OPyaroTUIiLJDrN9Y6xwwAXRuy3q+ubtCiTluaagkhxFAoupFHtx/IlLKUclkDKWMVq/UcixtZL4pSLqYQJINqqlL9Sc8q6zJ85/7sdnb/tttb1xX4bkqlEO9ZdTaMNTdWitxaz8Y+O9bBdSy31gokDwild+vFHr0YfDchHxhubxS4tV5gaqTHiYkWxZyEVPF0JJweNhYmqHKaKaYYIcWwYVus02StX3nZy3WMvnWZ6ofRSaq4OKSkrNDgOisEuHi4+Dj9qmlA9TkJnU8jwGOSar+a2iYiYYYxTjBBgw437Ao3WSWWtYoHmrJwmmle4DgaxSUW8HB4nVP3vPazKnp5cNti6RLRokdMgoNmhCLHqAHZ7IK67bBBi2U2WVb1ff3ahBBCZKxVtHo+rX4F2VGGaiHEcy1Roml0PbBZV2Frtz4nu53dT/9+teP6g7jaUC7EVPJxdlmInqlCGyeKm6tF5lcLGKuYHu2S91Pm1wrEydP18ogSlyjJpkMXcikr9RxLm3nGKz1OTLSpFGQ2kHgyEk4PCc86zDHBKaYokSMk5gqL5PEpkeMCx1AoTH+a4Fo/rK7TfLZpov3pulOMMM0oo5RQKHpELLBGh5ASAROMMr2tOiSeTFZNLQIQk7BCHQ+XVznJS8xx265znRXWaEg19YAZsyVe5zQVCiyzyRKbXODYfaex34/qb0u0feZBismm+BJhsOTwOcEEZ5jmml3iA65LNVUIIYYstXqwByuA5xi0zn43a2UHldB+XH1g9ZTtldP+LdcxZJVbh+srRbY+MuclVArxILSW8vEjO+cmqWJ+rcDNlSKpUUxUe1TyCfNrBRbj3WkwGacO9baTBfZizEbLZ7WRY6QYcnKyzUgxOjJrdMXeknB6wFVtgVNMcYwaLg7rNPmQJcoUBntsWixNunQIMRjyBJxlmvPMYrE0bGdbWG0QPqIKp6yiRnkQSLe60dZpc5nbpBhGKXGc8cEen2L3eLhM9KupdTqEREwywnHGadHjhl3mBivS+XXIAuvyMieYY4IOIe9xlWOM8xqnnvmxHTSj/anfW7pEtOlxiikqFPiJvUQo61OFEOLAiFMNuz6b1ZL3U/J+1qQpSTWbbZ/leh7YmmqcUM5HVArxjunAqYFbawVurJSIU02t3GO0FLGwVmCl//m7LbWaettHkVWV26HLu1fHKOcjpke7eK7BdSyu7l862eVR74QsHp+y1h6o0+9ray2MOVBDGrh6+RI/+vFfEid7e0CorWKWMU4xxRhlElIWWKNFl1lq9xy03m1r/80WPVJSfDxGKA46hrboDsLqGg26RHg4O6brerikGFZpsEETF4cJqlQoyBTdIUhI2aSFi8MIJVIMN1nlM27TUs9Bh4oDRFk4yRQvcRwHzWcsotGcYgq9D/82uoR4uCSkvMUl1lVrz59TCCHEwaGVpRjE+F6K1ooo1rR67qCJ0dZ04HbPJUocRoohE9WQW2t52uHjzerZPdmWPGHsED6kSqu1wdV3wqrr3Od2/9Jz+lsAednHSEV2J6UDxk/9PfLVC8Meyn1prajVHpxlJJw+JmssG/+f9zBpSoMOdTrUabNJmybdXWlKlLc+p5jiBBMEeLTocZMVfFzmGMd/zGmC99MjokGXmAQflyrFwZYtPSJ8PDSKkJglNmgTUiRgkhFyPHiTa7H/mnToETNGGQfNIhtc4TZrNGXK7x4btSVe5xRViqxQZ5ENzjO77/9GYhISDAEuH3CdayzLz14IIZ5jnpNSCBK8/prUbujguobJasjiRo5md9jHcpZqIcJzDVplFV+tVHbGl2xbH2sgNYrUKJJUk2xdpooH/ZEb7FXrGXzX4LspwdZ1L8V3DYFr8Fzz3IRYCae77KCGU4C1H15m8dINcvhUKOD1K5EGQ4MuddqD0Nqg83iNifoNjk4xxXR/H88lNlmlwQRVJqnuSaUyIqHRb8CjUbQISUgZocg4lUGVVRxcITENOlQoEOCxQYsr3OY26xyRLegODN+6vMQcJ5mkS8RFFjhOjRqVoY3JYmnTo0Se6yzzPtdkHaoQQog+i+8aoqdsdDRsCjuolDo6mwastUXrrAuy1halssZTqVHEqSJOHKJEk6T3W3Jm++E1C62Bm+1bOzV69GafSTjdZQc5nEZXN+j+aAEbpfT63TUjYgyQw6XcDwmQHTi26PUDa5vNfmjdak50p8HRJCXyhMTcZIWElDkmBus8hXiUreY5eXyK5OgQ8hmLXGd5Tzs3PxcsnGSSl5jDRXOVJQBOM4U+IOutW3QpkWeDFj/hEj0VDXtIQgghxL7bqqJurWt1dFal1RqwkFpIUk2caHqxQ5JqXj25wXglHPbQd5WE0112WMLpfd9PQpMOYT+w+riUyZHf1oWzTY82PcYoDxoc3WKdCgWOUZMGQ+KpbW1Hs9X5NybhGstcZZGeNM55IsoqJqlygWOMUmKNBguscYFjB3Ka+9Y61BTDW1xiTTWHPSQhhBDiQMqqsmZw60vnV/G9o3My/7CHU+nWu4t83Hum+SWkrNMiIiLB4OGQJ+AWazTpMcsYr3JySCMWR4lCDZplNegQk3COGc4yzYJd4wq3aajukEd5cCmrmKDCLDWmGcXHpUfEe1xjhlFe5/Swh/hAeQISUgyGr/ISH9rrWZVXpncLIYQQO1gUcepk1xR8slDltZMbz82a1INOwukec3EYu6u7boohj88J+faLPVKhAECHkDY9ZhhjjgmW7SZXuM2K7JcK3D+QxiQsskGTHmXyvMqJAzOF92FcHBw0LXq8xilGKPKuvSrrUIUQQoj7UigM682A2+t5ZmtyAv8gkHQ0BA5apu+KfVEgoEBARMIKdSoU+CovUafNFbvIIhsYDKa/Gfjz4E4gHWOasW2BdJMmXcrkmGaUuUP461GhKJOnRZfjjFOmwE/sRbqyDlUIIYS4h7Ea1zFcvl1hpBRRCHZ9o9p9FSWKdugyPuyBPIPDd/QlhHhiPi4TVEkxrNEgwOPznN3xMdZaLBbDoy9N/9rWfW16XGOZ5gGdNqysYrwfSGe2BdIlNqnToUK+H0gP86/zO0rk6RFRJMc3eZW37GVWVeOZH1dbRYUCI5QYoUgenzY9mnQHbyHxc3OiQwghxOGXpBqtDR/fHOFzZ9fQh/RvWJwq3vlsjCTVHD/XJV8d9oiejoRTIZ4jDpoaFSyWDVo0VAdj7SBLbG1bpPrXdl7e/R9oFBqHE0xymmnWbINrLHOb9aFPJ90ZSEfx8YhJWWKDBh1K5JlhlONHJJDeLYdPQkqM4au8yEf2BldYfPzgaKFMnhGKgzBapTCY4hwS0yFkltpgz2ToN4azWVBtDUJrtjevhFYhhBAHkTWKZtfj+nKJ01OtYQ/niaUGPrg+SjdyuTBbpxOXqA17UE9JwqkQz6Gt5kmj9sHd0h6H7f8XkbBOlwIBX+AcITHX7TLXWd7XKaWOzcL3DKMPDKTTRziQ3m37OtRXOEm1vw71ni2GbDYFfIQio4MgWhzsdxyTUKfNVZZJSPFxKPWD61Y1PtspOUWjyeExyyg+k4OniLeF1q23Fl26RBJahRBCDJVF4WrD9eUiY6WQavHw7HJgLXwyX6Xe9jk/W+f6yijHXvCGPaynJuFUCPHUtmqoOXxy+Fgs6zSxWM4zy3lmWbKbXGWJFep7EkKKNmCSEaYYoUYFB01CyuJzGkjvdmcdao9j1ChT4F37GQEeI5QY7QfRrT2aUwwNOtxklYgYF4ciOUYoMc69c4Q8uO/2OiExazSJiEkwaBQ5PKYZ5eS20JqQ0rAdNmixTpN1moQq2bPvhxBCCHE/idE42vDxzSpfPL+G6xz8hoLWwuXbZVbqeU5NNZlfLZLawxtMQcKpEGIXKRRjlAHo0KNFjxFKfJXRbF2qXeYmK0TPED60VdSoMMUIk4xQIgdAiy7XWSYkpkSOqec4kN5PiRw9Igr4fINXATBYWnRZYoMuERo9qKCeYnIwzftpBHiDwLtdSJwFUGJSDKAo4HOKKc4yA0DLdlmnyVo/sLbpSXVVCCHEnkuNIjUOl29XePF4fdjDeaT51QILa0Vmx9qs1gO6kYt3uLOphFMhxN4okKNAjhTDKg1cNK9wghc5zi27xlWW2KT9WKGj0K+OTlJlnAouzuBx51nFQTFGhVNMoSXFPNDWOtSbrNIjJIfPCEXmmHimIPokHhRaO4S06BKT4uEyzRgn+hXWkJh122StX1mt08HKFjlCCCF2ncLRhsWNPLVyj4lqOOwBPdDSZo4rixXGKz26kUOrd+8spsNIwqkQYk85aMapANCkQ5eY6f6+q5u0uWaXWGBtxzpIbbMK7FZ1tEwegDa9frCKKJFjgipTjAzjyzq0XJwD2ZV4a9ujLTEJqzSIiNFoKhSYYQzIpgJv2vYgrG7QIlEPbv+vrMLDwcfFw33A5Z33axSbtFmlwRoNeurwrD0SQgjxbFKTbS/z6UKVSmGVwDOP/qR9ttHy+WS+SrUQ4eqU1UZx2EPaNRJOhRD7pkyBMlnwWKFODp83OcMrnOCGXaFNjwlGmNhWHV2jwS3WUChqlDnJxKBjrDi6PNzBSQ3I1sJu0KJHhCULsxeYRaGwWBq2Q50ODvqe8On1Gzvdj8USkxCR9i8TFDDD2GBtbMt2WaXJGg1WaRBKWBVCiCMtSRVKwafzVV47tYE6QJOyWl2XD66PkPdTqsWIGyvP1tzyoJFwKoTYd15/39WtLW0SUk4zhUbTIWSeVbpEFAmYoMqkVEefew6aUe78AbZYWvTo0CPBEOAxSZWElIiEHhEtuiSkg/14Ldn2Rw4K3f/Pw+kHWLff+MlBoYj6J1BiEjxcjjHGqR1htcEaTQmrQghxJCkUhvVWwK31PMdqB2Mf916kee/aKI62zI51uHy7POwh7ToJp0KIodna0gagS8QqdfL4nJDqqHiErS7EW1O+AXpEg/vdfsh8Wn7/BMqWiJgV6kQk+Lgco8YppgBo2u6gqrpGU8LqIRVYj1D24xVC9BmbTe+9crvCSDGimHvw8pH9ECeK966NkRrFudkmF+crHMVfWBJOhRAHQh6fOSaGPQxxiN1vS5vd4uPdN6zeqazeP6yu0CCWrXEONgtnmOYVTlCnzbv2GnXVHvaohBAHQJJqtDZ8fHOEz59dQw/pvHlq4P3ro3QjhwuzdS4tVLBHMJiChFMhhBDiid0dVsN7KqvjnGIqm7pu2yyzyTKbj92hWuwPbRWvc4oTTLJCnTJ5vsErXLNLfMz8QxttCSGeD9YoWj2Pa8slzky39v/5LXx8c4RGx+PCbIPPFisYe3Rnl0k4FUIIIZ5RcN+w2iImpYDPCxzjRY5n99s6y9RZYZNQqqpDE1iXL3KBGmUuskCVIj4um7Q5xRQzjPGhvcECa3JCQYjnmEXhOoYbK0XGyiEjxf1bumEtXLpVZrWR48xUgxurReL06AZTkHAqhBBC7LrtYTVr3tSlTQ/V31rpeH87n81BVbXOBi3Zv3WfVGyBL3OBAJd3+YwzzAzWL49SoktEQsoXOMcJJnjPXqOtekMetRBiWJJU4/an937p/Cqusz+/q2+sFLm1XuR4rc1KI08vOvrR7eh/hUIIIcQQZU2aCpQpANk+rWs0CEnI4XGOWS5wLNtiyTYGYbWnoiGP/F6O1cwwRkRCiy4dwkNXVZyxo3yOs8SkvMs1XuUk/l2HQ3l8LJY6bUYo8ou8xmV7i0vcwsgJBCGeS4lRJEZx6VaFl+bqe/58ixs5ri6Vmah2afdcml1vz5/zIJBwKoQQQuwjF4fatj1cW/Ro08UCoxSZZQyAhu2wzCZL1FmjMdwQaOEYNV5mjjzB4O6ElJbt0aJLk+7gsk148KrAFi70p1dv0OIW67zJGfQDvrEKRZUiCSlNOrzAcY4zznv2Gitq7w9MhRAHjcLRhqXNPLVKj8lquGfPtN70+XS+ykgxRGPZaAeP/qQjQsKpEEIIMUQlcpTIAZBiWKdJSIyPx2mmOccsm7T5xM6zzOa+h9SKLfAaJ6lRoU6bS9xmggopBsimMI9RHkxVBjCYfmjt0aTTv8zC6zAqj47VvMkZjlHjJqsYUl7hxGN9rovDCCVa9NAovsqLLNg1PuC6bBskxHMmNRrXSbm4UKVaWCXwzK4/R7Pr8uGNEQq5hEo+5sZq6dGfdIRIOBVCCCEOCAfNGHc2Ve8Q0qFHgRw/xwus0+QTO8+qauz5WHzr8iLHOckkEQkfcJ0aZV7j5D17yCakbNKiR0xC1uHWx6VCnhlGBx9vsXRsSLNfYb3N+p53MM5Zny9zgSoFPuYmk1R3VK4fV4kcFssmLaYZZZIRPrE3ucYS9pBNbRZCPL0kdVDK8snNKq+f3kDt4r//buTw3rVRXMcyPdrlyu3yoz/piJFwKoQQQhxQBQIKBFgs67QoEPDzvMSKrfMJ82yo3d/WQFk4xRQvcBwXh6sskZLyEnM43L9L5FZ18W4phiZdukTEJFjAw6FAwCRVzjNLnTbX7TLzrO361i2jtsSXOI+D5h0+4wWOU+Dpp8cpFCOUCInpEfEap5hjgvfsVTZlb1QhnhuqP9V2Ya3A8fHOrjxmlCjeuzqKtYpTU00+Xahw6Bb17wIJp0IIIcQBp1CMUcL0p/1WKPB1XmHRbvAJ8zTU7hwcjdsKr3KSCgVWqHObDc4zs2Od6ZNw0FQoUOk3g9pisHTo0SUiwON1TvMyJ1iwa1xneVeqqXN2nNc5TY+IT5jnNU7h4jzbg/YFeAR4NOmQx+frvMI1u8zH3JS9UYV4DhibTe/9bLHMaCmimHvwtmDWgt26tArTv7QWzLbLSwsVwtjh/LE6lxaqPI/BFCScCiGEEIeG7k/7TTGs0aBGmV/kNW7ZNT5hntZTbneStz6vcJJZxujQ412ucpxxXufU7n4BfRpFiTwl8v2psm1iEo5R4ySTz1ZNtfAyJzjHDCvUWaPJG5y+ZyrybihTwGD6e6NOMsMoH9obLLGZTW9+Po8thXguJKmDow0/+2wMzzF3wiZ3h9DH/UVguXCswZXbFcxzvFZAwqkQQghxyDhoalQG29JMMsIMY8zbVT5lgY56vC6SjtWcY4ZzzGKxfMo8OTxe59SehLn7yabKFgGISFihvqOaesuucZ0VNmg9Muy51uELnGOKEa6xhIvDixzf0/Fr9D17o26JbUpC9hZz5/qD7rv7do9Itq4R4gBLjWK0FOFoi1IWBSjF4Dr926idv74UWTV1iwUcbbi+XCJJ77984nkh4VQIIYQ4pLa2pYlJWKfJLDWOUeOGXeEitx68V6qFWcZ4mRMUCFhgjTptzjF7z56f+8nHZYLqoJoakTBLjRODauoK86zet5patAFf5gWKBHzAdY5Tu+862L2SxwdggyZ1utlBKgqNxkHh4OCiyeHh4eD23x52EmCrgVSbrPNxtu1Q9nYY95gV4uhRbLSen21e9oOEUyGEEOKQ83CpUSEipkmXE0wwxwTX7BKXuUWo7qyHqtgCr3KS8f7WMO9wizNMcYzaEL+Cne6tpm4S4PM6p3iZOW7Zda6zPKimjtsKX+Q8YHmXq7zEHLl+WNxvo5QZ5f4dNi22XxU1/c7GPRJSUgwGi8VisP1b4KBwccnjM0cZb9ua2RRDx4a06A7C69ZlSCzBVQhxKEk4FUIIIY4IH48aHiERLUJOM81JJrlql7jBMmeY4RSTxP2tYcYo88Y+TuF9Glk1dQSLpU6bkIRZxjjBBA06rNkGJ5miRZerLPE6px/YVXjYFAoPF+8JP89i6RJRp01E3N9jVuHjUiTHJCM7vuaElJbdCqtd5lml/ZhTvYUQYpgknAohhBBHTIBPgE+XiC4h55jhPLMYLNdYJiZ56NYwB5FCUb1rbaqPy2mmWWSDJt19XSu7nxRqsK3Q3QyGDiFdov6qV4NG4eMxQpFZxjjDNG/bKyyrzf0fvBBCPAEJp0IIIcQRlccnj0+HkA1abNDkDDPPtNfnQbB9beoGLTxczjM77GENhUYPOh/fLSGlSRcHzVe4wMd2nsvckim/QogDS8KpEEIIccRtVd0O0rrS3aBQjO5j06PDxsWhQgGLpUWPl5mjSoF37Gekygx7eEIIcY/DM59HCCGEEEI8MYWiTJ4WXWYZ4xd4mbwdTsMoIYR4GAmnQgghhBDPgRJ5QmIK5PgGr1Kz9+8qLIQQwyLhVAghhBDiOZHDR5E1UvoqL3LKToId9qiEECIja04PscREOMpDKelsIIQQQojH4+LgoGnT43VOU6XI+/YaRu1hSrUwzSgnmaRFl0+Yl3WvB5yyihplFLBKAyuHm2IfSDg9pG6FV/iw9RfknTJn828y7Z9GKSmECyGEEOLRFIpSfx3qSSYpk+cn9hKhinf1eRyrOcEEZ5imSI4eEZNUmWWM9+w1lmR7mwMlsB6TVJlilAmqeDgA9IhYsGvcZJWG6gx5lOIok3D6BObnPyCfOOT7+6wNg7WGi523uNZ7n6o7QWwi3mv9L67on3Gm8CbT/hn0EEJqK9lgIbyEISWni/23EjldINDFoYxJCCGEEA9XIk+XkCoFvsGr/MReZFO1n/lxc9bnDFOcZBIPl3WaXGWJMUqMUMJi+QovcMuu8T7Xdz0Ui8dkoUqRaUaYZGTQ/bpLxC3W6BBSpdjfU3iKs8xQt23mWWWeNfm5iV0n4fQxGZPy7rv/nTSOeaX4NWaCs/s+hiyI/hmr8TzHggukJqVurlBxxklszPut72chNf8mM8HZPQ+ExqYsRde42fuYjWQJhUbjkHLvL6pAFcg5WWgNdJF8/3IryAa6IAFWCCGEGII8ATEJCvgaL/Ouvcq8Wn2qx6raImeZZpYaCrjNOuu0mGWMVzk5+DiLZZP2oEL3kb3BdVZkD9Z94FqHCSpM9QNpDn/w87jIAgYYo8hxxnG2tafpEtKgS4DHK5zkZU6wbOvcZIVFNvZ2Wrh4bihr7YF6Ja2ttTDmQA1p4NLb3+eTv/hTOqbBbHCel4pfxVXevjx3O63zs0b23BcKX2Ix+ox6svMPR9mpYUhop3XyusyZ/BvMBud3PfR10yY3e5+wEF4ksj3yusRscJ7UJqzG87h4uCrAUS6Ozs5/pDYhNj1C06Fn2qQkdz2qItD5QVgd82aY8c/haWl1L4QQQuwHi6VNjxJ5rnCbj7jxeOsM++tJzzJNjQoxCTdZJSTmJJMUCB74qRExPWIqFFilwXtcpaV6u/dFHSIlm+MFjlMiR4+YHhFh//sTEvUvs/ufNAgWbY4pRphihBplNJqYhGXqNOjg4zFJhRJ51CPOEFgsTTp0iamQH5zcuMU6N1llnaacZBgiz/P5+tf/N44fPzHsodyX1opa7cH7U0s4fQI3Pnmbn/z//hNBkqdtNinoCm+Uf4mKO76nz7sSzfNe689QKF4ofplL7bcJ7YPn+5edMSyGVrpJTpc4k3+DY8F5tHKeegzWGlbieW72PmY1ngcUE95xxrxZ1qJbrCXz2Mdo96dxCFSBQBfwlI/WbtbUCUvSD7Bd06JrmmhcZoIzzOVepOKMS+MnIYQQYh806VCmwDJ13uYSsUrv+3F3ryftEHKdZXw8TjCO9wQT9Bp0yOOj0VxigUvcxj4nlbjAerzAMU4wSYphnSYBLjl8fDz0fZJeRDIIqluX28NrSEwOvx9IRymRA6BJl2U2iUioUGCS6hP9nO5mMGzSJsEwRgkXhzY95lnlJqt0VPjUjy2ejoTTXXbQw+lb//P/JA57eCqHAmIbcaHwJU7mXtn18GSt5XrvAz7t/ISSM8Jc7kU+bf8Yw/3/SNyt5IwBlla6QU4XOZ1/nePBC08UUkPTZSG8yM3eJ/RMC1/l+9VYh9vhZbqm+ZRf3YNpHIrOCJ4OqMfLpCSUnRpzuReZCc7gKqmmCiGEEHupQ0gOjy4RP+YiTdUdvC9nfU4zxan+etINWiywxhglphm7b5h6HAkpLbqMUKJBh3e5yoZq7daXdOA4VnOWGc4xg0ZxnRUiYs4wjYtD2A+bEQkxKQZDigUsDhoXB78fYgO8HVNwt6QY1miwTgsHRY0Ko5QeWR19GhEJm7RwcQbPsUaTeVZYYJ3kASc5xO6ScLrLDks43VJ0Rminm4x7x3m19A0Cnd+V50ptwketH3Arusykd5KcLnEj/PCpHqukR1FK0UzXCXSB07nXOZ57AUfd/0yZtZaNZJGbvU9Yiq5hMYy600z4J2jEqyzH1x87ID8rF5+KWyO0XdrpJg4uM8HZrJq6xxVrIYQQz4d2usnV7ntM+aeY8A/mAd0whCRsBaGfcoUu0QPXk45R3rXnbdPDQRPgcY1lPubmkQo2ysIJJnmBY+TwudX/Pp5jmhxPdgLeYolJ6RERERP3I2yKJSQmwGWCEfJP+LjPqk2XFj2K5CiRJ8WwSYsE0//P9uN2dj0b9fbrdvDe+11PSEm3XRqsTCXuk3C6yw5bOAUo6Ao908JVAa+Vvsm4f+yZnqdn2rzT/J/UkxXO5N9kM15mPbn1TI8Jd4VUledU/nXmci8OQmpsIm6Fl5kPP6aVbuIqjxn/HDmnyO3wCq1045nH8CzKzhiezrEZL2FIqTjjzOVeZDo4s29rf4UQ4nkWm4hrvfdxVXbisOzU8PWD1xMedKlN+Kz7Dle772PJ9tw8HrzIC8Uvy9+VPoOhS0SxPy00JuVGv8L3qPWkz/a8lgZtqhTpEfM+11hUwz0OeWb9tbkvMUeZPOs0uckqp5miQmHYo9sTW42WIhJcNAqNg8JB9//baqep+reeLmFmgdXsCKw7L9N73r9Jm3W1+zMAh03C6S47jOEUyJoA6YCeaXEq9xrnC194qjWem/Ey7zT/J4mNeKH4Fa5136djGrsx/IGiHsFRDo10DV/lOJl/lW7a5HZ4hZSEilNjOjhNO2mwGF29b/fdYXIJqHg1emmbjqkPQvRc7kXK7tgzP35qE7pptu61Z1p00xaeDqi6E1TdiQdWnIUQ4igz1vCz5p/2+w7ckdMlKm6NilPLLt1xAn3wD7SXoxt80v4ruqbFtH+GolNlKbpGK92goCu8Xv5Fqu7EsId5YGzSZplNHJwnXk/6LHpEJBhK5LjFOu9z7am2L1E260pcJKBAjiK5/vWAiIRVGixTp057Typwo7bEy8xRo0KLLp+xxCxjjFPZ/Sc7oGz/vztV03srpncqqbY/gdn0P4tBbxOz7VIpyIqmqh9tt8LunTcXBxeNgzO4b2s8P+JTVlR9ON+QPSLhdJcd1nC6peSM0ko3qDjjvFH+JQrO4//SWQgv8VHrBwQ6z5n85/ik8yNSu3fBsOhUcZRHI1lF4zDln6bsjrIcXWczWd6z591NZaeGpwM240UMhqo7wVyQVVMfFCJjE2Wh07Tops1t17PL2O78+SrU4BeiQlF2aox4k1TdCUbcSfK6LM2ahBBH3sftv+JG7yNeKHyF5eg6rvJxlEtqY9ppfceJVF/l+0E1q65W3NqB+V3ZTZt80v4Ry/ENik6Vk7lXud77kHa6CWRjB4htj7P5z3E6/4ZsdTZkFkuDDiXyGAwfcZPrLN8TIrVV/eAZDMJnkRwFchT6zZa2pBg6hHTokcOn2t/DPiJmhQYr1FmhTldFzzT2og14iRPMMkaPiMvcZpQSs4ztybpPsdOdimq649JgqFBAo/lzPjhSjZsknO6ywx5OAXK6SGIjrLW8XPoas8G5h368sYaLnZ9wvfcBo+40NW+Wy92fwWN0v90NRT3CiDvJcnzjnmB2WHgElL1aP2A2cJXPbHCOvC5nFdB+8OyaFond+YdG45DTRfJOud9FOEArTWoSYhsRmR6xDQl0Dq08QtOmmawNtsPxVY4Rd5KqN5ldSnVVCHHE3Ox9zEftH3IieJmNZIlmurbj/QpNUY+Qcwq4yseQ0kkbtNPNwck9V/mDoJpVWscpOtV9C6zGplzrfcCVzs8Axen867STOovxlft+/NbJ5hF3ktdK33yik81ib2zfdmaNJstsUtgWRPN3TTGOSegQ0qZHr99SSKHwcQjwB5VThaJJhw7Z8cEIxcHazxbdQVhdpfHYa1996/ICxzjJJAbLZyzi4HCKyfs2LhL7z/YruG16/CUfkSoz7CHtCgmnu+wohFPoBx6nRCetMxuc6++Jeu9i9NiEvNv6M9biBeaCF4lMj6X42h6M/vlRccZxlcdGsoTF4OCSc0rkdYlAZwdOWmkSE5HYmNB06Zk2PdN67EZPHjlK7ii+zmFsSjvdHFQNsurqGFV3kpF+YD0oFQMhhHhSa9ECbzf/hJp3DIVmJb7xmJ+pKOgyeV3G0z4WSzdt0Uo3Br9rC7rCTHCWmeAsRae6d19DfIuP2z+kndaZ8E4w4k1xtfvuPScr75bXJSIbgrW8VPwqs8F5+V1+ADTo9LdZcQmJadOjTUhEjMHioPBwyeNTJEcO/4mqlFvrESMiPFxG+lukGCybtAZV1Q1a9+wD61jNGaY5zywazQ2W6RJxjpl9mwotHl9KikazwBo/5cqRaKok4XSXHZVwuqWoq7RNg4Iu37N+pZVs8rPmn9I1LS4UvsSt8PI9Z6PF0/PJM+pNEZkePdumZ9qDhhe7batqkHeKOMqlZzr3VFer7iRj3gwT/nEKev+qBUII8bTa6SY/qv83Al1gwjvBtd57z/yYgSpSdCq42ic0Xer9ZSQVZ7wfVM/s2prV0HT4tP1jbkdXyOsSp/Kvs9C7SCNdfezH0DjkdYm2qTPpn+KV4tfwdW5XxieeXophhU18vH6d092zabIhMfX+Xp4FAqoUUCji/lrVrcpqjTIvcpwcPrfZYJU655i5p6IrDpaIGB+P97nGVbU07OE8Mwmnu+yohVPYvn4l5ELhi5zMvcpKfJP3Wv8LB4fzhS9zqfNjokM6pVbc36C6qgIMZkd1Na/LTPhzTHhzjHkzT9U8SwixNyITktiI1Mb9tyTbZdAmpDYm6V8O3te/TInvXLfZhg55XaboVHe85XXlUKxhjEyPv67/N2Ibca7wOT5u/9WePE9OlSm5VXqmQytdBxQ1b4YZ/yxT/ilc/eRbYBhruNn7mMvdt0ltyqncq0QmZCG6yNMumSnqKh3TxFcBr5a+wbh//KkeRxxuFkuLLm2yNYoVCjs6Fm/Q4gYrnGKK6hHtwHsU9Yjxcfghnxz6Dr4STnfZUQynW0rOCK10k7JTo5muUXbGmM1d4GL7x3tW0RMHh0JRckb7a5Jj6skKhhQHlzFvdhBWc05x2EMVu6SbtrjafZeuaXE6/zpj3sywhyQeILExS+FV5sNPH7shnMbFUdmbq7zskuxSKxew9EybTtrYcfJRoSjoCoUdoXWEolM9MBU5Y1PeavwJm8kSrxS/xoftH+zL36mSM0JOl2mnG3RNC43DhH+CmeAME97cY53I24yX+aj9A5rpOmPuLJPBCT7rvLMrJ4BdfFzt0zMtTuRe5kLhS9Jj4Dln+lOAe8TU6TBOmQn2boq62DshMRb4c96n9xQdoQ+CU3aSGTVG7hdPMHfi1LCHc18STnfRs4ZT2NoTtc2EfwJPBcyHn+ziCMVh4pGj4o6htUszWaNn2kC2n+uEP8e4N8eIO4E6BBUWsVM3bfJZ910WwksAeCogsl0mvRNcKH6JojMy3AGKgUayynzvU25HV0hsTEFXmPJP97d6t1hrMLa/0YG1WJVirCG1SbZb3qByulU9Te55Dk/lyOsygc7hKA+lNLHp0TUtOmljR+jzVEDRqe4IriVnjOI+NuOx1vJR+wfMh5/yYuGrXOn+bCjN8irOBL4OqCerxLaHq3ym/dPMBGcZdafvWRoRmR4XOz9hIbxIoAucyb3BrfAK9XT3u89vNUsqOiO8XvomFXf8mR8zsTGNZJXNZJl6vEIjXaPkjHAsuMCEPychWIg9trXNTYMOP+AjjDqYeeRBTtspXuMUC6wx/nNnmXrh5LCHdF8STnfRboRTyPZEzTtlmun6Lo1MHH6KilMjp4uEtksjWcFi8VTAuHeccX+Oce/4od7s/nlwJ5ReBBTHgvPkdYkbvY/J6SKtdBNDwvHgBc4WPk+g88Me8nMpNhGL0RXme5/SSNfQOEz6p6i4NVajedaTW/syDoUip8rknSKu8nG1R2pTQtPpV1u7g489HrzAC8Wv4Cpvz8d1rfsBn3b+mlO511iJbw62WBkWhWbEnUArh814mZSEnC4y7Z9hJjhL2RljIbzIxc5PSGzEXO4lFHCj9/GgU/BeCFQh26bCRpwvfIFTuVcf+2SitYZWukk9WaGerLCZLNNKN9macpzXZcrOGPVkhdB2cJXPjH+G2eA8VXdCehYIsUcSUlwcrrHEe+rasIfz2LaC6W3WSUh58W98Ge/4wewwLuF0F+1WOBXiUfKqRMGtolCDqgEoRtxJJrzjTAdnZFuDA6STNvis+y63wktshdKcLnGj99GOgKFxKbtjNJIVNA6n869zKv+aVET2gbWWzWSZ+fBTlsKrpCSUnFGm/TOEtstS+NmBWvfvqYCCLuPpPMYmrCe3KTpVXi/9EhW3tmfPuxLd4KfNP2XSO0lqE9aShT17rqfh4lNxa1gsm8nS4CRebENG3ElmgnN81n2H0HT2bUxbVdRRd5rXSt8g75Tv+ZjQdPohdIV6vEw9XR3sY+4qn6o7TskZw1EeoWnRTDZoputYbD+Yu2zGSxhSirrKbO48s/45WQYixB4IiQnw+BmfcVOtDHs4j3R3MJ3zpih+46SE090i4VSInTQOFWcc38nRTZs003UUipngHGfzb0pIHaLtoVShmQ3OE+gCN3sfPTToeCpHwSlTT1YIVIFzhS9wLDgnU7j3QGR63AovsxB+SivdxMFlOjhN0RlhObzBZno4OjMGKo/BktiIC4UvcTL3yq5Xz5rJOn9d/2MKTpkRd5qb4Ue7+vi7LVBFSu4IAOPeMVb2sep9t4IuE5ouSileKn6VvK5QT5YHgbRnWsBW74Exqu44vi6QmJB2WqeRrhLb8KHP4RJQcWskRDSSrNtwzZtlNjjPlH9KTnIJsYtCYlwcfsBHbKr2sIfzQPcEUybA0xJOd5OEUyEerqhH8HWOerKCxTAbnOOMhNR91U4bfNZ9h9vhZRSaY8EFfJ1/ZCi9W16X0cqhnW5SckZ5ofBl6QC6C6y1rCe3me99ynJ0HUNKxRlnKjhFN22yGF195P6WB5Oi6FRpp5uMe8d5tfSNXZsaHpouP6r/V4xNOZN/g086P9qVx32eaFxyTpFOWh/cl9NFqu4EBV0FZekmLRrp6qBz+9PKGmpVaKUb9EwbR3lM+6c5FpxnxJ2Sab9C7IKIhJSU7/MBkbq3n8Cw3QmmGyQkWTAFCae7TcKpEI/Hxafsjm0Lqec5m3/zvlPKxO5op3U+6757VyjNcaP38TM1jCk5YyQ2pGfa1LxZLhS+vKdTNw8Say2GdMd92V6Fqn992/8fccAdmi4L4UUWehfpmEa/gc4Z8k6Z5ega9eTgT896HFljvRauCnhtF7Y0SW3CW43/m0ayxkvFr/FR+y/2dK3mUVdxxqn5x4hNl1a6SSNZu+c1vrvPN4GnfTbjJVIS8rrMbHCO2eA8Bfl7IMRT22qQtEGLv+Jj7AE65/PAYAoSTnebhFMhnoxLQNkdpZ4sY7EcCy5wJv8meefB//DFk2mndT7rvMOt6AoazfHgBTwdcKP30SOn4j2JijNOxzRIbMRscJ7z+S8cyjVl1hoiGxKZLpHpEtoukellt22X0PRv9+9/ugN3hdp2HRh0vR1xJ5n0T9JONlmMrpFyOLcEeBgXD1cH9EyLk7lXuVD44lPtl2yt5f3W97kdXeHl4te41PkJ8aGsKgsXj4o7TkoyOBEz6k5zLDjPdHBGpv0K8RQSElxcLnObj9SNYQ8HeEQwBQmnu03CqRBPxyOg5I6w2T8oOR5c4HT+DQmpz6CdbnKl8w63o8+yUJp7AVf53Ox9vKuhdDuFpuKO00xWAcXJ/Kuczr2Op/09eb6nYa2lka6yFt/qB83uIHhmofP+vyMVGl/l8HUOX+fxVba1ilY6q9TZwQeS3VCwVcPb8Wfhzvu208rF1zmWwqvPTTf0rWY8ZafGG+VffOJtiq503uFy923O5N9kMfzsmaebioMhr8oU3SrttE7XNCk7Y3yu/Lfk74EQT2GrQdJbXOKWGu7flkcGU5BwutsknB5esakBCldtog7g3Pznxb0h9QXO5N84lBW4/ZTahFa6QSNZo5ms0UjXqCeraBzmchfQyme+99G+VZU8AgpuhXqygqdynM1/jrnci+ghNk3qpE1uh5e5FV6hY7J1dQ5uFjT7YdNTAY720GgUKtv70yakNia2IbENCU2X2PZk6uguyekiiY0x1vBS8ascC84/1prDxfAq77a+l3UsTjtspIv7MFqx38rOGJ20gaM8Plf+W4x49zmYFeI+Ehuz0LtEbHtM+6cpuaPDHtLQ9Ihx0fwFH9JU3Ud/wh54rGAKz3c4/d73vse/+3f/jm63y9e+9jV++7d/mx/+8If83u/9HmEY8q1vfYvf+q3feqLHlHB6+Bjr0UnfJDIn+vdYNE1cvYGrNnDURj+wmoc+jthdHgFFd5R6soTqV/1O518npyWkRqbXD6DrgyDaTutsVeJc5VFyxhhxJ1FKM9/7eGhTHXOqhKd9muk6OV1kwpuj5h+j5s7i7kM1NTYhi9FVboWX2UyyzrYj7hQT/nG6aZt2uplVT233kDYZOhoUmoJTpp3WmfZP83Lxa3gP2Re5nqzy4/ofU3bHKDmj/b15xVGlcXGVR2IjXi19nZng7FDGkdiIK513KDpVjgUXpHHTARWaLjd6H3Gj99GO3+slZ5SZ4Cwz/bX8z5uYhIiE7/MBidq7deT389jBFJ7fcHrz5k3+4T/8h/zn//yfqdVq/KN/9I/4jd/4DX7nd36HP/iDP2BmZobf+I3f4Nd//df55je/+diPK+H08SRmhE76Gq5aI+98itrnfyRbYjNJK/kClhw5fRFDHtAY65PaKpZc/yMNjmpsC6sbOKqOUgfzZ32U+CpHwRkZhNS53Iuczr9OoAvDHtqes9bSNU2ayTqNNKuINtN1euZOW/hAFyg7NYpOBUf5RGk3+5x0fccepcO21aW5kaySkqBQVN1Jxr1jjPvHqTi1XduKxtiUlegmt6LLrEQ3sRgKusq0fwpQLMXXaKebu/JcYncVdZWOaRDoAq+XfolRb+qej+mlbX5U/68opTiRe4mLnbeGMFIxDHldpmuanM1/jrP5z+1rONyIF3m/9ed0TROAKf8UrxR/4aEnUcT+6qQNrnU/YCG8iCFlwjvBmDfDSnQdrRwiE9JIs22MRtxJpv2zTAend61r+EFn+n0NVmjw13wK+/TPZyuYLrJB/KhgCoc+nD716vg//dM/5e/8nb/D9PQ0AL//+7/P9evXOXnyJHNzcwB8+9vf5rvf/e4ThVPxcNYquulL9MwLKBISO0lkTlBw38XXt/dxHA6d9FVCcw5Nk4LzMzrpy1i2/4KKcVhDqx5KGYz9/7P333GSZXl95/0557rwkT7Le1/VVdVdbaabbrrHMgM7AoG0MqtdaZEQIBYJPawWJPHogQftCiSkQQhh9QKEkWFgBoaZ0cwwfmjvyntfWelN+Ijrztk/bmRWZXdVl0tf592v7HSRcU9kRkXE957f+R2PQK1Fs7n9/RhLlLFnwuoklqhiTqTOrUC3CKJhXJEmYxW42jrJtdZp1qd2s8bbhi1cbOFgC+e+GqosFaHyqcclanGJajxJNZqkGk8QtTe6FwgyVpEOu5+UzCGEwI8bNFSZqXCY8fDaIt+C91ZXJeoqaYDVZfdiS5dmXOV88y3ON9/CER7dzhq6nXX0OGvvuYxba00pGmXQP89wcJFIB7gixTpvJykry7g/wMXWkXm6dcZcqasyjvDQaF6rfI6t6YNsSR+cKQePdcTb1S8T6YBd2ac5UfuLRR6xsZCaqkpaFrjQfJt6XGJf7tvnvVGS0jHnG29xqXWUtMyxJ/ssA61TjAZXKEfjHMi9nw6nb17HsJw04ipKR2StjgU7eVCOxrncPMpwcLm9j/pWcnYnA83TjIWzmwAl2xgVaaoqpxsvc6bxCl3OGla7W+l3Ny5IRc9ikUhCIvrpYCdrOcP1eT/mPQfTFeC+H5GuXLmC4zj80A/9EENDQ7zwwgts376d3t4bv7S+vj5GRpbHBufLQaSK1OPHiXUHrriKFA0CtRYQ1KJncMQgGfsIlmjM8zg6qUVPoMjjyQtooB4/xrtPITnEdBNrbupbEmAx3g6soHQKX23AZ7rEKMIWU3jyMq68aoLqHAp0kyBqzoTUK60TXGkdn3UZgcQS9kxYtW5+z82f2+33LrawcYSHJzN4MoMt3Hl9Qg2VTy2eotYOovV4ilpUwtc37vcSm7zdSb+7BU+miXWMH9eoqilGgkvLeq1jhM9kdONEVFYW2+VVgslwmOHgEgA5q2MmqHY6q277ArQelxn0zzPkX6Cpqkgs+tyNFOxuSuHYzBl0Y/kItQ86KcG70HybiXCQ/bkXSMksx2rfoBKPszf7LKfrr/DOplLGytdUFVIyx3BwiUa5yqOFD83bco9aNMXR2tepxpOs8bbjCJeT9RcBjYOH1orXKp9lW+YQm1P7H9oyX6014+EAV1snGQ8HAPBEhm53LT3OWrqdtbgydYdrufdjToTXudQ6xmQ4iCUcNqb2YONwzT/DoH/ulj/XUJWZxml5qwtPZqnFUxyvf5OTdYtedz2r3C30uutXZIdoBxufkJ2so6TrjIjSvB3rYQym8ADhNI5j3njjDX7v936PTCbDD//wD5NKpWY9sGit7/mB5r2meRfb1ODilC1oLWipnTTj3QgCsvZbtOIts9Z4WmKKUPdRDj9C2jpNSp6d8zWeyaztLlpqFxKfrP0mzWgHintZd+AS0/OOwNrEZgohAoSAWOWpx08QqNVk7beRwqxjm0s3QmqGbmc1lnQQGrRItgBRKJSOUTom1hGRjvBVg0iH7c/DmS07bkVizQTV5C0983FKZO46xAaqRS2empkNrUVJIL251FZik7OKdDmr8awMAkkQt/BVnZqaWjH7Wr6XuipTbzcnEkg67D5ckSbQLa61TnKldRyJRafTT4+zjm5nLZ5MM+RfYig4P/M76rJXsy61k2ZcYSS4ynBwcTFvljEHavEUaZmjGk3wUvnTdDtrGAkusy19iIvNw8R65W2xY9ydlqrhijT1uMQrpc/waOHDFO2eObt+rTVXWic413gDSzjsyX4bV1snqcVTM5cJuXES5VzjDSbDQR7JvfDQlIgCRCrgun+Oq62TNFQFV6TZnN6Pr5r4qsFocGUmJBasHnrc5IRj0e677+Z4SitGgktcah6lGk/iijRb048SaZ8B/+w9PS5U48mZzugFqxfX8pgKhxkJLmMLhz53E6vdLXQ5axa1md9c83DwCXmMrXxTH6cu5r57/6YHCKZSCIqFNNne5bku+L7DaU9PD08//TRdXV0AfOhDH+ILX/gClnWjLHBsbIy+vnsr1VjKa06rlYVffxapQnu2tBNXXkOKGvXoIHDzP3JBrDuBAEvUacZ78eONZOzDuHJuZq5jnacWPZGMQ1xFiib16NF3jON+pYlIJ2G1vWeEpEKo11AOu8nZb+DI0Tk4jnGzQDcYCi7c1WUtbCzh4AiPlMy1P7eQwkJiIYRAYCGFRKGSzqzKpxpNMq4at9xncnaITQKs1npmNvTm7Ugs4ZCVxZlgJYQgVK2kIY8qU7nL27HSaRSl6Ma/FU9kyNkdWMKhFpU4E7426/I5q5Mt6UfROmYkuMw5s/ZwxWmqGhIbT6QYCS6zxtvOWHCVpqot9tCMRRboJhI7KQEvf5ZHcs+zytt85x+8g1Zc51j9m0yGg/Q46+hw+jldf+W2FRjJSZQ8U+EIL5U+zSO55+lx1z7wOJayelziausU11vniAkpWr3syDxFNRznSvPEzO/KwqbTXoUjPVpxnYvNI1xsHsYWDl3OGnra1TF305wo6bx7liut4zRVjYwssiPzBNVokovNI+950vluVOIxiKdPkvZjCZvR4DKD/jlckaLf3cza1HaK9sqY/XOxiVA8wQ6+pU8Qz+GE0Cbdz/4HmDFVWlOuNGmMVedsTHNp3hoiHTlyhJ/4iZ/gD//wD8lms/zIj/wIzz33HL/xG7/B7/7u77Ju3Tp+8Ad/kO/7vu/jYx/72F1f71INp1prPv+HLzF5/Ti2voAU89sUKZkt3UEz3oMgJG2dwFeb2yH0vUlqgECRxREDZOyjWPfZ9lpr8NU2GvE+BNE9jePB+UgiFFk8eY6Mddx0/F1mLGw8mcYVaWzpYgkbif2uEOurJr5uIICs1UHW6sAVSSVGoFo04xoNVZ7VyMi4V4K81UlK5khZWTyRYTy8NivMGitbh92Pjct4tLTXVxsLb7pR0rb0Y2xJH7zv8toh/wIn6y+htWJb5hAj/mVK8d2dJLexsWWKlqqxObWfbZlD8z7bNt38TaHosHtn+hHMh6R09xpXWieZCK8jkPS7myk43Yz4lynfxWNxWhTI2gUEkmo8MfOcmJFFetolwJ3OamzhzPxMoFoznXdD7VO0e1ntbWXcH2A8GpiX2zrNwqZg9yKEoBSOoIjptFezOb2fHmftsi/jjlFIBJNUGaNMi5AWIT4BLUICQvQ93sSbg2lAxIb7KeVd5g2RHmgrmT/6oz/id37ndwjDcGYrmVdffXVmK5nnn3+ef/pP/+k93fmWajhVSvPJ33qRyfEYUDhiGM+6jCOG57zjbDJL+Tix7sKR17FFmWa8E7iXZjVJqW+si4AmbZ0iJc/d01hjnaYePU6k+3DEMLYcpxnvvsdxPCiNJcrEugNLlMlar2FLs0n8SjMdYh2ZpqVq+Gp+100bhmEYs2VkgYaqsNrdyt7cs/e0XjBUPifrLzEcXKRo9bI6tY3zjTdmGtLdi5zVSS2eomj3cSD3wrxsWVKNJrnun2XQP5+s0W5zRYqi3UfR7pl5/6DdhEMVcN0/y7XWqZnS3XWpHWhgsHVuVr+EeyMoWj14MkOofcrRGIoYgaTT7qfHXUdL1RlonUERt5d1rGMkuLgoJyUdPHJOF42ohK+b5K0uNqf30+9uXtYlv0G7MszFedf3NBqfEP8dobXV/tqNzwOU0HMTTOHhDqfzYamGU4BzJ0f5xudeA1Un0h1oUgiaePIKnnUFSzxYmZTW3DRbOj1LuYlYdz3AtfpYokGsO7FEhYz1No4cv+M4ArWBRnwQjSBtnSSIVxMv4kJsQQOw0NhkrON48rxplmQYhmEYc2h6BrVo9/Jo/kN3teXYRDjI8do38VWDzekD1KIpRsMrDzSOlMgRah8hBHuzz85JuXGkA4b8i1z3z1KOxhBIep31dDj9lMIRhJAoHVOPSzMNfwCyVrEdVHvpsHvJWV13FaZqUYmrrZMM+ueIiSjavaxyN1MOxxkJLz9wGe07OaQoOF1YwqERl6nFJQSSVe5m8nY311tnZvoTLC5Bh92Lr5rt7tF5NqUfYa23fdk2UAqIqNEkJG7/p1EoBAKJwMHGw8bDxcNB3mIPmpAIB/vBgymYcDrXlno4/eYXzxL4MRBji0lAEOkuQGKLMTx5CVdev+fy01jn2rOU3ThiEFtO0Yx3MVezlJIqGgtNBldeJWMdRd5iAbfSLvXoMUK9FktM4FkDNKM96FucEVp4CktUiXURW4yQs9+Y9/JqwzAMw3iYuCKdbCUlUzya/zAFu/uWl4t1xLnGG1xpnSAjC2xKH+BC8805q3xJ+hJkaaoK671d7Mw+dc/hZXqLrOv+GYb9S8REZK0iq9wtxDpkKLj4rvEKBFnZQcrKYQubQCf9E8J2HwSJRcHupmj3UrT7ZpUDa60YCwe42jrBRDg4Ew4LdjfDwaUFbdKXk13k7U4mw6EHmJ2dXwWrB0VMLZ7CFSk2pvayPrV7Qfa+bcZVFJqstTAB7uZZVJ+IiJh2C0o0yTY1PiEu9oMFUzDhdK4tn3B6g6COJaoonUeRRRDgymt48hK2fO+zVLPXdMbJbGm8gZhbPxk8GI3FFDFFBIq0dRJPXpgp9Q3UKurRITQuaXmKUHUQsfSaEkiqKNLJqlr7TVw5uNhDMgzDMIwVI2l6ZxPpiP355+l3N836fiUa52jtG9TjEuu8XYBgwD81L2PJyg7qqkTO6uRA7gPk7I47/oyvmgz657nun6Eel7Gw6fc2k7M6GPOvMRUP39MYLGxydheeSCOEpKVqVKMJVHv2MykH7qUWl2iqKp7MsNbdgUZx3T83q8u88W45qxNL2JSjMSzhsN7bxcb03jnb4khrTUOVmQyHmQqHmYqGaak6EovHCx+j0+mfk+MsGSaczq3lGE5v0FhMIkVMqLsBC2tmz85rSDF77UWss+3Z0p72ms6x9prO+S5r8LFEs72Os0TGOkag1uGrzViiTMo6TyPai2Zu99SaWwFSBCidw5WXyVpHECJa7EEZhmEYxooxXea7PfM4m1P7Ac2l1jHON97CER7bM4e43Dw27+WinkgTE6O0Yk/2adZ429/Vz0RrxXh4nQH/LGPBFTSaot1Ln7ORpqoyElwi1HO3NZ0nsmStIo70UDqmEZdxZYo+dxPlcJSR8Mqcl+6udGmZx5MZStEoAsEabxub0/vJWsV7up6k8/8Uk+EQU1ESSKd3AHBFik5nFVmrI9nHW8c8Vfz4PR9jSTPhdG4t73B6sxa2KKF0GkURiHHlAJ68jC3G8dXW9mypbs+Wrl3wNZ2SChoHTRrQePIcSjuEehPcoh5+KZKUURSQNMjar+HIycUekmEYhmGsGDc3SmqqGqVohD53I3mri4vNw2gW6jWbIGMVaMRlVrtb2ZN9Blu6NOJqu7nROVqqjiM8VntbcWWaUf8ylXhigUYn8GTGdJWfA67IkLWLlMNRFDH97iY2p/ffdhsapRWVaGImiE5Fw0TtExEpmaXTXkXGKhDqkFo0QTkaIyaZ0LCFgyNSvK/4l3DlUp6UuQcmnM6tlRNOp2ksSkgZEKouwEEQoHGxxTDOgs2W3o7CFhM4cgg/3opibkooFpIgKZfRpEjJ06StU3PeQdkwDMMwHlbTM6i2cNiWPsSgfyHZ13IRZGSRhqqQljnSVp7JcBAQdDtr6HbWUAknGQ0v33ZfVWP5sPHI211U4wkiHdBlr2Zz+gCdTj/laGwmiJbC0ZmwmZEFOpx+0laeIPapxRMznYxvRZA0tyrY3TxR+M5l25Rp2mQ4xGh0lUP/01834XSurLxwerOw3UQJHDnc7oDbN3cDfKhppCijdAeWmCRnv/7A3ZMNwzAMw0ikRJZV3mautU7PBIHFkuyDbSEQrPa2IrRgOLg4q8uusXJILIp2L/W43F6/K6A9Y5+zOumw+0lZGVpxnWo8QSWauKcZfQubmIg+dxMHcx9YtvuvNuIqL5f/hIzM8+SH/xbde7Yu9pBu6U7hdHmfHlh2HCKdLLqO4hW2+HrRCZTuQFJD6Rzl8INkrCN48rLZcsYwDMMwHlBL17ncOr7YwwAg0C2EFuStbi43jy5gabGxGBQxU9EwyTY0fWStDtJWjmZcpxKNMeCfgQe4D8RE2MJjNLjM2cZr7Mw+NWdjXyixjjhc/TIAa70d+N7ybcJlwqmxoihyQIQUDRrxIXy1GU9euWVDKsMwDOPeaY056WcsOo2mEr/3vu3GSpNsDVSKRuf8miPt44oUl1vHSVt5NqT2zPkx5ovWmpP1l6jGk+zJPsv54C16xP7FHtZ9M+HUWIFslC4gKaO1TSN+lEa8H0cM41lXcMSwWZNqzDutk5MloepB4+LKISxRXexhGcYDiVQn1egZAKRoYIkGkgZStN9mPjYnAw3DWF4C3cIVaU7VXyElc/S5GxZ7SHflmn+aQf8cm1P7udw8grKXd5doE06NFSvpkpzs7SqFT6S7CaO1CPz2PrRXsETJzAAYc0JriHWBSPcQ6d52KL3R+a8ZP4IlyrjyOq4cMEHVWHa0tqlFTyJQ2GIcRYpIFVGsBqxZlxWE7wirycdW+2NByzz2Goax5AS6iSvSHKl+jSeL33nbDsFLRSkc4XT9FbqdtVSjSRqqir2kt4K8MxNOjRVOENNJrAFibDEGQuOrzfhqG5ao4MorePIqUrQWe7DGMpKE0Q4i3UOoeol0NxoPAEEDR4xiyRpKW8SqiBAKRYpmvJtmvAdJBVcO4MrrWKJiXqgbS149PogiS9Z6k3r8GCCBCEEdixZChAgRI1BoJEq7KJ1u/9tw33FtMZIGKesCnrxoqlkMw1gyAt3EER5vVb7E+4p/ibSVX+wh3ZKvmhyufpWUzFC0e7nYPLzYQ5oTJpwaDxGLSPe218y3sMU4Gpdm/AjNeB+2GG2vTx1ECNOC3phNa0GkO5OZ0Zkw6gAgqeGIYSxZR2mHSHUQ6LUQ3zSb1H7tLahiixoKl5baTUvtQVK9KaiWTVA1lhw/Xk+gNpKSp9vbn8n2d2w0BSIKyX38XRlTIWkhqSJEgCBCihiNINY5GvFBfLWJjHUYRy7MfpSGYRh3EmofC5s3q1/kqcLHcaS32EOaRWnFkepXCbXP7uwznKh9a7GHNGdMODUeUikinZQ9SCpI0UDpPPX4SepxiCuv48kr2GLcBIWHlNaSSHfdNDPaxfRDpqSKIwewRBOlXGI6CfR6iOV7XymgyRPq5CysoIYtqmhcWmoXLbW7HVSnS39NUDUWX6wzNOJHscQEsc7e437YEkUGyNwIrvrm71ZQ2qMavYArr5CxjiGFP4ejNwzDuD8KRSOu8Hb1yzxe+ChSWHf+oQVytvE6U9EwuzJPc7b+6mIPZ06ZcGo89BQFlC4ACosJpIgI1FoCtQlJHVdexZHDWKJqmnysYFrbhLqbSPW01412kqyj01iigievIKWPij0iugjURm7MHt3nMckR6mSvL0GtvQ7VoaV20FK7kNRuCqpmfbSx8LQW1KMnAfDkNRrxwTm9fkX7sVdMEqh1hGoNaesknrxgSn0Nw1hUGoWFw1Q0zPHat3gk9/yS2AN1yL/IldZx1nm7uO6fIdTBYg9pTplwahgzJDHd7fWpAbYYAayZGS0AQRNLVNpv1ZmPTWhdfpR227OiSRiNdQfJxt4KS5Tw5EWkCIlViphufLUF1Pw9KWlyRLOCag2waanttNTO9omS6zhyCFtMLOsX7kqnkllp1UWku4h1Dil8BD5StJIyUNFC0EIKf+ZjQWgC+gJrxruJdDcZ620a8SPzdBRJrLsAHymaNOIDN5X6mq1CDMNYPDEhjvAYCi6QbubZnjm0qOOpRVOcqH2Lot1LpEKq8eSijmc+mHBqGLfkEul+AAR1bFFCiBiNhdIZfLWZm//5JKG1elNwrbRnWlfW2azlTOlUO4gm60VjXWx/J8YWk3jyHFJExCpDpLvx9TaSsLrwZgfVentG1aKlttFSOxAEOHIERwzjyOElfT/T2iLSHTNhNNZd7TJPgBhLlHHEGBoHhUek8ig83tn9dfrykhZC+O0A67cDbOumUFtDEJgQOwdC1UNL7cIVV/DVJub/JYNHrL12qa9LNXoeV15tl/qahnWGYSyOUPu4Is3F5mHSMs+61I7FGYcKeLv6ZSzh0O9u5mzjtUUZx3wz4dQw7kCTJdTZm9ZJaQQNLCYRIkSKCI1NrNP4aiO0m+QACFqzZlml8NE6+c70JW71Xt/h+1I0cMQIQizvvazmS7LHaJZI9RDqHiLVgyLX/m6ILSZIWcMINJHKELe3f1mKNFkinazxEzTanX01oeolYD3EGktM4sphHDG0qOtUk997nkh1EuluIt3ZPgmQlD9L6lhiAk9eRQOxyhDrruR2IEju+QGSWnsWNQRihFAINCSbmKC1g9IZIt3V7pA8+wYLAqSoYolkBlq2S6YtUTPNzu6S0g716AkkdYTwiVXnwh17ptR3or3EYjUZU+prGMYimt5i5mT9L0jJDD3uugU9vtaaY7Vv0FRV9uae5UTtxQU9/kIy4dQw7plIAgPZd3Sn1MkMKlNJV0qhSMrVMvhqAzeH1rkR4YjhdqnnMFJEc3z9d6a0Q6R7UdpFimZ7T8Pmgo1FaRvVbtAS62zysc4S6SKaNAACH1uM48prgCCK88S6myhetSBjnEuaDJGebiwTYzOGlCGxTtOM99JkL4JmElTlcPsExvyFsaQ0+uby3M6ZLUMEIZaYwpMX2jPSHjFFQr2GML5dUwmBxpvZkie+RQOdm46ezJpSQhIgREQSZDUg2zPlvQRsfMcRmjdC66wAWzcne9q0hnr0GIoUWevt9rYxC00S627AxzKlvoZhLAFJQE1xuPpVnix+FwW7e8GOfbF5hLHwKtvTj3Ou/iaalft8ZcKpYcwZkYSH6a6UN4VWaGJRapem6ZnL3/j/zbOlzPr+rOufnnUVAiF0MjMYr4NY4YhRHHkdVw7NW7fLpINtN6HqI9R9xLrzFuMECJE020G1cdPHSYC1RLMdJu50PIEig9I3hc+ZIJqZCTHTBCFS1HHEGFLW0VoSqyKR7iOM187NL2HJsIjoJXl+0u2u03VA4qt17dLzuB3Mh9pNvep3vFatQeOidApNCqVTqPb7mc/bH+uZEy5J0yhHXscSrfberjli3UWk++bp9ks0GTSZ6V/B9FBmCEIkU+2y37D9TZtYpwjUmnfcf3R7ZreGFPWZmdtknWuYlAoTIkSAJASiFVs6HKhNhHodaXmCRryHxSpvT9xc6us8FKW+SrvJXtzxJjQWUrTve+37X/Jx0L5/BrPun9OXTU7ULPYtMYyVJ9AtbOHyVvVLvK/wl0hZ99K9/P6MBwOcb77JKnczY8E1fN2Y92MuJhNODWPeCSBDTObGTNCDmgm/MRZjSBkRqwJhfIhGrNuBZBBHDmKJ+38Q0xpi3Umo+whVH5HuJlkLqNqzYmeRIiZSBQBEe+ZKI5PSSzxCtQpNineXXobvmG1tAvqmGdBMe23izT8Xt4NuHUeWZgKH1hZKu2jSKJ1tl7s+TK/MxE1dpyFp6DWKEBGxyicdVmPa29QMYYlJNB5Ke+3fmYcihW4H0Vt3IY5mGhUlpcPj7RfHikilUXQSqE0sbpCZTeOg6US9a/9NjcDHYgIhWkhiEKCxUe01x/qOlQ76psB6czi4ERgkPo4cad+3l4dY56jHB7DFGKHqmqlAWGxJqa/GYnKm1DdtnSIlz6+YUt9IddBS2wjUOsDCFqNI0UJrO7lv6jRaF1C43LkSJ74pxAZ48hKedXUBboVhrHyRDlBatfdA/Z+wpTtvx2rGVY7Wvk7O6sQVaYbjS/N2rKXChFPDWNYsYnqJFSRb4UwiRQulczTiAxAfwBJT7e1IBpFU3/Ns+vSawVD1tsNo70yZpiXKePISUgTEKk2ke/H1Dt4VRm4RApLSSx8holsG2Eh1zKwdFCSzrUlH2kGEUMllld2excsmIbndsMq4FTeZsdQw/YJeilZ7P9WtwHQzh/bfR/gz5a5ChDOlwFrLmRfGSZhNEesimm7QSyeE3juBJkVM6hYzrhpJA0EFSdT+XcSATta+CkWSZEX7PmyhsZKGTjpNpN12sG2XLse0TxYlWwIt5dk+rSW16EkEMY4cpjlv3XnvlyAm6epriQbNeD9Bu9TXFmPLcqZQa0Gg1tFSW9tlzCGevIwlm/jxegJ1c+VB3K4GaCKotE+MKGbdP9Ht5dvT989k6UM9foJYF0lbxxb996S0iyAyZfTLgNaCRnyAWOfIWEewZXWxh7SEaGpxicO1r/JY/iNI8WBby91KrCPern4FjWZDag8n6yt3nenNTDg1jBVDEtPVnp3VSEpYsoHSaZrxPprxvvbMWRJULTGFENNdbPtmZkenZ0okDRwxmFyHconowVdbubeZMdEu/0zdtvRydoBtoYWH0rnkReiyDkBLhbjpfpE0VbLFFIIYpW20SLdnUPOAvM36zoeJQJEFsrPvs+/8+F00yexyiBRV0MmsspCKWGVoxAdoxAfaQXUAV15fckG1Ge8l1p1krTepz/F+pnPr5lJfm2r07e0ZwhqWqLcbYNVmPhf4ix7I3knpFK14M77agiaFpEZaHkfpFIHagFa3momZPhGSuvGlu7x/Sqq01A5iXSBrv7ooPQoA/HgD9fhRJE1y9ivYsrIo4zDuTGtBPX6CQK0HQirRh0jJs6St06axHKCIsXGZCK9ztPZ1tqQPkLe65mwfVK01J+svUY0n2JN9ljP1V+bkepcDE06NOZFvjbFj/DVqbhcXuh8jsrw7/5AxjwSKDpTqYHo9oiXqaBxaagcttQtJA4ja5XLTjYPGsGXygi9SnQR6Pdy2ec3cjfWWAdaYF5oMoc7c/AVjTgjAQeHAzb/fmTXBU1iyRayyNOKDNwXV60siqIaqj5bagScv0lLbuPVWPkvLdKmvLcbaFSMekepEsY7ZJ9HCduOrm4Nrsr5Y0Fqw4Ko1RLoLX20jUGsBgSNGcOQIQdxHU+1lvsriFXkEdULdRyV8P3nnpbtagz5XtLaoxwcJ1CYsMYHSWSrR+8lab5ty4yUoqaJ4ilCvIS2PE6g1IKCldhGo9WTst3HlyGIPc9FFBLgixUhwiZHgEmmZp9/dRL+7kaLd90BBdcA/zaB/jk2p/VxuHiFmcU4oLQYTTo0HIlXElsm32FA6SShdiq0xeutXONP7PsayG1lyp6sfSrPXIwpq2KKKbq8ddeUgAFFcJNT9SYMlwzDmkEDRiWoHVYsppGwSq9w7gur0jOr8NDS7HaU9atETSFEhWfddvOPPLB0i2QZq5iRLlDQHI2m6lszwyKSztCqiWMPsNdXRTaG1jhR1pGi0v9aYkxkirSWBWt8u3e1EEOLJi0jh48cbCOMDD3yMuxoHWcBHk6YSvp+c/cqCdD6OVIFa9BSKPCl5mkjl0XQiRZ16/ASR7iFjHTZlvktEEkyfJtSrSMujBHojMUWSivHkhEYtehZHDJC1jyz6ibXFFugWKZEhZ3cR64grrRNcbh3DExn63I30e5votFfdU9lvKRzlVP0Vup211KJJGurhKqc24dS4b52NQXaNvkQmqnI9v53QStFfu4AWNvuHv8ZYZj1net+H7+TufGXGgtHkCPX030Qtyy1VDGP5EsR0tvcNnQ6q0zOqj9KID2KLsZtmVOc3qCbbxhxC45CxTlCPFmPbmLlkoyii2i+mZ1cGxAhqWDTbe1THaARKu8Q6T6hX8c4Z42RNdr3dZbyRdB+/6fP36joe6zR+vAVfbUbjIUWFtDx+yz2xF46HJkISU42eI2O9Tcq6PC9H0jrp/FyPDyCIyFpv0ox3t8vmQek8kgq+2kykO8nZryzobK7xblpbVKNniHQvaesIfrwFRf7G98mi0VhiilCvphT2k7FOtPcgXsSBL7KWbtAKk+aTnsiSszvQaK77Z7nmn8IRHr3uBvrdTXQ7a7DE7eOXr5ocrn2FlMzQYfdxofn2Qt2MJcOE02UqFVbpr16k5nUxkVm3oDOUduyzbeJ11lbO0XDynOr7NjZMHSMbJmtHNFB1u+hqDvK+q5/mQvchBoq7YB4WixsPyvxNDGPxTAdVmAmqokWsc+8IqgPtZkrhnI/AV1sJ9WrS8iiNaB9Lqdvy3LPQFIgo3Ca4NrBotLsvxyA1tDuBx7pAqFdz6/DaaHcdbwdWfAK1llCvAcARQzhyDD9eRVPPX+nu3bNRWEhqNOJDxLpAxjo2p12Ptbapx48SqA3YYhRHjrb3y539nKMoIGigdIZK+EGy9uu4cmjOxmHcPa1tqtG3EeluMtZhWvF2FLeaXBDtbeSm9yA+iK82krXewpalBR710uPrOn6YnGRx8Oiw+5DCYjS4zKB/DguHXncdfe4met112OLG+nKlFUeqXyNUPrtzT3Oi9nA0QHonE06XE63paA6zvnyS3vq1pCsfUHM7uNKxj5H8FrSY33VCvbUr7Bx7GSducaW4F0tH7Bp9cdZTrQDywSShcGg5OXaOv8qq6gVO930bNa9rXsdnGIaxPLWD6kyH5ZuD6mM04v148hqevDhnLwAjVaARP4IjhgnUqnftG/xwsdDkicjfCK2zKnqnw2sy65rsI6oA66aZ136mX1YJAjx5DiFCgngjYbxmQW/NnQkUeSRVfLUdpfNz1igpUsV2GW+OlDxFpIo04323vbwmA0RI0aIWPUNKniFtnVgxWwQtB8kews8S6w4y1lu04t3trdzey82NydJUog/gyQukrROL1nBrqQnxmYySky0WNl32KqR0mAyHGA4uIZD0OGvpdzfR627gUvMIU9EQuzLv42z9NR7WphBCa72kbvnERA2lltSQZpw7Oco3v3iWwF/YLmVSRfTXLrK+dCoJfdJjsLCdWFp0NoZx4xbZsEzLznC1uJfB4k5iObflQm7UYMfYq/TXL1NzO7nWsZvNk0dIRXcuwak5BdzYx1EBVzr2canrIEqa8yKGYRh3prFI9vQNdRdgY4kJUvIirhy473V6Wksq0QdR2iVlnX/P8GDcjSS8JvswJ+E1mTld+s91gjqaNFLUydsv3ndprdbgq8004gMIAjLWiVllvHdxDUnQoYgtxsjZrz306xkXgtIu1eg5Yp0nax2mEe+5j/2NFZYotddUt8jaR3DE9Ye61Pe9CCRFqxdHelTjCVqqjkCg0azzdlKOxqnGE/d9/bab4unv/Nus2bJ3Dkc9d6QUdHfffsmfCaf3YKHDqRs1WFc+xdryGVzlU3M7GMpvIx2UWF27hKXbexGSlNEiBAV/glA6XC/u5lpxD4H9gBuoa83q6nm2j7+G1BFXOh4hHVRYVb90T4VJMZK610nBn6Bh5znT9zSTmbUPNjbDMIyHio8tSiidaXdf9fHkFTzrEpao3dM11aOD+GorGesNGrcotzQeNj6i/ayeNEoau6efTsp4HyNQ67HFCI4coxnv4X7uV0lY9hBE5OzX7nksxt1T2msH0xxZ620a8b7ZWxXdI0Ej2aaMPI4YJmMfNuuI70LR7sEVaRyZQqmY4fDiA12fCadzbKmGU601537p1xgcqDDl9FBO9dF08vOy1rPQGmN96SR9tUsINOOZdUxl1tJTu0Jna/g9Q2HNKRBZKYqtUZSQDOe3caVjH0333rsvpsIqu0Zfors5SCnVx0hhC5vH38ZV99+go2llQUA6qjOU38q5nicJrft/IDQMw3j4JGW/QsREuhuQ2GKElHURRwzdsRwyUKupRc/gyXOEatWshifGwyxG4qNIkbEOk7Iu3dVPRaqjXcabIS1PE6pOIlY/4FgCJBGKNGnrBCl5xszCzTGl01TC51Ck28F0/xyV9msk5fZ6VUHaOkVKnjVl2gtouYfTpV9vslRoDZUp+koXWaNPARBIj3Kqj3Kql0qqj0qq577LaYVW9NUus750kqI/RiQdrhd34dtp1lTO0Ts+cFfXkwsrEFZo2llaTp5V1fOsqZxlLLuRy52PUE313sVtVawvn2LrxFto4Fz343Q2htg5+uAbAKfjOhqoeN30Vy/SXR/gXM+TDOe3mm1nDMMw7oogpqu9tUMDS1SJdYFa9DSCBinrEp68fMuSSKVT1KNDWKKE0o4JpsZNLBTpdqOkx9qNko7eNlQkZbxbacSPIPHJWm+1S0LvtFbxbrgoHKSo0oz3Ealusvbr89IU7GEU6wzV8NtRuO1gegCNe+cfvCvJPuvJ/sLJ38+PN5C1316QrYuM5c/MnN6DymuvcO13/wtVkSG0UkityARlsmEZAI2g5nZSTvW23/poOoX3DF1O3GJt+Qxry6dJxQ0aTp7Bwna8qMHqygVs/WAPxL70qHud5P0JHBUylV7F5Y5HkpLaW4wr60+xe/RFiv4YE+m1TGTXsGXi8AOP41YC6RFaKbJhmcn0ak73PkPTLcz5cQzDMFY+hcUkQoj2bKrCEYOkrIvYYgwhkjBRjZ4j0l1k7aPUo0dZ/M6xxlIkqaHIYYsRcvar7wqFSjvUo0OEei2OGMaWkzTjXcxHeXgyljSSFjn7VWw5NefHWEq0FoCYt31fY52jGj6HxiZjHaEeH2Q+tzWSVNHYaNK48jIZ6/iC7+X8sFnuM6cmnN6DymuvMPZ7v0PcvHE2WiGouR34dhYlHdyoQT6YxFbJA3kovVlhNZlddcn5k6wvnaS/dhFLx0ymVzOeXU9nfZCe5vWZTrxzJcKikuolG5bx4iZVt5MrnY8wmtuMFhKhYzZNHWXT5FEi6XCp6yB91Ut0+qNzOo5bqTqdpKMaAsWlzoNc7dyHNtvOGIZh3BdBFVs0iHQnGhdJlZR1EaVdWmo3aesIrXj3HM6UGCuRoIEm1W6U9NLMuuZIdbbLeNOk5SkC1UNM/zyPxkeg0HhkrCN48uKKLLYKVR/16BAKD0eM4spBHDmIFMGcXH+s81TC5wBJxj5GPTrIwhRRJnujxroDQUzaOtneG3Vpvt5f7kw4nWPLLZzeSiA9am4noe0htCYTlMnNzK5Cy86RjmrEwmY4t4WmV2BV5Ty5oDTvt0EhqKR6cOMWmbBK085xvbCDVbWL5IISw7lN1N1ONk0dnWm4tBAiLBpekYI/SdXt5EL3ISYza01INQzDuG8RlpgC7Pa+hODI6yjlEdOzuEMzlonZjZJiXWyX8bZIWSdpxnvvo7Pr/dJIUUXpAq68RtZ6EyEWdveE+aK1RSPeh6+2Ialiywki1dfezkVji/GZoGqJxn0dI1JFqtGzAKTtEzSig7xz397518ISLWLdgSUqZKzDpuHVPFju4dSsOZ0HrvLpag3PfK4QVJ0OWk4OZTlYccBQfiuWCllTvYBTXbjyBommozWWrPt0uxFCs23yLVpWhlO9z7CufJpVtcsLNp5pNjEFf5KGnceNWxwc+jK+lWYov5WhwnYabseCj8kwDGN5s4l10mdAUsaWFSQtQkyndONueeh2o6Rq9O0AOGIIS5ZoxIdY2LJwgdIFJFUCtY5IFdtlvpUFHMPcS2ain0CRx5PnAYGvtpA0PptACp9YZ2nEByA+gCVKOGIQVw5iifJdzSBHqoNq9ByCiLR9ql3Svxgn/1PEOtXeG9WmGn07jhggYx/FEs1FGI+xFJlwugAkmnxYIh+WZr7W07i+eAMieTopBMkeSlW3k5rbxc6xl5GLvOFvJqqigMlUPwLYUDrBptJxyl4vQ4VtDOe2EFumFM0wDONeKIoE6t67thvGdKMkS0zhiusEqo8w3r1oo0m2UWqh8ahEHyQlz5C2Ts/bGs35orWgGe+mpXYiaZG13qAZ77ypSZkgpptYQ9IBt4SUDbRO0VK7aak9SOo4Mgmqtpi4ZZlsqLqoRc8iCEjbZxYxmN6gKJCU+k4Q6lWUw1WkrTPtrr7L6+9ozD0TTg3ywRT5YOk0GJBAV2sEgKaVoe52kI7q7Bp7me3jrzGW3chgYTtT6dWmw69hGIZhzDtBrDtp6g6WQhOtZC9OhSXKtNRuArWOrP3WsukGG+s8tegJYt2JK64iRZP6e+43nHTAVaqj/VkVS1QBB19twVfbEfg4cigp/xUjCKEIVQ/V6NvaZdjnl1gTNEGsu4EAS9Roxnvx401k7CPtLbEWe3zGYjHh1FjS0nGDdLMxs/1MaKXobgywqnaRlp1lKL+NocK2pCuyYRiGYRjzaCklBkmsO5HU0VhUo+fx5CXS1rElu+VMsv3ONhrxPgQRGestfLWFQG24t+shT6SnZ1ib2O3y3lCtIVCbgAhHjBLqPqRo4MlLSVnwkvr7TXOJdReSpOFWLXoGRwyTsY/MNOEyHi4mnBrLggAKflKGHGExmVqFRLFp6gibp44wlepnqLCd0dym+95r1jAMwzCM5UWRZbobrK82EajVZO3DOOL6kpp9i3WGenSISPfhiCFsOdEOjA/alChNpNNJx00CbDGKEDGR6sQSZVw5SDPez9IMpjcockz/HUPdTTn8MCl5rl2yHS328IwFZMKpsezYxDMNp5p2lrrbQSaosGf0L9gx9gqjuU0MFbZTSvWbsl/DMAzjoWTFAcXWKAV/gvHsOmpe92IPaR4lZceCJoKIWvQ+HDFIxj686I12tIZAbaQeHwAgbR0hiNfQjPfNw9FcIt3XDqox4LePs1xeC4l2Z/EIS5RoqZ34agMZ6xiuvGZe0j0kTDg1lrV0VCcd1dFA2e0msj36apdZUz1Pw85ztXMv1ws7wWxJYxgrm9ZIHaOkeVozHk5eWKOjNUqxNUJHc5RcMDkTSTZNHeHYqvczkV2/qGOcb5o0Go0lSoS6j3L4YTLWifaemgs/HqU96tGjhHptsh2MNUAj2gssRIWXRaw7FuA488Em1l0I6ggRU4+fxFdbyFiHsWV5sQdnzDPzLG6sCAIoBhMQQCgcJlOrsHXIrrFXWFM5z+nep6mmzL5+hrHSSBWyqnqJtZVT5P1JxrMbuFbcw1R6lamcMFYurckGU3S0RulojlBsjZCO6gBEwqaS6uVqxz5iaZP1SxT8MQ4MfYXTvU8zWNy5yIOfb6IdygIs0aQRH8RX68laby3otjOBWk09egyNQ1qeIFSd7b1FjbulyaI1WJSIdZ5K9MH2uuKTCHzzEL9CmXBqrDiODmfKfmtOkVRY44mBzzJQ3MWF7seIpdmKxjCWu0xQSvZlrl7AUQE1t4Oh/DZ66tforV+l6nZyrWMPI7ktZjbVWPakiij44xSbo3S0Rii2RnFUAIBvpSml+hjObwOtyAZliv4YXc2hWdfRsHPsHnsJL2pwqevgQ3DyxiXWbntPzVx725mzpK1T87pdidI2jfgAgdqEJUp41kma0V403rwdc6WL6QDimXXFyT6wyf67QrTa730kLaTwEfhI0Wq/902QXWbMM7axouXCMgpBxetmXfkUfbXLnO19itHspofgidkwVhahFb31K6wrn6azOYxCMprbQNXrpbt+ldXV8ygkk6l+vLjFntEX2TbxJtcLOxko7iKwM4t9Ewzj7mhFV2OQruYQxdYIhdYEkiRQ1Z0io9mN+E4WK/bJ+1P0NK5j1a+851VmohoNO8eWqcN4cZ0zvc+gH4IlL8memnF725ldBGrtvG07E6oe6tHjKDKk5Bli7dFYUtu3LGdWu9S32f7bxWhstLZROoXSHe0TALe6T+ubguqNAGuLCTzr+gLfDuNOTDg1VjyJpuiP07LSKOnwyPDXGc+s5Wzv+8wWNIaxDHhRnbXlM6ypnMWLmzTtLBc7DyB1zOrqBVbVLs9c1kLR1RpJtp9yu1DSZtPUETZOHWMkt4lrHXtNif8DcuIWofTMCb55YMc+qyvnWF8+RTqqoZBUvW4GiruILBcvrFPwJ1hdPY9Mut7ck0xUo2lnWVs5hxc1ObbqBdRD0eHeam87U4MH2HZGa9C4SRgijdYpFKl2OMoS6tVIamSsN2nFe1CYE2JzTZMmUO9cO62S8EkFQYAUIQIFIkagAYHGSoIsLpHKokjhsx3FCdLW6cW4KcZtmHBqPDRScRMdN6m4XXQ0R3jq6p9wufMAVzr3ocWDtnI3DGNOaU1Xc5C15dP01q8BmonMWq5k1tDRGGbT1NH3fHEugEIwCUDTztF08/TWr7K6dpFSqo9rHXsYy26cl5kjSwVk/RKpuE7V7abp5FdGkNOaTVNH2Tr5FlW3i2sdu03Z9BzJBiXWlU6yunoBS0cz99Fca4LO1gjryyfnbO4tHdVpWRm6G9c5dP0LHF79IUI7PUfXfmdCx2jEojQqvLFdySS+2kigVpOxjuDKATTebUJn8rFuf+/WM3MhkhaePA9oGvEhzGzpQpLtZljJ/Tiefmq45VOERhAiaCCQNOO9gCBtnVqgsRp3IrTW937qbR5NTNRQakkNaUbltVcY+73fIW62FnsoxgOKsGl4RQr+BHWnyOnepyllVi/2sAzjoWfHPmsq51hbOU0mrBJIj6HCNiLpsrp6gUx4/w1NImFT8XpIRzXSUY2WneVacTeDhR1E1r2vBxM6JhuUyQZT5PwpcsEU2aBEOpq9cXzLylBKr2Kq/dZ0CssurAods3v0RVZXLzCeWUs6rJENywTSY7CYlE37dnaxh7m8aEVPY4B1pVN0NweJhWQ0t5laqov+ysWZvb3nSyA9LB3h2xkOr/4ITXd+K4ksFbBx6jjrSycIrRQXuw4ykt+6aKXFgiZCRCidh/bs2rsvE7TXNDbb6xpDhIgB0FqitY3GuSnYpjDzPsuLoIEmQ0qebK9HXuwRPTjbTfH0d/5t1mzZu9hDuSUpBd3dudt+34TTe2DC6crTsPNIYlJRg6H8Vs51P7GgZ5ANwwC0puCPs7Z8mv7aJSwdU/J6GctvItuapL9+GUvHc3c4oOL1JLOr/jixsBnKb+Vaxx4absctfkAlYSxIAmiuHUbTYWVm9lYhabgFam4nvp1FSYkXNLB1iNQxeX8CL072W/St9Kyw2nCKSzqsOnGLR4a+SmdrhEud+yk2R+lsDVN1u1DSodgaRQNjuY1cK+6hnOpb0rdnsdmxz+rqedaVTpGJqvhWhuuF7e0y9fN48cK9xoiEjRYSLSSHV3+Iaqp3zo8hdMy68mk2TR7BVT6j2Q1kwiq5YIq6U+Ri16OM5jYt0n1GY1HCliVEu7xXaQu0hdIOiNRNodNUWK1UNwLqqaQT8DJ/+DLhdI6ZcGostOSFajd5f4pY2pzvfpzBwg7z4sow5pnQiv7aJdaXTlDwJ4iEzUhuCw23QH/t0rzPHAHU7QKBk6XQGsHSionMWobyW3GjZjuIlsgGUzPhWAMtO0fN7aTp5IktByvy8aIGubA8K7DeLMai4nUT2imkVuT8SVJxA4DASjGVXkUplYTVutuxZB5/MkGZA0N/jhfVOdf9FOvLJ8i+Y/a6aWVpuAUK/gSOCqh43Vwr7mY0t3n+Sn61JhNWKLZGyPlTDOe3Lvm1xJmgxPrSKVZVz2O3S3dHc5soNEfpq1+daXi00GIkkeVhq5Bjq16Yu71QtWJV9QJbJt8mHdWZTK9mNLeR9aVTZMIyNbcLS4dkwipVt4sL3Y8xkVm3ZO77xsPlRkA9Tdo6sWzvhn68kVis4sPfs5912x5Z7OHckgmnc8iE05UtkB6BnSEXTFFK9XKm9xlqXtdiD8tYjrQmG5TobgzQ3RjAiQOG81sYzm8jMDPzSBWypnKODaXjpKM6dafIUH4rXtxgVfXizBYZC8mXHjWvi1wwNTNz5Vsp6m4nDadAaKWQKsKLG2TCCtmg9ECzuXG70Y1vp7G0IhdMkoraYVV6MzOrpfQqam7norxg72wM8sjw19BCcq77cbaPv46r/NtePsKikurBi5tkwwqBlZqzTslCK/L+BMXmCB2tETpao7jtv5NGoBGc63mSgeKupRVuble663XRX71EwZ/7jrH3IznpksWLGpzue4ahwo4HuDJNT+MaWyfeJBeUqHjdDBZ2sLpyjuI7bu/0yWE3bpGO6pRSvVzseoypzJoHuj2GcT8kjZlOy2nr+JJ6KLkTraEZ76WlduGIQZ776CPsPLBvsYd1SyacziETTh8OVacDL25iq4BrHXu41PUo8RLoZih0TM6fTGaYpMtYdoNpRLKEWCqkszlEd32AnsYAqagOQM3tIBYORX8MhWA8u4HBwg4mMmsWpSHIYnLiFutKp1hfPoWj/GTmKL+ZrtoA3c3rS6J9iEIwmV5NYKXJRFWyQWlBwnKMpOZ14dsZBJqcP0m6fR8KrBRXOvYxUNy9YP/m15TPsnPsJZpOgWsdu9kx9tpdz+xNBw4trHbJr2A0t5lrHbupeL13FR4tFVJsjd60r+cYlo6ApMFVKd1P08ljqZC8P5HMdIdlRnKbONX3bYu+n/WtSncHC9uBmDWVCzMl3kvJdFVAOqpxsetRLnUeuOegX2yOsG3iDTpaozScAteKu+mpX6OrOfie/76nS+1TUR0vbjKZXs2F7seopPoe5CYZxj1bjgFVa0k9foJArcOVlwCbD33vx9i4rXuxh3ZLJpzOIRNOHx7TLxSL/jgtO8NAYTd1t0jDLdJ0CvPfwKG9kXrBH6fQGqfgj5PzJ2e9OIykw0huM4P57VRSd/eCz5hDWpMJy3TXB+huXKezOYxEEQmbycwaal43Uvl0NUfI+xM0nCJNJ0ehNY6rfFp2lsH8doYK22k5t3+QXglSYZUNpROsqZzF0jFjmXVMZdbSX7tIsTW22MNbkhSSqtuF72Rw4xYdrVGado7z3YcYzW2ev3/vWrNt4g02lo4zkV5DJdXDpqmj933ioGnnaLh5is1xbB1S9noY6NjDSG7TrC7pbtSgozU6MzOaPN5pNIKa20Ep1UdoZ3CiBoVgkvxN+35OqzsFMmGVppPn2Kr3L0rlixO32Dx5mNWVc0nprteblO62xha1dPdeNOwcmajG9cIOzvQ+fVfPd1l/km0Tb9LTGMC30lzt2Eu+NU5//fI93XcUUPF6yYQVXOUzllnPxe5HqXlL80X2SiN0TCaokA1LZIMbb76V4Uzv+2i6xcUe4oK4UeJ7lrR1bEm/vFLaoxo9Tay7SMsTBGo1ltfHhz6+24TTuWLCqbGUtKwMsXTIhuWZrykETadAwy1Qd5LA2nCK1N0ikZW694NoTSqqUWiNzYTRvD+B3Z4liIRD1eum6nUSWR5O1CQbVNBC0NEaxdJRUhZZ2MZQftsDl88ZtydVRGdziJ7GAN31gZmurHWnyERmDaGdItsq0dUcvG35Y4ykkupFoCm2RgGYzKxlsLCDsez6FbWtUc6fZOPUMfpqlwDBSLsT6Zry2Vn/pow7qzsFBJAJK5S9Xs71PEE53T+nx5AqZO/IN+mrX2WgsBMnatHfuDIn1x1hU0l1k4obZMIqvpVO1vbGLTpaI2TCKgCxsKh4PVRSPcTSIRVWKbTGyYbluwo5vpVCkMy8nu15XzJbuQCvLIVWrC2fZsvk21gqZCS/hZrXRV/10rtKWZeDpp0lHdUZy6zn+Krnb7sXaiqssmXybVZVLxBJl2sde3CjJmsq5x4oiMdIqqmemcqFkdwmLnY9euuGZcY9u9FpfHYIvXnN/PRMet3toNgaReqYC92HuFbc/VBU/UwHVE+eI2MdXZIBNdZ5quG3ofDIWEdoxTtR5HA9y4TTuWTCqbHUaKDidhHYGSLpIoTAjn1SYY1MWJn1BBxIj4bbQd0ttANrBw2nQMvJz5x9dqPGrBnRfHsmDUCJZLak6nUT2GmsOCATVCgEE7csAwuESzXVjaMCCv5EUpKYWctQYfv8Bp2ZZiSjFFujWCqi0b7NDadAwy0uiVLouZAOykkYbVynozmMpWNiYTOZXkU11YsVB3S0Rij4E+3Nvu9ew87RcIvk/ClScYPASjGU38ZgYfvyfRGmNR3NYTaWjtHTuE4kbAYLOwgtl7XlszNNgIx7p4Gq24UXN/DiFiPZTVzoOZRsTfOAvKjO/qGvkPcnOd/1GH31K/MSqqbLN7WQdLRGCaRHOd03s64245cptkYf6H6igKZTJBuWGcpv5XTv07cNV3OhozHEzvFXyAUlJtOrGMttYuPk0WV/X29ZGby4ScXr5siaDxPedPLViZpsnjrC2vIZtBAMFHahhWRd+dTMidW5EGFRTfWQ9yewdMxQfiuXug7ScvJzdoyVTKqITPjuEJoJqzPPVxpB00lCaNNuN3mLA9woWTueCcugNQ23SD6YopTq42Tfsw/FLOqNgHqejHVkSQXUUPVSi94HKDLWMRrxATTJcgYTTueYCafGcqEQNO0cLSdHJF2UtJE6xouShinuTdsBKGSyPkqHM01PNIK6W6Ti9eDbWaSOk1kCf4JUVLvnMrqaU6Tl5Mj7U3hxg0B6jOS3MFjY/sAlUZYKKbTGKLbGKLZGKbTGZgJ1KF0i6ZCK6rPG3LIyNwKreyO0tuzcou1rd0dak46qM+vcOptDMzM6DafARGYNgZUhEySzo3O15YNCUPF60VJSbI4g0ZRS/Vwv7GA0t2l5rC3Wit76VTZOHaPoj880w5EqYk313KI0OVqpFIJKqoe8P4nQmmsdu7nceeC+9moFyLfG2T/0FWwVcLb3KTZPHp5Z7zqf6naBlpOl6I9jq3Dur98pkgnL1N0Ojq96f9IFeQ55YY3tE6/TX7tM085xpXMfqyoX6PBXTqm6L1PYOsS3s7y95iOEVooNU8fZUDqO1DFD+W203Dzrp068Z7OsBxUKh5rXRcEfQ2gYLOzgUteBFVkplApr5NphXOoYqSMsFSF1nHxNRVg6+VyqeObj6ctIHc1czlH+zPPydNVX3e2gZeeILRs78nHjFtmwTCYo33G2u+4UcOMmUqv2LOqeFb+kaCkGVD/eSD1+DClqpKwLNKIDwI3XVSaczjETTo3lTgOh9NodPj0iy0UgsOMWkXRoOQU0yRNQPpgkG5TvecbtvUwHHSUtOpojSBRVt4vBwnZG8ltmnf2+9Q1oB7TWGMVmMjOaC6Zmxlh3ipRTvbScHFYckA3K5IIpAitFYKWJLRclLaw4xIvqZMLKrGAyHdTrNwXW6QAbSm9Bn+hEe//JjvbtLLZGZ2aoI+lQSvUlJbg6prM5QqE1dsttQuZSy0pT8zrJBFUyUZVIOgzntjJY2L4kt8oQOmZ15TwbSsfJhhWadp6Bwg7SYZnV7T1LjfkRCoe610GxNUYkXS51HWSguOueKiZ6alfYN/JNQsvlYtej7Bh7DVvPfVBcLL5MIdBYOuZ079MMF7Y98HVKFbGhdJxNU0cBuNqxFyf2WVs5O6eP5UtFJBy0YOZ+5cYtRrMbKaX72VA6MdP8bSEEwqWR6qTQHEULycWux7jauTQ7kt4rqUI2TR1l49Tx24ZEhUAJGyUtYmGhhJ28lzaxsFFCokXyPS0kkeUSSwc79nHDRhJCw+oDllxbNNwC+WCKqVQ/p/qepek+ePXGUnYjoF4gYx1etICadOTdQ0vtxhaj2GKCltoF75jOMOF0jplwahhzx5cpql4XXtwkH0yhkIxn1zNY2M5kZi1aSKSKKPjjM+Gs2BqbmfWNhEMl1UPV6yKWDl5Yp+BPkA2m7iqkaSCQKZpugUB6xJaLAOy41S6Lnv0kGUqXppOnaedpOjlaTj753MnPyYyrHfsUW6N0NEeSGWB/fCY8Ne0c5VQfDaeA1BFZv0TRH5vX2YD3Ml3+GEsnKZ3WMTW3g4rXQ9XrpuZ1tf8uC9+VNCkVq9DdGGB96SRe3KTqdTGY30ZHY5i+xrUV+SJ9qWpZaQI7TcGfpOHkOd/9OGPZje99okdrNpROsG3idapeN8P5LWwbf2PeT74shmTGKE82rHC9sJ2zPe+7v2oErempX2XH+Gukoxqj2Q2U0/1smjyKs0iPEwslRhLaKepOkfHcJtaWT5ELSos2Hl+m8J0sBX+Cyx2PcKH70PKdwdOavvoVto+/RiqqM5zbQtPJk2qX3koVJ5sl6RgL1Z45TWZIb545lTpe0I7nN8+inu9+nIHi7uX7N7gLNwLqRTLW2wt+U5OOvIcI1AZccRmNINQbb3lZE07nmAmnhjH3NFBzOvGdzEy3WN9K49uZma6YAA0nTznVR8vOIXRENqgkAW2OyldvppDtsuhsUhZtOQitcKIWqahGOqzNCq4KgW9nZ8LqjRCbvEXSnf3EqDXpsJJ0AG2N0NEcnWnCoxDUvC7KqV5CK40bNcjfpgPoUjC9B6fUmmxYmvX3aDj5pGGW203V66LWXq88FywVzDTNyASl5OOwRDqszYTPyfRqxnIb6K1eprM1siS2g3lYJU2TNJmwSinVz7meJ5JO3u8gtGLn2MusrZxlNLsR30qxvnJmEUa8sGpOkVxYpup2cmzV++9pzVwmKLNj/FW6G9epO0Wudu5lXek0+WByHke8tGgELTs70whuKZhu3DRQ2MWZ3vctu3CUCUrsGHuV7uYgNbeTax27WT91gtwyaRj3rlnU/mfnZA38UnUjoF4iY721YHc3pV1q0dNEuoe0PEmg+oi5fSWVCadzzIRTw5hf091iEQKtoeZ1EVkuXlgnH0zOCquLNkZh0bKy+HZmpiwJBJYK8KIG6aj2rsCczLrmaNn5mU6405cJpUs51UvN7UJLSSqoUvTHSYeVZRempks5A5lCWUm5VjYoz3rB6Fvpdofnrvb7blp27tYv3LTGiVtJo4xwOogm729u6HKjHLuDlp0lthxkHNLVHH6oXqAvdRqoet2kwhqu8hnObeFC96GZ7Yrs2OeR4a/R1RziSsc+cv4k3c3BxR30AmpZaSwdI7TiVN+3MZrf8p6Xt1TA5skjrC+dIJY2VzoeIedP0V+/tOweO1aq6YA6lN/Kqb5nl25Pg5tYKmTz5OH2/crhcucjFJrj9DWuLMv7Vd0p4MVNhNac7z60omdRFzqgxjrX7sibbnfk3Y7ivRuCmXA6x0w4NYyFEwtr2a0J1EBgpWnaWUKZIpY2SjpIHePErXZjJk051Zs0oVJhu0R3fMWW3sXCou4U8e0MsXSxdEQ6rMxazxxKl5qblAL7doZMWGl3bizP+r3EwqbuFqk7RQI7gxYSK/bxomS9UjqsLvrJC+POkrXnPe0TB5prxb2M5Dezb/gbpMMq53qeYG3lzKKWZi6WGInv5MiEFQaKuzjb8+S71+lqzarqBbZNvIEbNxnKb6Pp5Nk4dWxOu9Eac6NlZUjFDUazGzi+6oWluyWX1vTXLrJ9/HW8uMlg+361aeoY1jK/X0VYtNwCuWCKqdQqTvY/u2K7Kk8HVFdeJmu9OW8BNVQ97Y68kLGO0oj3o7lz4zsTTueYCaeGYTyIWFj4VoZUVF+SJboLRSFo2Pmkm7TlIVCkwjq5YApLx+1tj5KmVKGdRsNMSXU2rOC9o/uysTyFwqXuFeloJR1kQ+lxtucJtk28MWfdpper6TLfitfNsVXvn3khnW+Ns2P8FTpaY1S8bq4Xd7Jx6jiZsLLIIzbeS8tKk4qbTKTXcHT1B+Z1+6D7kfMn2TH2Cp2tEapeFwOFXWwsrbz7Vc0pkIqagOZC9+MMFHetyFnU+Q6ofryeevw4kjpp+xz16ABwdyddTDidYyacGoZhzA9NMsNQdzuxdEQ2KC1awydjYTXtLDWvi6nMGraOv7HsKibmS8vKYOsQDZzteR8drRHWVM4SWikudeynu3Gdnub1xR6mcZd8K4Ub+5RTvRxZ/aH73l5pLtmxz5bJt1lbPk0kXS537qezOUxP49qKPQE4axY1vYqTfStzFnU6oDpiGFuOYYk6kjpSNBAE9xVYtYaW2k0z3oMtxrDlGK14N+/syPteTDidYyacGoZhGIaxUJIy3yyZsIpCcL24i1hYbCidfKirL5arQHrYKqTudfD2mu+48/Zp80VrVlfPs23iDZy4xWBhB76VZmPp+ENzcqjmFNt7u2vO9zzO9UJ7FlVrpI6xVYClQmwVYqsAW4VY7ffv/vjGZWJpUU71td/656wJ4P0QNAAbzTs754dYojETVi2RvJ/+XIp3b9mVdOR9jEBtxBVX0ShCvemex7Tcw+ky2NndMAzDMAxjfliopMOx18t4fgPrpk7NagZmLC+u8gmlSyYocWjg87y99jvw7eyCjiHfGmfn2CsU/THKXi8Xux5l49SxJdXpeCHkwvLMLOqusVfYPHkEqWMsFd5V7wIlJJFwiKVDJN1kr3g7g6MC1pVPsbF0Arix00Ap1U851Ufd7ViwUmJNBoGPxQRC+AgipNBoBEq7KJ0h1L3A7DJzQTArrFqiTqDWEuleUvIUoeoh5t3d1h8GJpwahmEYhvHQ6/DH6PDHFnsYxhxwVEAkbLyozqGBz/PW2o8uSFmpHbfYOvEWaytnCK0UZ3qepLs+wK6xl+f92EuVTUwumKLmFGm4HYSWi8JCS8l0qapAI1SEVApBjNQKqWJsfWPW1IvqZFWIRM0sUWm4HUSWi60CuuvXWV29AEx36O+jlOqjnO6n4vXc397Gd0njEeMxk7dn5W6FoImkhBAhUkQIFBoLpV1inSfU/SSRLCZjvU0r3opi5W7JcycmnBqGYRiGYRgriq0jIiwc5fP4wOd4e+1Hkxm1eeDELVZVzrN56iiWChgo7CKWDtvH3zCl4W25sDxn+7cKIB03SDdvVDhE2Ex5fYR2CkGyN3FPYwBI4mHV66ac6qeUTsqBAzszJ2O5M4kmS0w2Ca3vmjCO2+G1gRQtmvGeu+rIu5KZcGoYhmEYhmGsODYxsdJIKXls4PMcXvMRqqmeObluoRVdjeusqZyjp34NiWIq1c9ofjMbp46Riupzchzj7thEdPqj0O7xpxBU3E58O4uSDl5YY23lNBvK7VJgO89obhMXug8tcjdhC02OmBzx0my5s+BMODUMwzAMwzBWJAuFUAGR5fHY4Bc4vPrDlNP99319maDEmso5VlUv4MVNAulxvbgT307TW7vKzrFX5nD0xv2SaArBFARTM19rWRlKbpHISuHGTTaVjhFaKa527lvEkRrvZMKpYRiGYRiGsWJJNE7cIrDSPDr4RY6u/iCTmbV3/fN27NNfu8TqyjmK/jgKwURmLaX0KvKtCdZUzj40HXiXs1TcIHVTKXAgPbZOvEEp1Ucl3beIIzNuZsKpYRiGYRiGsaIJwI2b+HaGA4Nf5viqFxjLbbz9D2hFV3OI1ZXz9NavYOmYulPkQtdBLBXRX7tEb3tNo7E8TXd2fmTka7y6/nuWxL64D0RreupXKVSqwO7FHs19M+HUMAzDMAzDWPEE4EUNWnaWfcNf41TfswwXts26TDqssLpyntXV86SiOqF0GcpvpeXk6a5dY8vkYRZzhaIxt5K9ViP2jnyTI6s/tMjrT+9fOiizY/xVehrXKaV6UZOjwNLc5/ROTDg1DMMwDMMwHgoCSEd1mnaOvaPfwtIhw/lt9NUus7pyjs7WCBqYTK/hWnE3WX+KVdWL2Dpa7KEb80AAkZD0NAbYUDrO1c5HFntI98RSIZsmj7ChdAIlLc53PUYuKtPjtRZ7aPfNhFPDMAzDMAzjoZKOajTtHLvGXmH7+OtYOqbh5LnUeQDQ9Ncus33ijcUeprEAHB2115++Sbm9N+qSpzV9tUtsH3+dVNxgKLeVhldk0+RR3JSDWKYzwGDCqWEYhmEYhvEQSkc16k6BcqqPuttJV32ATVNHEe/ejNJY4abXn+4b/jqvbvhuIiu12EO6raw/xc7xV+hsDlN1u7jU9SgbSsdZXbvQvoSzqON7UCacGoZhGIZhGA+lbFghG1YWexjGEmCrAIuIvSPfWpLrT604YMvkYdaVTxJLh3Pdj1NsjrBr7MUVtQ7ahFPDMAzDMAzDMB5qAoiX4vpTrVlVvcC2iTdw4yaD+e34doYtk2+vyC2MTDg1DMMwDMMwDOOhZy+x9ac5f4KdY6/Q0Rql7PVwsfsgmyaPkY5qizqu+SQXewDGynG916GeMncpwzAMwzAMY3lylU8sHR4Z/jpOvDhdb+3YZ+foyzx57c/IhBXO9DxJKD12j768ooMpmJlTY45cWuPymRc66KhE/LUvTZEKTDMBwzAMwzAMY/mxVYAmYs/INzmy+sMLt/5UK9ZUzrF14k0cFTBQ2EksbbZNvLkiS3hvxUxzGQ+slLP44jMFOioRlazF554rEpt7lmEYhmEYhrEMCUAJSU/jOhtLxxbkmIXWGE8MfI7dYy/RcIuc6X0fvY1rbCodf2iCKZhwajyg0ILPPVcADY+eaWLFmoF+l68+njeN2A3DMAzDMIxlaXr96ZaJt+hoDs/bcdyowe6Rb/HEwGfxojpnet6HQrJr7GVSUX3ejrtUmbJe475p4CtPFhjvsPnQq1W+/nieyBakWoqT29J0VyIeO91c7GEahmEYhmEYxj27sf/pN3h1w3cTzuH+p1JFbCgdZ9PUMYRWXCnuBRTbx19FPsRTPCacGvftyI40ZzaneOJ4ndf2ZYnspB6/lZJkmjHfejRHRzVmy/VgkUdqGIZhGIZhGPduev3p3pFvcHj1Rx58/anW9Ncusm3iDVJRg9HsBkrpfjZOHcNbpAZMS4kp6zXuy/Veh289lmPTdZ+RLptKzpr1/UbaItNS/I9niox1mHMghmEYhmEYxvIzvf60uzHIxqkHW39aaI3y+MDn2DfyTUKZ4lTvM6TDKjvGXzfBtG1OwunP//zP85M/+ZMAvPTSS3z84x/nIx/5CJ/4xCfm4uqNJaaeknz+2QL5ekzvVMTVNd4tL9dIW1hK85kXimaLGcMwDMMwDGNZmtn/dPL+1p96YY29w9/giYHPkYpqnOl+kpaVYdfYS+SDqXkY8fL1wInh5Zdf5tOf/jQArVaLf/bP/hm/8iu/wuc//3mOHz/ON77xjQcepLF0xAI+/2yBwJE8caLB63sz73l53xE0Xclnni8SWu95UcMwDMMwDMNYklzlE0mHfcNfx4nurqeKpUK2TLzF01c/RW/9Cpc79jGS28j2iTfobQ6wQBvULCsPFE5LpRKf+MQn+KEf+iEAjh49ysaNG1m/fj22bfPxj3+cL3zhC3MyUGNp+IvHcgz2uTz3dpVvPpa7c929FAitGe2y+dLTxYd4ebdhGIZhGIaxnNkqwIl99o58E/R7vKrVmtWVczx95Y/ZPHWE8cx6LnY/yprKOTaUTyNRCzfoZeaBwum/+Bf/gn/8j/8xhUIBgNHRUXp7e2e+39fXx8jIyION0FgyTm/0OLwzw/6zDY5tyxC4d3f3iRyJF2jOb/B4eX92nkdpGIZhGIZhGHNvZv1pc5BNU0dveZmO5jBPDPwZe0b/gpad5XTvM2SDKbaPv4Gr/IUd8DJ0351qPvnJT7J69WqefvppPvWpTwGglELcNJOmtZ71+d3o7s7d75DmXz7N2GKPYZGMdVh85akCq0cDWo5gvPPe7jq+J0k3Y17fl6WzErP7sln0bRiGYRiGYSwvM/ufTr5NKd1PKb0KgFRYZfv46/TVr9CyM5zteYqu+jV2jb20oOMTQlAsZujqzS/ocefKfYfTz3/+84yNjfHd3/3dlMtlGo0G169fx7JuLCwcGxujr6/vnq53YqKGUkuz+LNSfTj37PQdweeeK+KGii3XA1589P5OIDTTFplmzJefylOox6wdC+d4pIZhGIZhGIYxv27sf/p13lz7MdZUzrKhdBItJJc69+PEPtvGX1uU/Uq11pTLDeKx6oIf+25IKd5zMvK+w+lv//Zvz3z8qU99itdee42f+Zmf4SMf+QhXrlxh3bp1fPazn+X7vu/77vcQxhKggS8+U6Catfjgq1W+8tSDnYVppC1SLcVnnyvy1784SbFuau4NwzAMwzCM5SXZ/1Ty9NVPIYCh3BbqbgcbS8dxVLDYw1u25nR/D8/z+Lmf+zl+9Ed/lO/8zu9ky5YtfPSjH53LQxgL7LV9GS6t9Xj6SI0XD2ZR8sH7irU8QWwJ/vSFDnzH9CkzDMMwDMMwlpfp9aeTmTWc7n2aoj/Otsm3TDB9QELr92o1tfCWdFnva68w9nu/Q9x8ONZLXl7t8qcvFNlxxaeSlQz3unN23TLWaAHrR0K+++sl5NL8kxuGYRiGYRjGbWkEYgntRyHTaVb9wA+S239wsYdyS3cq653TmVNj5ShnJV94pkB3KSbtqzkNpgDKEjiR5upql28cWsJNsAzDMAzDMAzjNpZSMF0JTDg13iW04HPPFUHA/vNNjuxIz8txAleSbsUc3ZGZt2MYhmEYhmEYhrE8mHC6TCkBpzanOLfem9N1mxr42hN5xjptnnurxrcezcE9bgd0L5qppIPvNx7LcXn13M7OGoZhGIZhGIaxfNx3t15j8QS24AvPFLi0zgNAKM3q8ZCNQwEbBwP6piLuN04e3Z7m1JY0jx+v8/reDJE9/w2LGmmLdFPx+WcL/M9fmqKnHM/7MQ3DMAzDMAzDWFpMOF1mqhnJZ54vMlG0+ba3qwz1uAigkpW8fCDHywcg3VJJUB3y2TAUkPHvrhZ+qMfmm4/l2DjoM9ZpU84v3N2jmRJ4oeYzz3fw1784eVdjjiQ0UpJ6RlJLW9TT8qa35PPeqZAnTjToqpjAaxiGYRiGYRhLmQmny8hIl82ffXuRwBF88LUqrzySpZa1km9qTWc5oliNQAour3Y5vTkFWtM3GbGpPau6aiK8ZWfcekryuWeL5JqK/omI1x7JLuyNE4LAhsgS/NnzHXz0xTLNlKR2U9CcfpsOoq3Uu6vSpdJkWopMU5Grx5xfn+L0phTbr/o8caJBbyla2NtlGIZhGIZhGMZdMeF0mTi/zuOLzxRI+4rn36zxjUM5QuemcCYEU0WbqWLyJ7VDxZqRgHSgqaUlr+/J8Nq+LG6g2DCcBNWNQwH5piIW8D++rYDvSl54s8qXn8wvym3UUiBCxXCPw+98d8+s74l26Mw2FYV6TP9kiBtprBiUhMCC0BG0XEkjbVHNWox22bihZu1ExOU1Luc2ptgy4PPE8TqrJpdXSA0tGOh3ubI6eXNDzYGzTXZcaWGrxR6dYRiGYRiGYTw4E06XOA28sSfDSwdz9I+HbLsW8JUn82j53mtBI0cy2N9uMKQ1hWpMZzVGahjqcTi/IQVAdyki01Jc73d54fXqvDdAupPIkWTrMdsGAiyt8S2I7OnQKallktB5t2MMXMH1fhcnVKwdCbje63Dxo11sHPR58niDNePhPN+i+1fKWVxe43J5jctAn0tsC+xIs3YkoJKz+POnC7x4MMv+c00eOde86/Jtw1hKSjmLk1uSxyNLaaS69XsrBqnbn8caSyWVEje/9wJl/h0YhmEYxjJmwukSFkv46hN5Tm5Ns/1KCy/QvPjofZTbCkGlYFMpJH9uGWtWjYVkW4qmJ7je63DgTIPj21L47uI3cK5nLY7snNutZUJHcr3fxYo0a0YCRrodPvmRTtYNBzx5os66kfC+m0jNlciCgT6XK2tcLq1xZ9b8FqsRey80yfqaybzk2iqPRkrQOxFgacEr+3O8vjfLrkstDp5pmIZSxrJxtd/h888W8d32v74HPDEmlOZ9x+o8frJxy+ULhmEYhmEsbSacLlFNV/C554pc73d5/ESdkS6HcxvnZqsVZQmGe52Zz7P1mMurXcqFlX93iG3BYL+LjDWrRwMmixaf+mAnq8cCnjzeYONQsKAhtZyVXF7rcXm1y0C/S2QLrFizbiRg1yWf0IahXodj29Pvmi0f607uD8VqRL6hOLMpxYltadYPBTx6psGmwYW9LYZxtzRwZEeabz6Wo7Ma88zRJqc3pRDtGVChNVIn74UGoUnCptZILYDZX5caBDBRtHn5QI6BPpfveLlMtmUSqmEYhmEsJys/jSxDU3mLzzxfpJK1eOG1Kkd2pmfWks6H+nRTpYeIsgRDfS6iHVIrOYs/fX8HfRMhTx6vs+X6/AS7SML1PnemXLdUmJ4djdl9sUmuNT076nJljXdX11nO25Tz4PmKVeMhEx0Wn3mhg85yxMEzTXZfauKYyVRjiYhkspfyya1pNg/4dFZjvvb43C0nKFQjBvsc/svHuvjoSxXWjyzd0n3DMAzDMGYz4XSJudbn8LnnikgNH3ytyl88mqN5i660xtzQ0yFVaVaNBdTTFp99voOeqZAnjzfYOuDfdXmgEtD0BI1U0k24kUq6CzfSknr749EuZ2Z2dO1IwI4rPrFMZkePb3v37Oi98D3JwKobgTtwJV97Ms9LB7I8cr7J/rNN8s0H756kSbY0KuUtSnmbUt5CyyRgd7Tf8vUYy0xaGe9QTyUVIUO9LodO1hkv2ry1OzOnx6jkbZxAYSn41Ac6eOp4gyeP102Zr2EYhmEsAyacLiEntqT46pN5itWYg2ebfPXJPLFlCjMXgpaC4V4X2iG1mbL4/HNFOssRT56os2o8umXYbNz0vpGStwyXbqBmOg3vutQk39BMtWdHr97l7Og93ZZ24EZruqdCvBDe3J3hrd0Ztl31efR0447dijVQmwmgN0JoKWdRzluz7pdWrJFKz+oeLZUmX78RVqffirWYQs0E14fRaGeyFVbLk3zg1Qpv78rMW0VI6EpCR9Ndjnn1kSzXex0++lKFbMu0tjYMwzCMpcyE0yVAAy8ezPLmnizrhwL6J8I5LXMz7sF0SNWa/vGQwBV88Zniuy4mtCbdDpyZpqK7FOGFyXo5EISWJrQFgSNoedOzpjbX++dm3fBdEYKJzmRtca6edGu+vMbl7KYUq8cCHj3dZNVEOBM43xlCY3t2AC3WYjoqMetGA9ww2cKn6Ypk79mMRSqISPkKJ9IIBL4jqGUkg73OrOAqlKZQnx1Ypz8u1OP279BYSc5u8Pjz9xVI+YoX3qjxrcdy8998TQgmOmwK1YjhHoc/+M4uvuOlMhuHTZmvYRiGYSxVJpwustCCLz5T4ML6FHvPN/BtwRv77qMjrzG3hGCkxwGt6RsPWTMRgYbQTt58pz2DmpaMd9gPVI67EGpZi1rWSrbUGQ6o5JOZ4ZvJ6QBajVk7E0AFLQfqaUk5n2xto+5hNl8oTbap6J0KSPkaJ1I3gmtaMtTjELizg2uuoZJxtGdZO6rt97UYLzRTrsuJBl7en+X1fVlWjYVsvR7wlSdzC/rvZbrM1440f/L+Dp440eB9x0yZr2EYhmEsRSacLqJaWvJnzxcZ67B55u0aF9Z7SSAylg4hGO1xGF0hf5fQkVxfNV2+HNI/FWFHmqYraKSSAHp1tTtn5eRaiplg/O5vJkG0ZyogFWicSCOBwE6C64V13rvWW6d8NRNUC7U4Wefa/jjXVCZwLCG+LfjiMwUurfPYfaE5UyGyGKbLfLvKMa/vyzLYLvPNzcEabMMwDMMw5o4Jp4tktNPmz54v4juCD7xW5bVHslQfwq65xiKRyXZCN28ptODEewfXTEvRPx6Q9hVuqLGVILIEDU8w0mVzfr2HkrNLj/P1pEw4X1d4gcILNalA4QU6eQunP07em7Wv86OUs/izby8yVbD4trernF+fWvwTb0Iw2S7zHel2+IOPdfEdL1fYNBQs7rgMwzAMw5ghtNZL6uXZxEQNpWYPKY4jpqbGiKLFfRGhfJ+4XocH/JVFFjQ9idDghZqWK1hSfwRjUWk0Y/4Unx/5Oo24tdjDWZLsSFOoxWSbMW6gcVVSQtpyk3Wu9bSF74pZ4fVWnOmwGt4IrKl2qHUDTWc1Yts1H9tsxXPXrvY7fP7ZpGT8ubdqvHwgSz2ztE68OYEiHWgqOYvHT9R5+qgp8zUMwzBWBplOs+oHfpDc/oOLPZRbklLQ3Z277feXxczp1NQYqVSGbHYVYhGbBMX1GqGYAHV/pWBKQC1joTxJR5S8CK5lJNl52VHTWLa0JhsU+E5e4I8Gv7DYo1mSIjuZBZvsePdDmFCaXFORLcW4YbLO1YrBUmApjaUEAo0mWVMbWRDagtAWVDOS8U4b3xEza2G/0VLsvdBk/7kmhYYpA70dDRzZkeabj+XorMYcONvka0/kZzXWWipmynxLEW/sTbr5fuylCnnz9zUMwzCMRbUswmkUBYseTB9U05PUMhItINNSaAHVJTabYCwRQiBdl16vc7FHsixpKahmrfsqk7fiZAY121R0VmKsWIOAt9pb8WwZ8Dl4tsna0dCcUrpJJOHrj+c5sS3Npus+XZV46Xccb5f55msR4532TJnv5kFT5musHFdWuXzrsRybr/s8c6RuHrcMw1jylkU4BZZtMI2sZDYmcCROpEn5ikZaEi/x7q7GIhMCYV5GLLjYEjTSFo307K/n6jGdlZiBfpcLG1J0lyIOnGmw63ILZ5mV/CqRPC4lW/48uHpK8Llniwz1uRw6WWe8aPPW7swcXPPCqObspLw71HzmhQ4eO1nnmSN1sx7ZWNZaruCbj+U4tSVNyle8sTdL4AheeKNmnlkMw1jSlk04XW60gHoq2W5EaMg3YnxHmKZHhrEMTTdusiLN2hGfpmfx1acKvPhojr0XWuw/26BYX7ySUCWg6SUdl5M3i0aq/XnaopGSyeNRStLyBFoKpNJ4frK8IOUrUsH0x+983/5e+/3NoXa00+bPvr1Iy5N84LUKb+/MMFVcfk8roSMJ7aTM9609WQb6XXZdbrFmNKS3FJn1qMayoYHz6z2+9nielid47GQd35FcWeNydEeGyBJ88LWquU8bd6QEVDOSUt6iWFN01JbZmVhj2Vp+ryKAH/sPf0GlPvelV4Wsyy/+6LPveZlf+u3f5MSZU4RRxODIMBvXrgPgL3/Hd/HR598PQOAIKlmLWApSvsJWmlpaom+a/f3y55K1hB/6ro/ecVx3c9k/+v3/yr6D+9m1b+8dr+9+vfXq6wxfH+Q7v/e772n8d+vX/t2/Z/cj+3j+wx+cs+ucdvLoMT71B/+Nn/r5/3vW18dGRviXP/FT/Pvf+c05P6ax8sS24Hq/B1rTMxnihZq3d6Z5a1eazdcDDpxtsGF4bkt+IyvpflvK21RyScCsp+VNQVTS9OQt9w614qTrcaapKNRj+iZD3EiDEIRWMlMc2oLASSo8xjptWp4keo91ojJuh9ZAU8lapHzF+1+v8s1DOXxX3vbnlrybynybKck3D+WBpGnW6vGINWMBa0ZDVk2Ey2623Hg41NKSrz2e5+J6j97JkKePtXh1X2amIVm2HnNya5rYEnzk5YoJqAaRBeWsRTlvUcol78s5i1I+WRoz3VTQjjR/5ctT9E9Gizxi42GwLMPpfATTu73ef/i//wAoxfDYKD/+f/80v/7//MLM95RMXuC1XIkdawr1OJmpsN79gu1eQt3dXPb0sePs2b/vrq/zflw6d37m47kMpYax7AjBeFeyNUq2EdNViRnucbi0rpPOcsSBs012X2olQfAuxAIqOYupfPKioFRIwuhU/t1b7VixJt1SZFqKfF3RMxXhhRqrPXEbyRsNnpqepJmSlHMWI932e64BnV5vW6jHeL7CiTR2pLEVSJVcv0CghJ4JtX0TId1lxZefyt8yHC9H1ZyNjDX94wHZpkJqwVRB8soj2WQ9uNL0TUasGQtZMxqwZjwk7ZtX+cbi0cDxrSn+4tEcsRS870iN8Q6brz4xe913PWuRbcSc2ZQikvCxlyozjxvGyuU74l3Bs5xL3t75/OIGimItpncyYtNggBNDaAnObvT40xc6+GtfnFzUKiHj4bAsw+lSo4HzE8P80n/6NWrVGhnX4/v/3t9l3b4d/Nq/+yVc1+Pi2XM0G02+52/8VZ774Pv549//rwB839/6G7z4tW/wJ//tkwgh2LJjG3/vH/4Itn3jT3PzZX/kf/nfefLZpzlz4hSWZfGj//SfcObESS6eu8B/+vf/kR/7qZ/E9Tx++5d/jVq1iut5/O0f/gE2bd3Cr/27f0+tUmVkaJi/8f3/G//5V3+TZz/wAkffOozfavHDP/6P2Lx9G6eOHecP//MfEPg+9Xqdv/UD38+qtWv4yv/4IgA9fX2Mj47OjOmtV1/nk7/3B2il6VvVz9/90X9AsbODf/R3fuCur//xp5+65e/2F376X/Kh7/oYB584xH//nd/j8oVL/MTP/gumJif5uX/+0/z8r/4S3/jSV/j8p/8UgM3bt/J3fvjvk0qn+aG//r+xefs2SlOT/M3v/zsz13n5wkV+8xd/GYANmzfN5V3BeAjVMxb1jIWMNGtGAnxX8vUn8rx0IMueSy32n23SWY3RJCVSU+3gWcpb7Y8tKllrVrhzA0VHNWbtWEjmmo+twLeg5QkaKYtmSlLNWox2vXfgvBe3W2/7MFKWYKTHvfEFrekqxxTqEU6cnIQ8vCM9s7a2qzwdVkPWjAUU6sqs6zMWRCln8ZWn8gz0u6wZCdg+4PPqviwt79ZVDPWMRaYRc2FDis9agu/6VhnbZI0VZ6Jo8dq+LFf7XVqp2feFTFNRrEWsHQ3I+Borbm9x6CZ7j5dzFmOd73huURon1nz6A538tS9NmhNyxrwy4fQBRVZSwvtv/v+/xPd+71/m2x9/isNXzvGvf+4X+Lf/6VcAGBka4qf/3c9TLpX4qX/44zzy6MGZn58cn+D3f/O3+Je/9G/p7unhV/7NJzj82hs8/sz7bnm80tQUew/s52//8N/n93/zt/jSn32Ov/UD3883vvRlvvd/+ets2LyJn/7xn+Tv/IO/z6atWxi4eo1f/Nl/xS/8ZjKWXCHP//nTPwXAf/7V3yRXyPOzv/hv+OJnPsuf/vc/4sd+6if54mc+xw/8ox9hzfp1nDh8lN/99f/Ez//qL/HBj30HAM9/5IMzgblcKvFbv/yr/P9+4V/R29/PZ//o0/zOr/4G/+if/V8zx7ub679dOD34xOOcOHKUg08c4syJk0yMjaPimKNvvM3BJw5x9dJl/vS/f5Kf+cS/Jl8o8Nv/8df51H/57/zNv/t3qFYqfPyv/mX27H+Ek0ePzVznr/7CL/K3fuD7eeSxg3z6v/z3Wd8zjPulbMFgvwta0z0Vkgo0R7elObwzQ7EaUctYxNaNJ3s70nRUo5kz1G4EYfsFQrVdZjWX4dN4AOIdWxdpTa6u6KxEpMLkpMG5DR7HtyXJPttITiysHkvKgHumIhMAjDmlBLy1K8Mrj2SxlOa5N6tcXu3yjXY5+ntpZCwyzZjLaz0+80IHH/9GyZSqrxBTeYtXH8lyZqOHE2m2XfNJ+xqpwHeg5SYnNkt5m6Ee5+6fX6QgVlDLCD7zfAff+5Upc58x5o0Jp/dJi+TJYbJo4TeajA4O8dTT76PqSLbu3kUun2No4DoAz3/4g9i2TXdPDzv27ObMyVMz13Pu9Bl27NlNd08PAP/gn/zjOx57/+OPAbB+40ZOHz8x63utZpOL587z65/4pRtfa7WoVioAbNu5Y9blDxxKrmvdxo28/tIrM2N4+7U3ePVbL3L+zFn8Vuu2Y7lw5hxbd2ynt78fgA987CN85g//eM6u/9EnD/Fvf+b/odloArBhy2YuXbjIkTfe5CN/6bs4ffwEjz71BPlCYeb4v/GJ/zDz8++8vdVyhdLkFI88dhCAb//wB/j6l7582+Mbxj0TgonOpOQ304jpLsfYsWbjUIAbQiyh5QhqWUk5b3Nhvb1iSmIfGkJQy1nUcjdK4jxfsXY4IB0oIksw2ONwdmMKSMqiu8oR/RMRfZMh/ZMR3SUTWI37M9Zh8+X35Rntctg84LNmPOKV/VlC5+7XfDfSFulmzEC/w5+8v4Pv/nr5rpchGEtPOSt59ZEspzelsBQ8erqJpTRHt6dn9ux+UMoWOKFiuNvmC88U+a6/KJt1y8a8MOH0PviuYCpvowWkfA3NpGQvuOmJQaOJ4+S0kmXdeAGjlcKSNy5nWdas8q9KuQxAoVi87fFdt11uJpLj3EwpheM4/Ktf/sWZr02Mj5PLt5t7uO6syztu8iJaCEAn1/Wz/+SfsXv/Pvbs38fegwf4j//63952LFrrd30+fbvn4vq7e3vRWvPaiy+xY89uih0dnDh8lEvnL7Bj9y4uX7j0nsd3PW/2FYrZY5aW6Z5szJ9GxqJh9jN+KPie5PqqG4+vVqTpH0vWrdpa0PAEF9Z7nGjPrspY01OO6JuI6J8M6ZuM6C5HZg2gcVuRhFcfyfLm7gypQPP+16qc3JLixYPZ+7q+Ztoi1VIM9Th8+gMdfM/XSnihSRvLSSUjeX1flpNbUggN+8818QLNkR1p/NuUdj+I0JGkfMXF9R7fOJQzWxMZ82IZt1ZceJFWlLKSUs5GaJA6aTyiu/P0rern9RdfBpLZ0PJUifUbNwDwyrdeRGvN2MgoF86cY+e+PTPXuXXHds6fOUtpcgqA3/+N3+LNl1+757FJy0LFikw2y6q1q/mLr34dgGNvHeZn/69/ftfXU6tWGbo+yF/5X/8mBx4/xJuvvIpSauYYcTz7ldO2nTs4f/osYyMjAHz1f3yJPQceua/rv50Djz/Gn/63T7L7kX3sOfAIX/qzz7Ft106kZbFn/z7eeuU1atUqAF/7wp+zZ//tj58vFOjp6+Xt194A4KWvf/POvxTDMIx7FNuCkV6XixtSnN3oMbDKRQNrRgO2Xm2x7ZqPE2rObfT4ylMF/uvHuvjVv9rLf/uOTr76RI7jW1KMddjE5pWfAVzvdfiD7+zijb1Zdl5pse9ck28eyjHS4zzQ9bZSEjdUjHbZ/PEHO2h65g63HNRTkq8fyvG7H+/m5OYUey80efJEg7MbU7z2SHZegum0lifJNGOO7sjw5jLa09pYPszM6V3SWjOmqigbcs2YSqxnNrOHpFT1t375V/mj3/+vOI7Dj/3zn8R2kieNwPf5qX/040RhxN/9hz88U4IK0Nndxf/6g3+Pn////gxKKbbt2snzH/7APY/vwKHH+K1f/lV+6Md/jB/5J/8ffuuXf5XP/tGnsW2bH/3J/xNxl+sKcvk8L3zkQ/zED/0olm2z98AjBH5Aq9Vi1769/Pq/+/cUO2/M6hY7O/i7//CH+cTP/hxRFNHT18sP/Nj/cV/XfzsHnzjE5z/1p+zcuxsvlSIKIx598nEgaWj0l/7nv8LP/l//nDiO2bxtK9//f/zwe97Gf/BP/jG//on/wCd/9w/YvnvnXf1eDMMwHpTvSQb7ZlevpFqKNSPJujBLC2ppwZmNKY5tT15cWrGmbzJk74UWuy63zMzqQ8a3BS8+muXY9gz5WswHX032Ez69ee5evvmehRsoJos2f/zBTv7yV6fItswM6lLU8ARv7MlydHsaLWHXpRYd1ZjDO9M00gtXpdNIJ421Xnw0R64Rs+uKv2DHNlY+od9Zl7nIJiZqKDV7SMPDV1i1auPM54u1z2mpPklcq9FICdRdhr353LvTWNlqE+P8yqX/stjDMAxjgaWbiq5yRCpQWBrGO2wmiza5RsyjpxvsPd/CM+sDV4xYtLd/ctr7Dre3gqpmLV7en6WWkRw420RrOLYjPW9r1J1AoaUg14j5vq+UyDXNmZCloukK3tqd4ciONJEl2HmlRXc55siONLVFXDqSacS0PMn3fL3E+pFw0cZhzCbTaVb9wA+S239wsYdyS1IKurtzt/3+spw5fa8AOZ8cN0U100QtrTxvGIZhrCDNtOR6evZWNr2TAUILvvVYnlf3Zdl/rsnBMw0zwzWHGp7AdyVKJPuWJ+9v/lgkn8/6+FZfg2g6ZDqC0JY3Quc7AmjoiFldvN+pqxTxwdeqvL43SyU3vyEkdCV2qKilLT754U6+78tTFBomoC4m3xG8tSvD4V1pAluw/arPqomIwzvSnN68+P0MGhmLdFPx2eeK/NU/n6KnbFr4Gg9uWc6cLpZ62GCyNWXCqbEgzMypYRjv1FmKSAeawV4bS8Huiy0eO92gs2peFN6Phic4tyHFmY0eQ+8ouX5QdqSxI40baZxI44QKt/01OwY7Trb4sBQIkjd0smQolhBLge8KLq7zFnRLKStSSA1eoPm+r5ToqJn71kILbMHhncleyr4r2XKtxbqxiCPbU5TzS2xeSWu8QGPHmr/2xSnyZsZ90ZmZU8MwDMMwFsRUh80UkKtFdNQVp7akOL4txbZrPodONlg1GS32EJc83xZcXO9xZqPH1VUuWgq6yhGPn6gTWRItQZNsGacBBMTt97qdIme+L0T7Y41GJD8rxI3Z0/ZMadMTlHMOkcWS37s4tiVEmtCBT364g+/9SonuigmoCyGwBUe3p3lzd4ZWSrLpus/G4YCj29JcXJ9a7OHdmhAEDsSW4E/f38Ff/fMp0/XZeCAmnBqGYRjGMlPL2dRyyfrU7nLE1VUu5zekWDcc8PipBhuGArPFw00iCZfXeJzZ5HFpjUdsC/L1mINnmnih5uoqhzf2ZJZ8cFwosS3QMaDhjz/UyV/+aone0tyc+Ijb5dLTkhnjmz5ufz79sdDvuOwK5NuCIzvSvL0rCaXrhwK2Xg84vjXFNw7lF3t4d6SlQCvFVMHiz769yPd8rWT2cTbumwmnhmEYhrFMNdOSgbSLHSrWDgdMFSz+5P0d9EyFHDrZYMdVH/mQTmIoAdf6Hc5uTHF+vUfgStItxZ5LTfJ1zUCfzZH/l727Do+rSh84/r0yFvekLqm7AqVADYr7YvvDvUhxh8WlUIoVKVJgcSmysAvLAsVqSIW6e9wzPvfe8/vjTqZJk7ZpSZukPZ/nyZOxe+6ZmWTmvvc95z09PVh7qcBQa2dpCmHAHRHMGJfCqTMryG4gM2+qEHCpBFwqfrdiX3arsd9+l0ogervfrRJ2NsEyJ0KgCOi6JcRRc6tbbYGwoMMevruwlz18t1NeiM75EZZ1cfHDsB0Pe2yJTF3FGbbYmu3kf4ckcczsqv32ZIK0d8ngVJIkSZJaOcOhsjXHiWIK2hSF8XtU/jsymTkDoxV+1wZwHAAjMwVQkK6zsrOb1R1d+D0ajohF7pYQ6ZUWhakay7t4MHR52NwYQlMIKBAXEnwyNoXum0LRoHPXwaZiCTwhgTtkERe0yCg3cEUEDlPEhkNvezCx7Gnty2Inb1PYobCqo4uPxuuc9GMFSb7Wk6oLOBUW9LKr74adKl22hOhYGGZJrocfh7qau3t7LOxUcQdNVnV2k+g3OWyhr7m7tEO1/9yklqVVBqfetyYiAlVN3q7iSSLhvGd3+pjpz7/EymXLMSIGhfkFtOvYHoBjTjqRUeMbv1zMt//+GoAjjz9mzzu8jxUXFvLQbXfzzBuv7JP9/TH3V9avXsPfzvv7PtmfJElSayc0xS7sIwRZpREsVeHHYXaF34GrA/a6iPthgZuSZI1Vndys7OymKkFDMwWd80LklAYpTVJZ197Fii5NkLE7EKkKfhck+yzWt3PiCQk8QYvMcgNnxC72pEYn50Y0gaEpGDoEHSrBaAa1MtGx06rEe8odNKmOU3nv6DRO+LmSdsUtezkTv8teEubP7h4iDpXczUHaFRks6ebmx/Ytf/huYwTdGnEBkz/6xJPotxi4KtDcXSKiQWmKTnGKTkmqTnGqTkmKjissOPGnSrLK5Vz9lqRVBqd7IzBtbLsXX30llhCxQO3RqU/v0b5aU1DaXIYechBDDzmoubshSZLU+igKRekOANLLDRyGYF7/eOb1jye93CB3S4hum0NkVBjNkjkQQFW8SshpL7NiaESDGsX+rUVvq31d33Z7RLeXYDG0aLGhRB3FErQvDNNvTYDqOIW1HdysbalFZFobVaEyQSWt0iTkUKjYS8Hm7gq6NTRD4DAsPhmbwtjfqum7Ltjc3arH51b5o3cci7t7MDTovilEm1KDP7vtn3+jfo9GnN/kh6EJxPtNum0J75P9CsDnUSlJiQagqXZAWpGoxdYHdkQsMioMem4Isb6dk4+PTOGEnyrpKNdpbTFaZXDaEuVv2cprz72At9qLy+3m/CsvJbdHd16a8gxOp4t1q1YT8Ac45ZwzOHzcGGa8/R4Ap597DrNm/shn73+Eoih07dGNSydeja5ve2v8Ph8vPvkMhXn5ZOVkU1ZSyg333E56RgbvvvYmyxcvwbIsjjhyLMeeehLL/lzM5x98jMvlIm/zFtp37sQ1t96I7nDw83cz+fqzL7CERZduuVx41RU4nU6uPPt8unTvRkV5GQ8+PZnXn5/Glo2bqKyooGPnTlxz2007fO6zZv7IlzM+RVVVMrOzueqWG3A6nfzrwxnM+3kWlmXRf8ggzrn4AgKBAM9PepKK8nIATvv72Qw95CD+88nn/PzdTBRFIbdndy659ip+/N93LF+8hCtvvI7VK1by1kuvEolESExK5OJrryKnbRseuu0uuvbozsqly6murOT8Ky9j0PChe/fNliRJakVKU+3vk6Qqg/QqE79b5dd+cfzaP54kr0nulhC5m0O0KYnstfmpAihN1tia7WRLloOtWU4C7sZlMlXLXqZCN+zlVzTTztbppsAZsYgLCHqvDxF2wNr2Lma3ab3DIls0RaEspeUdNpq6gk9TSam2+PaQJMqSdUYu9LaIudbVHpU/+sSxJNeDpUKPTUEyyy3+7O5mdaf9LyitzR9nZ1C/PjSZ074vp21J02YnTQXKk7RYAFqc6qAkVa/zuZLoM8koN+icF8ZhCgJOhcoEe5v8DAcISAhafD46hfFzq+i5MdSkfZT2TMv7lGmlXpj8NCedcRrDR45g9YqVPPPw4zz56gsAFObnc9+USVRWVHD3xJvoP3hQbLuyklLefmU6Dz37JOkZGbzwxFMs/PV3hh16SOwxn7z7AW3ateWmf9zJulVruPfGWwH4/uv/AfDwc1OIRCJMuvs+unTPBWD18pU88fJUUtPSuPfG2/hz/gKycnKY+fU33PvkYzidTt5//S3+PeMzTj3nTKqrqjjxjFPpM6A/yxcvRdd17p8yCcuyeOSOe1j42x+xtrf30Vvvcv+USSSnpPDua2+Qt3kLleUVrF+zlgeffgIUhRcnP82smT9iWRYZWVnccv89bFi7jtk//MTg4UP514czmPr2dFRV5ZWnp1JWUhpr34hEmPrYZCbeeSu5Pboz7+dZPD/pSR58ZjIApmFw/5RJzJ/3Kx+99Y4MTiVJkhpQlaRTlRQNVKsN0qpMIrrCou52lVBP0KLrlhC5W0J0KAj/pWqbAihJ0dmS7WBrloOtmU6C0YPGBL9Jx/wQyT6LkEONVm8VCNVegsWKLsViaYq93ud2WdOQU8HnUTGj96MorOvw118fqRVTFCqSNJKrDOb3jqMsSeOYWVXNViipKk7l9z5xLMv1IBTouSFIeqXJoh4eVnbWmqVPzcHv0fAELf41KoWzvilv9HrMAghG/8/tHw2fR8XvUfG5VSoTNUqT9Vj2XjMFaZUGnfNCJAQEQgGvW6E8SWdzjpP17XeQ5VfA61FJ8ll8PTIZv7uawSubfxjygU4Gp00gGAhQmJfP8JEjAOjeqycJiQnkb9kKwKijxqHrOukZGfTo05uVy5bHtl29YiU9+vQmPSMDgKtuuaFe+0sWLIrd3rVHNzp07mTfvnARm9atZ+mffwIQCgTZvGEj7Tp2oH2njrE223Voj7faS0nRYgry8mPBrREx6NJtW8DZrWcPAHr370tiUiLffPEf8rdsoSAvn2Bwx8Nkhhw0nPtvvoNhIw5m+MgRdM7tyjuvvs7alau46zo74xoJhUnPzGT0+HF8+ObblJeWMuigYZxyzpmomkb3Pr245/qbGXrIwRx32smkZaTH2s/fmkd8QgK5PboDcPDhI3n1uRfw++yJ9gOGDgGgfadO+Kq9u3q7JEmSDnhViTpVifYhQLzfJL3CAEVhVScXS7t5cEQsOueFyd0SovPW8C4P8i0lGoxmOdia7WRrpoOQyw5GE30mnfNCpPgsgrpCUZrOmo7uFjEsVNq/VCbpJHpNNrZx8uH4VE76sYLkfVgoqTJe5be+8Szv6gYBvaNzvBf18LC864ETlNYWcCk4I4JPx6Rw5v/KUYTA79bwxtmBpj8WgNYKQt1qg58P9kgJiySfSd81AdwRCOlQHa9SmqKzvIt795eDUhSqEjSSqk1+GpqIz60ycpFPFkpqRjI4bQKWqP+lLRCYpn2GSNO2fSAJy0JTtw050DStzj9AVWUlAEnJybHbVFVFNLQPy+Kciy+IBcXVlVW4PG7WrFiJ0+nY9kAFhBBYpsXBh4/kgisvA+yguqaPAE6XPRTqj7m/8vHb73LMSSdyxFHjqK6qbnD/Nc6/8lJGrz+SBb/+zotPPMVp/3c2lmVxzMknctxpJwPg83rRNA23x8MT057nzz/mM3/eb/znk895/KXnuPGeO1izYiWLfp/PpHse4OpbtwXpDe5bCCzL/sJxRJ+rouzgsZIkSdIO+eI0fHH295QzZNE+P4RuKWzJdrK6kxvVFHQoDNNtc4guW0PEBwWWAkWpOluzHGzJdpKX6YhVbU3ymnTdGiLZaxFwKhSn6azuJINRad+oTtBwB028cSrvH5PG8T9X0r5o784nLEzTWdgzjpWdXKgC+q4NkOAXLOrhxu85MIPSGMUe8WC6FV47NaPBh7hCFvFBi3i/RdvqCK6IPWxfEXYxo4iuEHYo9tJEbpX8DAebmnj4flWiRpLX5I++8fg9KuPmVaPJQ8pmIYPTJhAXF0dWTja/zZoTG9ZbWV5Bh04dAZj78ywOOuxQSoqKWbtyNZddfw0b160HILdHd954YRoVZeWkpKXy9svT6d2/H2OOOSrWfr/BA5n9w0906tqFTes3sGXjJkChz8D+fP/1/xh88HCMSIT7b7mDi6+5cof97D2gH//+5DNOOfsMkpKTmT71JbLb5HD6uefUedyShYs45PCRjBo/jqL8Apb9uZh+gwY02KZpmtxyxTXcM+lhTj7rb5imyYa16+g7sD8fv/UeY48dj8Pp5KkHH+XwI8cSCgYpyi/k3MsvZuCwIUy84DK81dU8eOtdPPj0E3Tv3YuyklI2rd+A2+MBoE37dnirq1m7ajW5Pboz96dfyMjKIiFx/6hsJ0mS1FKEXSpbogd9miFoWxjGHRGUpOh8d7ALhCCz3KAyUSPssIPR5GqDbptDJPosgk6FwnQHKzu5sWQwKjWTmkJJzrDFp2NTGPNrNf2auFCSqcDaDi4W9vSQn+nEEbHovzpAXEiwqIen0XOqDwSWpqBHLHqtDxEftBCAoUFYU4g4FPweDa9HJS+reQttVSVoJHoNlnf1EHCpHPdL5QGxBFdLI4PTJnLVLTcwfeqLfPz2ezgcDq6/63Z0h53RC4dC3H3dTRgRg0smTiAxKSm2XWp6GuddcSmT7rkfy7Lo1qsno44aW6ftU845k5enPMvtV11HdpscklNTcLqcjDvuGAry8rnr2hsxTZNRR42jz4D+LPtzcYN97NS1C6f9/WweueMfWJZFp65dOPHM0+s9bszRR/H841OY/cPP6A6dHr17UVRQRN8G2tQ0jb+dew6P3nUvTqeTpJRkrrhxIskpKWxct4F/3HArlmUxcOhgjjhybKwg0m0TJqJpGn+/5AKSkpMZe8x47rn+FpwuJ23atWPU+COZ9/MsABwOB9fefjNvvvAyoVCI+IQErr395j18pyRJkqTGMHWFvGwnYK9ZmV0SIT4oCDqh+4YgiQGB32UHoyu6uLFUGYxKLYepK3g1lRSvyXeHJFGWonPYgr9eKCngUlic62FxDw/eODvbNmKRF79LZXlX9w7XfT3QGQ6VFV1afhGo6gSdeL/JxrZOZoxL5eQfK/CE9m4KNeSw170tSdHpsTFI7ubQAZ21VUQLGwdZWurFsup2qaBgIzk5nWLXm2udU1/ET1mwvMFhvDvy0pRn6N2/H6OOavwaqNv75fsfyMzOpmff3pQUFfPQbXcx5bWXUFX5Abg/85aW8ML6d5u7G5IkSXUJe7idkMGo1EokVxlUJul0ygtx7KwqXJHdP/QtTtFZ2NPDys72EPX2BWG65oXZnOVgQ1un/H/Yz7iDJhGHSqLP5NSZFSTthbnLwWhQurCnh7BTxRMwCXg04v0mfdcG6b8mQEJg9/erejzkXHYFCQMGNXmfm4KqKqSnJ+zw/laZOd1ZALk/atu+PdOnvohlWSiqwsXXTpCBqSRJktQ8FAUhj8OlVqQySSfRZ7A5x8kH41M56cdKUry7Hq9pKbCunZOFPePYmu1ENwS91gdIrRYs7+LipyE7PsCWWregW8MZtvB7VD4Yn8opMyvIrGiaMb7bB6W5m4O0LzJYnOsm2RtBAX7tF8dvfePI3RJiwKoA7YsiB0yRplaZOW0ue5I5laQ9JTOnkiRJktR0XCEToSgoAk7YSaGkoENhaa6bRT3iqE7QSPSZ9F0bIORQWN7VQ9AlEwQHCs0QOA2BqcKJP/214lpBp8KCnvWD0j+7uSlPrpsvTKw2SPVaFKbphFwqaRUGA1YH6LU+uMvq6TJzKkmSJEmSJEktXMiloRqCuJpCSb9V02/ttkJJpUkaC3vGsaKLG0NXaFsYZvAqP1szHMzrFy+H7h6ATF0hoEJ80OKzMSkcPbuK7ptDu9VG0FkrU+qwg9J2RQaLu7lZ26HhebjViTrVidHCdAUhQi6NH4YnMmtQPL3WBxmwOkBG5f5ZrUkGp5IkSZIkSdIBwYoWSkquNvnu4CRKU3Q65IdZ2DOOzW2caKagx4YgGVUmKzq5+GmIXBnggKcq+DwqiT6L/xyWxJjfvAxYE9jlZnsSlG7P1BXycuxK6enlEeJCgmW5Hhb3iKNdYZgBqwP7XQElGZxKkiRJkiRJBw5FoTJJJ7nKYGHPOBb2jCPeb3LQYi+mqrAs18PyXDl0V6pFUahOsKszzzwoEZ9H5ZDFvgbngQaiQemimqB0U5B2xQZ/dm98UNrQ/ktTHZQCnoBFRkWEikSNrw5LJt5v0m9tkH57WECppZHBqSRJkiRJknTAqUzSSao26LkpRGW8xu994uX6vNJOVSVoJHkNfu0fj9+jMua36tjyRIFamdKIrtBtc4i2Jfac0rUdm24ZnYBHZbPHCZYgp9gulDSvXxy/RgsoDVtnktNke9v3ZHAqSZIkSZIkHZCqEnV+6ysPh6XGq0rQSfSaLOnmwee2A9TF3T0NBqVrmjAorUdVKMh0ANsKKG3OdpKfIegVCdJaa0m3yv/G2395gOqwt8nbTXQm8Nhh/9jpY6Y//xIrly3HiBgU5hfQrmN7AI456URGjW/8Wqbf/vtrAI48/pg973ATmT/vNwq25nHcaSfXu++Vp6cy7rhj6NqjW5Psq7iwkIduu5tn3nilSdoDuO7Cy7h70kNkZmfv1nblpWW88sxUbn1g5++5JEmSJEmSJNWoTrDXI13fzsn69hkgxL4LShvqT60CSp2KLAoHOMncpz1oOq0yON0bgWlj27346iuxhIgFWY9OfXqP9tUSgtIa61ev2eF9l11/zT7syb6Vmp4mA1NJkiRJkiRpt/niNDwBk3ZFEdqWNk9Quj1TV9jUOQHiPM3aj7+iVQanLVH+lq289twLeKu9uNxuzr/yUnJ7dOelKc/gdLpYt2o1AX+AU845g8PHjWHG2+8BcPq55zBr5o989v5HKIpC1x7duHTi1ej6trfGW13NK09PJW/LVnSHzrmXXkzfQQOYP+83PnrrHYQlyMrJ5pJrryI5NaVOJnHZn4v55J33uXvSwzx021107dGdlUuXU11ZyflXXkZGdhbfffVfADKysigpKmLNylWUFBVz9EnHM/enXzjt/86mz4D+/OvDGcz7eRaWZdF/yCDOufgCAoEAz096kqqyclTg/846k0GHHkRQVbCi08Q3rF3HK09PBaBjl86x51VZXsFrz71AaUkJqqJy1oXn0XtAPyZecCmPPPcUyakpeKuruW3CRJ554xWWLvyTj99+D9MwyMzJ5tKJV5GYlBRrz7Is3nr5NZYu/BNFUThs7GhOPOM0lv25mM/e+xBN1ykuKKRrj+5cdv01VJSVxbK4L015hri4eNavWUt5aSmnnnMWo8aPw+/z8eKTz1CYl09WTjZlJaXccM/tu52llSRJkiRJkvYvAY/Gmk4aazo1d0/2H7IUWRN5YfLTHH3SCTz2wjOce/nFPPPw40Qi9kK9hfn53DdlEnc+9gDvvvYGFWXlse3KSkp5+5Xp3P7wfTz+0nNYpsXCX3+v0/ZH/3yX7LZteGLaVCbcdD0f/vNtzJIy3nzuBR6+7Vaee+4pevbpxRsvvrzLfpqGwf1TJnHu5Rfz0Vvv0L5jB8YdezTjjj06Niw5Eg7zxLSpHHn8sbHtFv0+n/Vr1vLg00/w8HNTKC8tY/bMH1jyy2w6pWfw+uRJ3HL1BJYvXUZaxKRNyCAjYhBvWrw0+WnOvuh8Hn5uClk524K6f057lVHjj+ThZ6dw47138tpzLxAOhTn4sJHM+2UWAL/OmsOwEYfg9/l4/41/cvuD9/LI1KcYMGQw70//Z53n9t1/vqasuITHnn+aB556gl9nzWFB9LVctXwl511xKU+8/DyRSIT/ffGfeq9NaUkJ/3jiEW78x528+9obAHzy7ge0adeWx196jtP+fjabN2zc5WssSZIkSZIkSdLuk5nTJhAMBCjMy2f4yBEAdO/Vk4TEBPK3bAVg1FHj0HWd9IwMevTpzcply2Pbrl6xkh59epOekQHAVbfcUK/9FUuWcvWtNwKCXp06Mu2xR/j119/p0b07bbKzcBgm54wdyzkffoJT7HyhowFDhwDQvlMnfNUND2PO7dmj3m1LFi5i7cpV3HXdTSgCjHCYzmlpHD9uLG//810Kyss5ZNhgLjjzdEKKglAUdCFQysqpLC1jbP9+BEyL0UeO5Ydvvo21mbdlCx+//S4ApmlSmJ/PyLGjePvl6Yw/8Xjm/PAzZ17wf6xZsYrSohIeuuNuACzTIiGx7tpjSxct5vAjx6JqGi5NY+ToI1i68E+GHDKcXv360LZ9OwAOGzua77/6huEjD6mzff/Bg1AUhQ6dO+Gtrrb7uGBR7D3p2qMbHTrLU2OSJElS6xBnWpxTUMlGt4Pv0uIRiqxEK0lSyyaD0yZgNRAQCgSmaQKgadq22y0LTd2WsNY0rc4aSVWVlQAkJSfXeYwDyAqbOIVg7dY8qlVQAIcQhBWFiKJgmSZZYQMdBY9hoSIwDbNOvxxOu6qXooDYQSDrdDrrP0fL4oSTTuCck07EbVlU+3xEdB09Po7pzz/Dgj8WMPv33/n0sy9489mnUFUVARiKigVoAlIME9MUaAISTQthWtz16IOxILO8rIzk5GRUTcNbXc3aVaspKymle+9e/D5nHj379uame+8CIBwOEwoE677m2z0fgcC0dvAeaPUHDWx7bba9I6qq7vB1kiRJkqSWymVaXJxXQduQQfdAhJywyQfZSURUGaBKktRyyWG9TSAuLo6snGx+mzUHsLOhleUVdOjUEYC5P89CCEFxYRFrV66mZ78+sW1ze3RnzcpVsaG+b788nT/m/Bq7XxOCQX16s+D7H9EELMnP5/YHHmJ4bi7LV60iv6gIpxD897//Y1C/vgRVhaSkRMrWbyQnZLBk9twGFwiuTdU0THPHi/a6LcGhffoy6/sfsHw+KhHc8dgTzJ01h+/+9SVvv/s+Y0aO4IbLL6W8shJfIADYwXNGYgI5mRn8+tvvhBWFr6LDdZMNk6H9+zHn83+TZJgUjEvT9AAAceNJREFUbtzE7ROuIxQKAzByzCimP/ciI0YfDkC3nj1YvXxlLBv92Xsf8s5rr9fpZ9+B/fn5u5lYpkkoGGL2zJ/oM6A/ACuXLqespBTLsvj5+x8YOGzoLl4VW7/BA5n9w08AbFq/gS0bN0WfmSRJkiS1TA5LcFF+JTkhgxlZiRQ4NPr4QkzYUk5yxNx1A5IkSc1EZk6byFW33MD0qS/y8dvv4XA4uP6u29EddiYuHApx93U3YUQMLpk4oU4Rn9T0NM674lIm3XM/lmXRrVdPRh01FhVBgmGRaFpcevaZPPbCNC654SYcqsrd111LWkoKN0+4grsfewLDMMjOzOS2qyfgtgSXnH0mT78yndc//IiDBg3EaQmywgaqaPhsRK9+fZk25RmSU7dla1UEcabAaQkSTYv+Bw1j5caNXHnbnQjL4uDBgzhuzCj8gQAPTHmaC6+7EU3TmHDBeSTGx9dp/+7rJ/LYc8/z2rvv07dnDxQgoihccfklPPvCS1x37Q0IIbjz+mvJcToJCMHIMUfw8Vvvcu3tNwOQkpbK5Tdcw7OPPYFlWqRlZHDVLdfX2c/YY48mf2sed1xzA6ZhcOiYUQw/9BCW/bmY1PQ0XnryacpKy+g/eCBjjj6S0pKSXb6vp5xzJi9PeZbbr7qO7DY5JKem4HTVzyxLkiRJUkugW4Lz8yvoGIzwWWYix5T6SDQtqlWF9IjJNVvK+WebZDa7Hc3dVUmSpHoU0cLGLJaWerGsul0qKNhITs62uX7Ntc6pL+KnLFje4DDeHXlpyjP07t+PUUc1bg1UJRoUJhkWGgK/qhJRFRINc4/S3CYQURV0AboQWIBfU/FpKpEG5p7oQpBgWsSZFioQUhQCmorLsnBbYq/kDA1FwVBAFfYwZSV6W1BVCKgKYVXhr+y5dsXi3fXL9z+QmZ1Nz769KSkq5qHb7mLKay+hqnt/0IG3tIQX1r+71/cjSZIk7R9UITi3oJI+vjCfZyQwqsJPirFtZFQYCGoqHsvi46wkFiU277IXkiQ1Pbfm5qK+59Avo3dzd6VBqqqQnp6ww/tbZeZ0ZwFka+ayLFIMKzaPtErT7EDR2PGQ213RAM0SCCCsKAgF4k2LBNMirCj4NBW/puAQkGhYuC17XwFVJawqxJkWKcbeHQKkC4EejfdrAlWFmn6CBQRVlYBmB6x7J0RuWNv27Zk+9UUsy0JRFS6+dsI+CUwlSZIkaXcoQnBmYRV9fGH+nR7PyMpAncAUwAnopkW5rnFOYRVZYYNvZaEkSZJakFaZOW0ue5I5bQxdCJINE48lMBWFak3FbVq499Jbs302VWAHgzVZVQuItwRaM/9p1PQTwGkJVEAAQVUhqKoENQVzP57/KTOnkiRJUqMIwWnF1RxUFeR/afH09wbJCe/8xHKxQyMzYrI43sWHslCSJO03ZOZU2mMq9vDdeNNCAFWaiibsYkF78yti+2yqpSiEFXs+as1w3pagpp9gB86haF+dQuAxTIRh9z+gqQRVBUOe+ZWkA4sQpEdMugUi5AbCZIUNyhwahQ6dQpdOkVOjyKFjyINuaX8mBCeUeDmoKsiPKXH09oV2GZgCZEZMynSVfr4QaVvLebNNMlW6tsvtJEmS9iYZnDYDBUG8KUiKBqF+VcVUIGEfB4YK2OuiCkFLn3WiAq5oX2uCalMBPbpEDdhFlgKqQlBTCSsgq+o2LNEw6eML09cbwlLgtyQPy+OdWDK4l1qBRMMkNxChmz9Mt0A4NmyxUlPJd+mkh016+MKxLzcLKHNoFDll0Crtn44q83FYZYA5SR46BsK0DxmN3jbNsAslZYRNrt1sB6hbZKEkSZKakQxO9wEFgUPYpd2dlsAlBLoQsWJDCaaFbrWo0dUt2rag2h7mG4nOU61ZPzXJtDBrAlVVIbSP56m2RKkRk77eEP18IToGI6hAiUPDYQnOK6ikUlP5NdnDb0lueeZcalHcpkXXQIRugTC5/jDZ0WUw/KrCOo+D35McOCxBu2CY3EAYVUCprlLh0AioKqaq4DEtGbRK+6Ujyn2MK/fzR6Kb9LBBl90ITGskWoIwIFC4Yms5H2Ul8acslCRJUjORwWmTs4v7OIUdiDotEatAC9vmUXo1Dc8+KDa0v1OwK/w6dlJQSQDhWKCqElE4IILVzLBBf2+Ivr4Q7aIHLPlOnR9T4jBUhVxfiA4Rkw1uHQWFo8p8jC3zsTzexdxkD2s9DlkkQ9rndEvQORghN5oZbRcyUIGwAhvcDpYkuFCEQnY4TG4gQj9fuF4bWYZF1naFYEwaH7RW6nY1c0sBU1EwASs6WqPuZfsxVvQxZp3L9mPC0RNkYcX+/AlFr4eU6O3R63LkgrS7Dq70c1ypj8XxLtymSY9gZI/bql0o6e/RQknfyUJJkiQ1Axmc7qbpz7/EymXLMSIGhfkFtO/QHgU4+cTjOW7cmFjhHiA2/NQXLTKksi17+t//fAXAyceMb/I+en0+Hn3ueR6+/VZKysp4/PmXePyeO5t8P3vi8edf5KSjx9O+TU6sj4311fczWbhkKXdMvGaHj6ld+bfmRIBAiRadsgALE2IHiUFVwdxfvnyFoF3IoK8vRD9viKxohmmjS+fb1DhUAd0DYUZX+OuE5p2DduBaqquUOnW6BML084UocWjMS/LwR5Ibv9ZSZiJL+xtFCDoEDTszGgjTKRBBx/7/3ezW+TnFQ0RVyQxF6Bo06BHYswNwjcYHrY7oslmqsD/Pa35r0SJyGnbBOFVE58bHLkd/C7umgC4aP7kgotjz6muC1ZCqRgNaBb+mMD/RzUaPXGNZsg2pCnBqsZcVcU4iCIb49zwwraEC6YZJsa5xZLmfrLDJR7JQkiRJ+1irDE7X3jgRs6qqydvVkpLInfLsjh8gBFdOuByHZVGWX8Rtd/2Df055wr4LMIS9BIsZzdzpQuC0wGXWXwpmbwSlNap9Ptas2wBARlpaiwlMAW69egIA+UVFsT7uLdsKKok6w38VwGUJ4qLL5kQUZVs2Q1WwWlFWVRGCTsEI/aIZ0lTDDr43eBwsTHTjtix6+sOMK/fv8lmlGxbpRpgwsM7twCkEx5d6GV/mZXGCm7lJHja5ddhfgnmp2Tgti+7+MH18YXr5QsRHpzXkO3XmJXvwaRqpkQhdAwZHVAT2eoG4hoLWhgjszKoVzZrGMqWKnTUNK2rd+xAYikJEVaK31dwOpqIiFGLV0lUEmgU6At0SOIXAbVkkG/aJtYOrgmxwO/gpJY7l8U6Z0TqA9fMG+VtRNWs9Dio0lUOqg03afqZRt1DSP2WhJEmS9qFWGZzujcB0V+0KIXBVleAW9gFMtRAIBaqjGaX8LXk89/yLeL1e3C4XEy+9mN7du/Hos1NxOp2sXLMWnz/A+WeeztGjR/H6+x8CcNHZZ/K/n37mrY9moCgKvbrlcstVV6Lr296a19//kOLSMrbm51NQXMwJR47jvDNOx+f3M2nqixSXllJaXs7QAf259eoJPPvqdErKy7jrsce55uILuf7u+/jg5Rcoq6jg8akvUlhSgqZpXPZ/53DwkME7bH/tho1MfnEapmnidDi4/dqrad+2TZ3XZdSpZ/Djpx8BdTObZ11+FeNHH8GvCxYSDIW487pr6Jmby3V338uFZ5/Jh//6ItbHh2+/la9n/sjHX/4bYVn0yO3K9Zdfisvp5L8//MhbH80gLi6OnMwMPO49mwez/fBfwbbqvyqi7hDgaLAaVFUiauOHANfMbFVEzcGmHTwq0f2rtaYV1xyUxq6z7brAjqB1IDtkRA+A6w4hbBuK0Ncboq8vTKJpEVFgjcfJr0kOEgyLXv4w4wO+PXqtnEDX6PCwAodGta7RxxtiSHWQPKfO3GQPCxNdhOV6r9JuSDJMevvC9PaF6Oa3h9H6VYWVcU6KXDpJYZNOIYMRlYEWUzF8ewo1mVJw1Ex6r/Of3LQsop9HCuS5HKQZJucXVFLk0PgpNY4Fie79Z+SH1Cg9fCHOLqhis0snz6lzeGVgr+ynplBSZtjgms12gLrbhZKEXWMj0bBINK3Y7wTDIqgplDg0Shw6pQ5NzvOWJCmmVQanzUFRFCxXHP6IH1XYBw1KtAAPwM1Tnub/TjuVI0YczNKVq7j3iSd5+3k7C5tXUMgLjz1MWUUll998G8MGDoi1W1xayvPT3+TlyZPIykjnoaefZc4f8zn84IPq7H/dxo089/ADeH1+zplwDaccdwzz5i+ge5fOPHDrTUQiES6YeAOr1q1j4qUXc/3d9/Hw7beSX1QUa+OZV6YzuH8/zjr5RPIKCrnmznt4dcrjO2z/wy++5MyTTmTMyBF8/f0PLF21ql5wujNJiYlMe+IxZvz7K97++FMevO3m2H21+7h+02a+/N+3PP/oQ7icTl5+6x0++PwLjhs3hpf++TavTXmCpMREbn/o0T0OTrensK36L0SHACsKQlHQhIgVVrKAUHToryKiASei1uWaYFQ0+QG1MCxu2Fy2w/tDin1gn+fWSQmb9PGH6e2vP//ur8iJmORETAIKrPY4SDQtTiuu5rgSLwsS3cxN9lDo2r2PEUXY87CdlsAVzRDVzM/2aSp5Lpmd3S8IQZuwQZ9oQFpTQbRU15gbzY62CYbp7Qsz2Btq5s62TCrgFgK3gORgBBN7mL5bCP5WVM34Uh+zUjzMTfIQkkPv93tdAmHOK6ikyKmzJs7FuPI9OwHZWImWoOY/s3ahJFXYJ3RjQWfty7HfJommhbOBczcm9kmeGjXzvEscGiVOnRKHRrFDo8SpUaFrcj621OQclqCfN4hDwO9Jbvk31sLI4HQ36A4PiSE/CIuqWvk0fyDA1oICjhhxMAB9e/YgMSGBzVvzADh23Bh0XScrI51+vXqyePmKWJtLV66iX6+eZGWkA3D39RMb3Pfgfn1xOBykpiSTlJiAz+fnyMMPY/mq1Xz0xb/ZuGULldVeAoEgSYmJDbaxYPESbrnqCgDa5mTTu0c3lq9avcP2RwwdwtMvv8avCxZw6PBhHDps6G69XgcNHgRAl44d+GnuvB0+bsGSJWzNz+eq2+zhxxHDoEfXrixZsZJ+PXuSlpICwFGjDmf+n4t3qw+NVTNvrCZYjUSH3wkUHELgsixENHMpsDOfBtuynEJRdpg/qcmkbp+DtR9f/wNRRG8OqQpLkzzR6wrbBgAK/JpKStiglz/MAN/eP7D3COgeiCCArU6NgKYyrDrAiKoAG9wO1nocsYDTKaJB53aBpyt2/873VaqrLEp0syjBvduBr9S8NCHoGj1J0scXIsWwT/Bsdul8lxoHQBd/mBGVAeQgwd2nAZ1CRuz/0FIUji31MabMz7xkD7NSPHL45X6qfTDChXmVlOsaixJdHFO6dwPTGi7AUatQ0okl1cSbDZ+MDagKXk2lSlPZ4nIQ0BQiql1zQxPgsiwcQhBnWLiE/T0WUhQsVcFpCZINk0HVQTy1Vi8wsKtqlzg1ih16NIDVKHFo9si1AzioUGt917qjJ3tdloVb1Fy2V4bQRO358/YUAjV6gr3u9e1ur7VNmUNlVkocxc7W+52sCEGXQIQh1UH6e0N2ggIYWh3g/exkyh3ys7OlaL1/ZS2IEA0cbQuBadoFaTRVq/NYTdt2Xdd0lFofrhWVlQCkJCfXac7p3FYIQ0FBIJjx76/4cfYcThx/FEOPO5b1mzbvdICZJbabUyXAjGZ+G2p/9KEj6NuzB3N+n89H//qSOb/P59arr2zw+SuKgrFd5WGn0x4CpChKLOhrsF+mxeiRh3LdpRcDdrBvWhbz/1xc57Wt/brtbduGANvzVY3oe6SKbZnTmoDzr9nx9l5LcEjV3hmytacUoF3YBEyqVbt6apphMa7cHx0OvW0ObzhakdTrUKPBvoIRC/rtgJ6aIdDCQheCOFMQbwlGlfsZW+6nwKnxZ4KbRQkuSlvxl2JTUYRocXMN40yLnr4QvX1hevrDuIQgrMCaOCfzkhzEGxY9AmHGNmLes9Q42/4PoUjX8Okqh1f4GVnhZ2Gim59S4yiS/y/7lCrsAMFtWbiio0CqNbVJ/l9zQgYX51Xg0xTmJcVxQmn1Pv1fqimUtNWpU+iKFg5TFLu2hmUHmk5TkGjZQ3Y7hCK7PAnZEAFUqQoFbh2/qhLWVAT2Z0xaxKS7PxybmgP26KEFiS6+zEjcb4YFx5kWvX0hMsNmNOC0/55ctQLOmtt39zW22DZn3opNE4perrldUWo9LvoYYGh1mIOqgiyLd/FDalyrWgs3LWIwtCrI4OogaYZFUFFYnOCiyKnTyxeibchg4uYyPslKZHGCXEKpJZDfXk0gPi6OttnZ/DRnXmxYb1lFBV06dgBg5qzZjD70EAqLS1i2ejW3Xj2BNes3ANCrey5PvfwKpeXlpKemMnX6mwzq15cTjhq3y/3+vmgRJx59FEcdcTgr1qxlzfoNWJaFpmqYVv0laob078e/v/0+Nqx38YoV3HjFZazduLHB9u+bPIVxhx/GSUcfRaf27Zg6/Y16j0lOSmT9ps106diBWb/9TnJiQqNes9p9HNSvL+9//gXn/+10UpKTmDLtFdrl5HDyMeN5+pXXKC4tJT01lZm/zCYhPq5R7TelmvmqUl2JFiQGIljARpcGihLLkiaYFq6IXa10TwYblmkKxU6dOEswvszH+DIfW1w6ixLc/JngovIAOsMZb1gM8gYZWhUkO2xQoauUOO15WqUOO4NQ6tAod2h7b/6hsE8apEZMUgyT1Ii9DFabkEGnYAQNqNJU/oyeRMgMRegZsAseSXtXlmGCYVKhKZQ6dQZ4gwyrDrIszsmPqXGywu8u1GSfHKL2gb8dZLotgdusdblWYLD99YYCBQMod2iU6xrlDpWy2GWNMl3Dpym7zPxlhA0uySsnoij8mBrPycX7NjCtrV3YoF1499dRbSwFSLYEycH6+zCBcl2lQtdiGVm3JTikKkiHkME7OcmUtdLvheSISV+fXdSwS8Beh9wAgtGTvMHo6gJeTaVMtwus1ZzotbBHbSmKffJcs6xoNXH7uEW3amVFUepUH6+pMK7EsqUCR3TFCaVWhXIVQUiBQreD3Gg1/zUeBz+kxrPG42iR2WuXaTEgWiujS9A+TlnncTA3OY7McIR+vhDDooXEQoBX1/i/girmJYX5MiNRVqhuZjI4bSJ3XT+RKS+9zPT3P8DpcPDgbbfgcNhnlkLhEJfffBuRiMHNE64gOWnbsNuMtDSuveQibnngYSzTok/PHhw7dnSj9nnGCcczZdorvDPjUxLi4ujbqwf5hUUM7NObrIwMrrvnPm6/9qrY4ydeejGTX5jGV9/PRFEUbr1qAulpqTts/9zTT+PxF17izQ8/xqHr3HjFZfUec8V5/8cdDz9GWmoK/Xv3orKRxarSUpJjfXzmwfu48KwzuOEf92MJi25dOvP3007B5XRy3aUXc+O9D+Jxu+jUoX2j2pb2LRXoFGra9XrTTEFadMmQ4ugSN0mGxfGlXo4v9bLB7WBRgovFCW68+v43104Tgl6+MEOqA/TyhdGArS6dX5M8JJsmSYZFp0AEd62TJhb2wVupQ6N0u+C1bBeBqyIESYYdcKYaFikRk1TDJMWwYgHp9gffIcUuaPJLioewqtIhEGZQdA6PtO+lmIKUQASfAps8DjoFI0zYWrHfV/hNMkzahQziTKvOVILt57O7al2u/ZjGHgSFlJq1su3fflWhQtcIq2Aoaqxys8A+6HdZFqoAt2XPwWznjcSqUtcIK8QC1ZogtsyhxgJYtyW4NK8CRcC36QmcUlzVYouF7W0akGFYZGxXVbtSU0mPmFy7uYwPspNYEe9qng7upozwtmXfOkTn4xc6NH5O8RBSNToEQiSZ9t9tkmnGTvY213+wQ0BCIEIIu5p/Vtjk0rwKtrp0fkiJY0mCq9k/XxQh6BYIM6QqSD9fCIeAYofGd9Gl9Pr5QhxX6q23nQtwRUwKHBrDq4J0DkR4LyeZAjmtqNkoosExqc2ntNSLtd0HeEHBRnJyOsWuN9dSMlbQi/CWwvbDY3fi0WenMqhfX44dO6YpuigdQAorykmcuZOljQ4wAih02gdwaRGT7IiJBaz1OPgzwc2SBBeB1lwURgjahgyGVgcZVB0k3hJUayqLElz4dI3eXjtDoGAHolWaQll0bc6wahfzckXnbaVHzDrztmoXHCl16Pg1heRagWeyYdWbA+pTlegBs4pXs4fYgYLTtIebJRomqYYg1TAP2APmliwMbHHbFX5TDIsih8bPKXGsjXNQrmvNfiC5J5yWRfugQYdQhI7BCO2DBskNLNVmUXd6Qc2Ug7CyLetU81MzlBGiB/7RbFNNHQI9Gsi6Ra2sqmkHuI39u7cAr6ZSrqv4o32KqBqaEHgsiyTDItWo+z9bs11IVfgyI5FTi6pkNmEHwtiZrzTD5PvUOP6XFt/y/r6jn+99fSH6+UJkR4flb3HprIxzAgrdAiE6BI1W8XlqYPc90RSkGyYlDo2fUuKYn+je50OsM8Pbhu0mmxYBVeHPBBdlDp3uviBdd+M1rVJVVOyTAv9OT2BusqdFZoZ3xa25uajvOfTL6N3cXWmQqiqkp+94pGWr/Kzb6VqkkiTtlxQgJ2ySE7aD0q1OnSpdJStscHpxNScXV7M6zsmiRDer45z4owFbS5dgWAyqDjK0OkCbsIkBLI93sdntoH0wzMFVgXrZSJVopsysP/zNAqprz9uqCVxN+wz8AG8QdzTwLXeobHI78GsqEdWuSO02LVyWRYppkRKxaBtqHQdLUl01S0LVrvB7enE1YGcBC50a+S6dQqdOvkunwKm3qJM7qhBkhw06RIPRDsEIWeFtJ0JKdY31HgdlDg1NQJJhoAs7oKwpvFaTIU02LZyi+TJPKpAUrQDfEBOo0lUKHRq+6P9izUiHzW4HJxfLwHRnnECqYVLo0Bhb7qdDMML7Ocn4mvnvuWYd8r5eOyBNjRaI2+B28L80Nw7Looc/0irn4+tA55CBBWxy6TiF4LTiao4s2zcVxD2mxUBvkCFVQTqGDExgdZyTWR4n2aEIg6pDuMTur/+bZFkYQKlD45QSL90DYWZkJeHf239LQtAlGKFNyCDfpbPVpR/Qy/W1ysxpc9mTzKkk7SmZOW0cE8hz6Xg1lTYhg5ToAaAJ+DUFv6biU1V8Ws2PUuuyil+z5/L4VXWfnfHVhKC3L8TQqiA9/Paw3S0unSXxLhIMiwG+0A4PZP8Kgb02c6mu4gRSIiZxVmNX8pVaM4H9N1bq0FBQSDJMssMGcbW+bys1lQKXTr5TpyAasBY79+Jc5ljnBMnRQjodggYdghHa1yqq41cVNrsdFDp1IopComHSJmzQJmTIoE2qo0xXSTItvJrKOznJbN7HhXs0Icj1h+nrC9HHFyLRFBjYBeI2uu11yHsGwmREmnYqTHMTQH70s6JDyCCoKsxJsiuIe/9CBXG3aZEeMcmI2COC7MsG7YL2/36BU2Nxgh3oD/CGSDOa7nuzyKGRHjHxairv5ySxfi/M308wTIZWBxleFazzN2FhjxTb4nKw2e1gi9v+PG7skjetPXMqg9PdIINTaV+SwenuM7DnZpY6dSwFdMseCuix7HX54kyLOGvHw/FCilI/eFXtANcfDWz9aq3Lmtr4A3chaBcdtjswOmy3KjpsN6CpsbVAZaAo7SteBYpdOj5VJaKpOC27Kmpm2IwFfSZQ7NQoqJVhLXLq9vIg2BlOLZqxrBkKu/11VYBeq8BKzWOcQtA2GpDWnIwxgHyXzhaXA7+u4jIssiIG7UN1A2lJ2pGAAhFVJc60+DIzgblJe39oZptQhEMrAvTzhfBYwl6HPN5JnksnNWLSyxducAj6/qimgninYARTgT8SPfyU6qHM0fCpJE80AN0WhBqkh+3rCdv9z1dq9vSUApdOtabR0xek81783vSp9t9SkmExMzWO79Li//KaqKoQdPeHOahqW02J9W4Hy+NdZIUixFmCiKoQb1q0CRmxueoRxT4Rv7kmYI2ebGzob1sGp01MBqeSZJPBadMR2PPNfJqCT1UIRtfXM1UwUDGjGVMtWq3QbdlBrB3Qith6aA0JKQr+aEBbO5CtuezTVJINkyHVQXLCJhEFlsW72Op20CG6JqjM/kgthYWdfSpzaARVFUtRiDNNMiN2saymVuLQ2OzSKXdoqEIhNRKhfcggzbDkiRppj1nYf1tZEZMFCS4+yUpq8gqsihD08YUZWeGnazBCWIHFCS6KHQ4ywxF6+cP1imAdSMo1lTKHTqdgGBX7tVkZ5yTVsMgIG7GAdPvXqEJTKXXahcL8mopQ7Oq7caZFmmEHsvvyRJWdxdRpEzbY4HbwfnYSFXtQGTo1YjKsKsCwKnturFdTWJDgJqCp9POGaNtAJWy/Yu/bG60TkBwxaBM26owqsbOrOlvcDja77MfK4LSJyeBUkmwyOG1esWHB6rYiJmFVwUTBrDUnrCagdQmBJ5qZrcnQ1rbJpbMsOmx3oC9E4gFyFl3aP4QVe5hblW4X4QpoKgrb5nAqCFSxbQ1oez1oe9VsBXs5CoSCJrZdVwXEW4IcOTxX2kuKo0Mzi5wab+ckU9IE6/+6TYthVUEOrfSTZlhU6Cq/Jblxm4Kh1UGZ4d+OV1UodOq0Cxm4hbAL+kUL9JXHAlAFl2kSXysA3b5AWHMr1VUSTIGlwIysRJY0Yk1UTQj6ekMMrwrQPbr03mqPkzVxTtoFw/Tdbu3eXalZC7jIZdcIUAWkRwyywmasqGGFrrIiIY42Y66iX2afPXmqe91+WRBJkiRpb9OARFOQaDZ+bpCFvTadX1MpdkBIVQkrCgFNoW3Y5Ogyn8wGSa2SU0D7sAnh/WuunLR/y4yYVKl2dfJrNpfzUXYiSxsRVDQkI2xwaGWAoVVBXEKwwe1gdkocHf1hxpb561U8l2wJliAhGCGowJ/x9rzNNMOifcigW3TJuNYg3bAIAT5d49xdrImaHTIYXhVgcHQKT7mu8n1KHAIY6AvRs4ElbRojthZwrdfNAkp0O0sdik7PcJgRdG8pZO7RbpqdDE4lSZKaiAp25tQyyQDs/KskSZLUXJIsQQSo0jXOK6jixxSD/6Y3cu6gEHQPhBlZEaCXP4yBPTw1z+1gYHWQE0r2LMg4ELkFDPCFm7sbf0ntNVEPrgrSKRjhvexkCl06zmhRpuGVATqFDAzsKTyb3U66+IOMqtg7JzBUIMuwyDK2vbaW00O8lrgX9rZv/KXgdOrUqXz11VcAjBo1iltvvZXZs2fz6KOPEgqFOPbYY7nhhhuapKO1vfHcbAK+pj/b4ol3cOG1h+70MU9Ne4Uly5cTMQy2FhTQuX17AE4/4XiOG9f4tUw///obAE4+Zvyed3g3XXf3vVx49pkM7td3n+1TkiRJkiSpOTmAdMOkyKExqsJebubdnKQdVpJ1WILB1UFGVvjJjphUayozU+3M17CqIIO9oX3af6llyYmYVGoqiYbFNVvKWB7voqcvjEsIihwa36TF4zIFg7xBBvj2/d9Ka1+EZo+D09mzZ/PLL7/w6aefoigKl156KV9++SWTJ0/mrbfeok2bNlxxxRX8+OOPjBo1qin7vFcC08a2e8MVl4GwyC8q4vq77+O1pybv0b72ZVAqSZIkSZJ0oMuKmJRrKh1CEa7bXM47OUlsqLVESHLEZERlgIOqAsRZgjynzn/SE8gORTiswr9b8wOl/Vuyaa+JWuLQ6OELsyTBRaFTp4cvyJFlvlYfIDanPQ5OMzMzuf3223E67X/q3NxcNmzYQKdOnejQoQMAJ554Il9//XWTB6ct0eateUx+cRpVXi9ul4uJl15M7+7dePTZqTidTlauWYvPH+D8M0/n6NGjeP39DwG46Owz+d9PP/PWRzNQFIVe3XK55aor0fVtb83r739IcWkZW/PzKSgu5oQjx3HeGafz1fczWbhkKXdMvAbYlhkd1LcP0956h5/n/oqmqZw0/ij+duLxdfr7zoxPmTl7DpZlMXzQQK48/1wUReGVt99l/uIlVFV7yUhL5d6bbyAtJYWTLriYnrm5lJWXM+2Jx2L9u/C6G7nv5hvp3KE9D0x5mvi4OG668nKWrFjJWx/PYNLdd+72vk6+4BKOOORglqxcSZzHw903TKRNVhZLV67iuddeJxyJkJyYyE0TLqd9mzZcd/e99OrejT+XLaeyqoqJl17CIUMH87+ffub9Tz9HVVXaZGdx1/UTcTmbfp0qSZIkSZJah1TTIqDYc/Uu21rBV+kJbHbrjKwM0NcbQgGWxTvZ4HHRpzrAsaVeWStAapCOnUUtcKj08oUYWh1s7i7tF/Y4sO/evTuDBg0CYMOGDXz11VcoikJm5rbZt1lZWRQWFv7lTrYGDz39LKcffxyvP/0k11x8Ifc+8SThiJ2JzSso5IXHHuapB+7lxTfeorS8PLZdcWkpz09/k8n33sObzz6FaVnM+WN+vfbXbdzI5Hvv5qVJj/LOJ59R7fPtsC8/zJ7LkuUreP2ZJ3np8Uf56vuZdfY5b/4CVq5dx7THH+XVJx+npKyM//34M1vy89m0NY/nH32Id154lqzMDP73488AVFZV8/fTTuG1pybXCZxHDB3C/MWLo33cxOLlKwD4dcFCRgwbukf7qqiqom/PHrz+9JOMPWwkz746nUgkwgNPPs11l13C9Kcmc9LR43lgyjOxfhiGwYuTHuHqiy/ktXffA+C1d99n8r338MqTj9MmK4tNW7fu3psqSZIkSdJ+xyMg0bQodWicUOplwtYKuvvDzE728G1aPG1DBieUVNNVrj0tNUJOxDqglw1qan+5INLq1au54ooruPXWW9E0jQ0bNsTuE0Kg7OZitQ2VFi4qUtH1fZMg39l+TE2l/ipE4A8E2FpQwBEjDgagb88eJCYksHlrHgDHjhuDrutkZaTTr1fPWAAHsHTlKvr16klWRjoAd18/scF9D+7XF4fDQWpKMkmJCfh8/h32c9HSpYweeShOhwOnw1Fv6PEffy5m+erVXH7zbQCEwmGyMzIYP/oIrrrofL789js2b81j6cpVtMvJiW3Xp3u3evs6ZOgQPvziS4b070eXjh3YtGUr5RWVzJu/gAduvYkZ//5qt/fldDo4eoydbT9mzCheefsdNuflk5AQT+9oH8aMHMHkF6fhjQbpBw0eBEDXjh2o8toFCg4dNpSr77ybww8+iCNGHEL3Ll12+JpJkiRJknTgULGH+RY4NJYluPCYFgdFK/FKUmumKgrJSXHEZbbOokh/KTj9448/mDhxInfeeSfHH388v/76K8XFxbH7i4uLycrK2q02G1rn1LIsjL2w+HdDdrYfawfrEja4VKwQmNElKDRVq/NYTdt2Xdf0OgF8RWUlACnJyXWac9YajmqvHCfs37V2bUT3p+t128wvKiIlKWnb87As/nbC8Zx18okAVPt8aKrKyrVreeDJZzjzpBMYNeIQVFWt89xcLle9p9m3V0/WPvc8vy9azKC+fUlNTuaHOXMxTJPszMw92peqqLH+W0KgqdoOX2PLsuq+PopiLwQFTLz0Yo7fsJE5f8zn4aef5cKzzmT86CPqtyNJkiRJ0gEpJ2KSU77jE/6S1NpYQlBZ5cdXXN3cXWnQrtY53eN0ZH5+PldffTWTJ0/m+OPt+YwDBw5k/fr1bNy4EdM0+fLLLzniiP0/GIiPi6NtdjY/zZkH2NnQsooKunS0597OnDUbIQQFRcUsW72aAb17x7bt1T2XZatWxYbdTp3+Jr/8+nuj9puclMjGLVsQQpBfWMjaDRsBGNCnNz/NmYthGARDIW594GFKSsti2w3u349vfvwJfyCAYZrc/ejj/DhnLguXLGNQvz6cfMx4OrRty5zf/4gFfzuiaxq9u3fjk3//h0H9+jKkf3/e/ngGhwwZvMf7CoZCzPrNfg2++m4mBw0ZTId2bamqrmb56jUAfD9rNtlZmSQlNnxWyDBN/n7VtSQnJXLu6acyfvQoVq9f36jXVZIkSZIkSZKkfW+PM6evvfYaoVCIxx57LHbb2WefzWOPPca1115LKBRi1KhRHHPMMU3S0ZburusnMuWll5n+/gc4HQ4evO0WHA4HAKFwiMtvvo1IxODmCVeQnLQtoMpIS+PaSy7ilgcexjIt+vTswbFjRzdqn0MHDuA/383k3Kuvo2P7tvTv3QuAIw45mJVr1nLpTbcihOD0E46jQ7u2se1GDh/G2g0bmXDbnViWxUGDB3HMmNGUlJVxz6TJXHjdjQD07JZLfmHRLvtxyNAhLFy6jE7t25GWmkJ5RSUjhg39S/v6YfYcXn3nPTLSUrlj4jU4HQ7uvekGnnnlNYLBEImJCdx7046XKdI1jYvPOYub7nsQl9NJSnIyd1x7daNeV0mSJEmSJEmS9j1FNDhesvk0NKy3oGAjOTmdYteba51TK+hFeEtBNH6I8aPPTmVQv74cO7bxa6Ae6EadegY/fvpRc3ej2RVWlJM489nm7oYkSZIkSZLUWjg9eMZegd5xUHP3pEG7Gtb7lwsiNYedBZCSJEmSJEmSJElS69Mqg9PWpGYNUqnxZNZUkiRJkiRJkg48+2Z9FkmSJEmSJEmSJEnaCRmcSpIkSZIkSZIkSc1OBqeSJEmSJEmSJElSs5PBqSRJkiRJkiRJktTsZEGk3fTUtFdYsnw5EcNga0EBndu3B+D0E47nuHGNXy7m86+/AeDkY8bvdh8ef/5FTjp6PL265TZ4f35REdfffR8fvPxCvfsefe55LjrrTHKyMne5n9ff/xCAi84+c7f7KEmSJEmSJEmStDtaZXD6+Ut3EfRXN3m77rhETr7y4Z0+5oYrLgNhxQLA156avEf72pOgtMatV0/Y420XLl6KOKtFLW0rSZIkSZIkSZLUOoPTvRGY/tV2N2/NY/KL06jyenG7XEy89GJ6d+/Go89Oxel0snLNWnz+AOefeTpHjx5VJyv5v59+5q2PZqAoCr265XLLVVei69vemtff/5Blq1ZRWFzCaccfy8xfZnPh2WcyuF9fXn7rHX6YM5fkxETSU1MZedAwBvXrSygc5v4nn2L9xs0kJMTz8O238uX/vqWkvIzbHnyE5x5+kHc++ZTfF/2JoigcfvBBXHjWGfWe1/LVa7jq9rsoKSvj2LFjuOjsM/nq+5ksXLI0tkzOdXffy4XR7OpbH8/AoevkFxYxcvgwPB4Pv8z7FSEEk+65k7SUFD75z1d888NPBIMhdIfOP268jo7t2nHW5VcxfvQR/LpgIcFQiDuvu4aeuQ1nhyVJkiRJkiRJ2r/IOadN5KGnn+X044/j9aef5JqLL+TeJ54kHIkAkFdQyAuPPcxTD9zLi2+8RWl5eWy74tJSnp/+JpPvvYc3n30K07KY88f8eu2HwxH++dzTnHLM0bHbZv32O38uX8Gbz0zh8XvuZPX69bH7KqqqOPPEE3jj2SmkpSTz/S+z+L/TTyUjNY1J99xJIBhk3vwFTH9qMs8/+hAbNm8mFA7X2295RQVPP3gfL0+exPuf/Qt/ILDT12H5qjXcdOXlvDx5Ep9+9TUpSUm8PHkSXTt34vufZ+Hz+/ll3m888+B9vPHsFEYMHcIn//k6tn1SYiLTnniMk44ez9sff9r4N0CSJEmSJEmSpFatVWZOWxp/IMDWggKOGHEwAH179iAxIYHNW/MAOHbcGHRdJysjnX69erJ4+YrYtktXrqJfr55kZaQDcPf1ExvcR+8e3evd9vuiPxkz8lAcDgcOh4PDDz4odl9GWmpsm84dOlBZXTcrnJGehsvp5Oo77mbEsKFccf65uJzOevs4eMhgnA4HToeD5KREqqq9O30tunTsQFZGBgDJiUkMGdAPgJzMTKp9PuLj4rjnhuv47pfZbMnLY96ChXTr0jm2/UGDB8Xa+WnuvJ3uS5IkSZIkSZKk/YfMnDYBIRqYwykEpmkCoKlancdq2rbruqajKErsekVlJRWVlfWaayhw1FQVYVkN9qn2PhVFqddHXdN4cdIjXHLOWVRVV3PV7XfFguk67dTqq4KCQNi/azVnRJ8ngEOve76j9vYARSUlXHX7XXh9Pg4eMphjx4ymdmNOpyPWZxp6XSVJkiRJkiRJ2i/J4LQJxMfF0TY7m5/m2Jm+pStXUVZRQZeOHQCYOWs2QggKiopZtno1A3r3jm3bq3suy1atig31nTr9TX759fdG7XfogP78OHcekUgEn9/P7N//QEHZ6TaapmKaJqvWree6u+9lQN8+XHXh+XRu355NefWD04YkJyWyccsWhBDkFxaydsPGRm0HsGL1Wtq1yeHMk06gV7dcfp73K+YOAmxJkiRJkiRJkg4cclhvE7nr+olMeellpr//AU6HgwdvuwWHw84ChsIhLr/5NiIRg5snXEFyUmJsu4y0NK695CJueeBhLNOiT88eHDt2dKP2OWLYUJauXMWlN91KYkICGampOF31M6zbb3Pbg48y+d676NuzBxdddyMul4t+vXpy8JDBjdrv0IED+M93Mzn36uvo2L4t/Xv3atR2AMMGDeCz//6X86+9HiFgUN8+rN+0qdHbS5IkSZIkSZK0f1JEg2NSm09pqRfLqtulgoKN5OR0il1vrqVkrKAX4S0F0fhM36PPTmVQv74cO7bxa6A21pIVK9mSl88xY0djGAZX3X4Xt11zFbmdO+16Y6nFK6woJ3Hms83dDUmSJEmSJKm1cHrwjL0CveOg5u5Jg1RVIT09YYf3t8rM6a7WIj1QdGzXljc//IgPv/gSy7I4ZsxoGZhKkiRJkiRJktQqtcrgtDWpWQt0b0hKTOSJf9y908cIQKAghGL/BixqLtfcDqoi0DHRFSv6KEmSJEmSJEmSpH1HBqetVEjo+IULU6ixALShoFPsokBSQzTFsgNVLHSl9m8TVZGBqyRJLZMpFDT5GSVJkiRJrZYMTlsRU6j4hAufcGEIDQWBFs10qgg0LBSwM58KKMqOQ1MhtgWu9qGcEttOCAUTlYBwYAlXne1UBNp2AaseDWbtvkiSJO0bQaGzJpLDykgbVkbaUGilkKAEyFC9pGvVpEd/Z6he0tVqktWAPMEmHXAKzSTilDCJarC5uyJJkrRLMjht4QQKAeHAZ7kICrsSr1MxSFb9GEIliDOaPW0gLNyTY7Ba26hYOBUDtSYnG1161EIljEbAql8ZWFdMtFoZV/uyhSaDV0mS/iJTKGw206PBaFvWG5lYqDgw6KoX0tORT5Xw4LXcrDeymG91RtRaMU3DjAasdrBaJ3jVvLiVSDM+O0lqWkHh4DP/UOaEeqBjMty1jtHuZeRo9ddSlyRJailkcNpCRYSGV7jxWy4sFFQsEtUACoKAcFJpxe31PliohEWtpXBrBa4KIhaIqggURYCw57OaqASi/d5ezZDhmqC1JuOqY6EppgxeJUmqo8RMYGWkDSsibVlt5BAQLhQEbbVyjnCvIE4JU2AkscbIYaXRLradizA5agVJahC3Esapmghhn+yrtOLYYGQQ2G5kSLwSjAWvSYqfRDVIkhogSQmQqAZIUgMkKCGZfZVavOXhtrzvH0Gl5eFw13KKrSR+C3VhTqg7vR1bGONeTg89H0V+6UqS1MLI4HQ37O3DEUso+IULr3AREfZb41HCOBWDsNCptjx7uQeNJ1AwhIaBVnNDHXbW1ayTdYVtQ4aDwoFVO/CN0hQ72+rAxKGY6IqBIzp0WJKk/Z/fcrDaaMOKaHa01LLXhU5RffR3bCZDq6bC8rDeyOaHYG/YwSmtEE7yrTTyG/joULBIVgNkqxUkKGGcagQdQQQVr+Vmi5FGldWOEI4Gt01QQiSpdsCaqARjl5OUQPRykCQlgEcJy4N/aZ8KWA4+Cwxjbqg7WWolZ8XP45tAf8qsBOII0kUvZrORzgvVR9FWK2OMexlDnBvkd6wkSS1GqwxOKz9YiggaTd6u4tZJPqtvg/cJIdhaYTH9lX+yeuViDMOgsCCPDu07oACnHn88x48bg9bID/jPv/4GgJOOGU9IOPBarthZfIdikqT6sVAIWC4Cov7wWQDLNHni0TsoLirg4itupE/fQbH7vvvmXwCMG39SI59906qddZ357Ze43B4OPWwc06ZOonffgYweczQOxUDbVsIJsIPemvmuvlpZDQVhB6vRoNWhmDgw5FBhSWrlLKGw3siMZUc3mekIVFxEyHUUMty1jrBQ2RTJYH64y7YTYn+BQKXCiqeC+AbvdxIhWQ2QoJThViM4MaInyywEChGhERBOfJaLAiuFauHGbKBfLsIMd61jlHs5WVrTr80tSbUtC7flg2i2dLRrKSHh4H3fIdScwPHjZrXRBg2TrlohXuHmHd9hfOEfwuHuFYx0rSZeDTXvk5Ak6YDXKoPTvRGY7qpdRVFI9GhcddmVWCgUFBVx37238OATr8Uek2cSK1K0faEgHatOtdvjjz4Wn3CRb9oVdxUE8WoQHYuAcFDViGG7ZWUlbN60nqkvf1TvvuYKShuyasVSevcdWOc2CwVL6OxohpeKiM13VRHRJXDsjKt/u6BVV8xYptURvSyHCEtSy1ZqxjMv1I154W5UWPEoWHTQyhjjWoZLNck3klhttGFZpP0+71sYB8WWg2KSwNzx4xQEcUqITLWKBCWIS43gjH4WqYqg2vIwJ9SdX0K96OPYwmj3MnroBTKbKjUpv+XgM/9w5oW7RbOlc/lvYADlVsOL3JtorDOzAUEHrQQVwb8DQ/gmMICDXWvkyRRJkppVqwxOm4vD6aQqaCEsERt2WxNA5edt5qWXnsXrrcbldnPBxVfTObcPL019HKfTybq1KwkE/Jx6+rmMGj2ODz54C4C/n/V//DHra97/+AMEKl1ze3HJlTeh13pnykqLeeXFyfj9XirKSjl89NH87eyLePKxu6iuquSe267knPOv5P23pmFZFu07diEzKweA08+8kNk/f8dnM95GUYi1X1VZ3mCbtW3auJbXpk3BMk0cDieXX30rOW3as+CPOXz03nSEEGRlt+Hiy28gOSWN6686h7vue4rMrByWLV3IJx++ySmnn8v832ezbMkCUlLTAVg4fy7f/vdzKivLOfm0cxl71An1XmsLhbCo/+epbBe0ArFhwjsKWp2KiSM6PLixme2WyGc52WRksNHMwGe5tlvmp3715Pq/6z/epRg4iaDKg+VmI2rN1bZQMYWKiVLnsins+zTFIkOtbrVD8AyhsjjSgbnB7qw02gDQXc/ncNcKSswE1ho5fB/qy46G6rY0AgWfcOMTbvuGBgLZeAK008vZZGTwQvV42mjljHYvZ6hzHY5W+j5KLcfScDs+8B1ClfAwxr0Uv+Xkfd8IGvc/pLDZzAAgQ60kVQswJ9SdWaGe9HVsYYx7Kbl6UZOcTLGEQpVwU27GU24loCoWyYrfntOtBuT/giRJMTI4/YtqAqinnpnMiaeew/CDj2DNqmVMmfwATz37BqpiUVy4lUmPPElFVTm33DKRIQMH4lLsLG1xaRkvv/4aD056ifT0TF589hEWzp/LsIMOi+1jzqzvGXHYWI4YfTR+n5eJE87m6ONO48bbHuLh+27gwUkvsWzpQgryt/D0C+8RF5/AjA/fAOzA9u03X6jXfmHB1gbbTExKju33qy8/5rgTz+DgEaP56Yf/smbVMjxx8UyfNoV/PPQcmVk5fPn5+7z52rNMvOm+Bl+ffgOGMmTYofTuO5ABg4Yz55fviUTC3P/oC2zZvIFH7ruxweB0R8ROglZ7mPDOg1Y1+jhdMWMZDgdGiytwIlAwUPkp2JONRgYbjUyKrSTAfq4uJYIhVMzowOi/QsHCrUTwKGE8sd+1flT7d9wO7ncpERQggkZQOAgKB6Htfu/ockg4CKLHLitAghIkQQ2SoARJbOh3tChNSw3Q/JaTfDOFfDOFgujvciseMxZs2jOxt12uP/d6Z1QsMtUqsrVKcrRKcrQKcrRKMrUqnMpO0nzNqMBMZk6oG7+FcvEJNymqj7HuJehYzA934YvA0Obu4l7jw8MqwxMbSukTLt7zHcoX/sGMdK3iMPdKkuQSH9Ju8lsOPvUP59dwN7K1Co5xL+KbnWRLd6XESqbESiaeAG31ctYZmSypPoYOWglj3MsY5Ny40zWEQ0Knwoqj3IqnzEyg3Iqv81NhxTU49L1GXHQetz1ve1vQmlzrerIaiB07SXuuZilBwbYlBaPVQez7Ueo9xq1EWux37u4whMoaI5vl4XYsi7QjjM5hrpUc5lqJR5XV2lsKGZw2gWAgQGFBHsMPPgKAbj36kJCQxJa8rVhC5bAxxxLW4olLjad7z34sXr6CUDTAWrFyBT169iU9PROACRPvrNf+8SedxbIlC/j3vz5gy6YNmEaEUChQ73Ft2nYgLr7uF9OaVct22H5DbSayLTgdPOQQ3njtWf5c8BuDh41gyNARLFr4K1279YplZsceeQJffPrebr1eQ4aPRFEU2nfoTHV105S0t+eB1R8mXCdoVUSsIJPfcuGrFdRpWNFhwUadIcI1c2H3NkNohIVOCJ2wsH+8VogZ/oNJVAJ00ksY4NyEhkW5FUehmYKFgkZ0yR7FQsVCU0Qsq6wgUJVtc3q16Hq4Ss0auKJmjq8dKIWFRhidkHBQYiUSsJwEhLPBojC11bTfmCCrJrB2EcGtRHAp9u9ENYhDMbCESkA4qbI85FmpO5zLB3axsJpANlEJRAPaUK1A1i5Mk6gEiVNCTT6UMiR0Cs3kWCCab6aQb6RQKbbNY3QpYXLUSjrqJdvek9h7s+21qzksUBRA2O8P1FqLWNh3hoUWXe/YSYGZwpJIh9jrriBIV6vJ0SqjgWtFLIBtjgO6kNBZEO7E3FB31htZqFj0dWymq6OEdeFMZgb77nZg3prVHkrZXi3BoVp8ExzAt8F+DHWuZ5R7Oe318ubuptQK2NnSEVQLN2PdS/CaLj5odLZ053x4WG140DHI1QuotOL4p+8I/hXwcYRrBeladTT7GU+ZtS0IjY0eiFKxSFIDpKo+Oukl9FFCOBXDXh7P0vELNyomLtUCBBFhn9z0Wm7WGtlUWZ4dzOOORANXO2Ad4NjMYOcGOVQ+yme5WG1kszLShlWRNpRb8bWqemwLQHeXjklnvZjujgK66QV01ktaTbBaasazPNKO5ZF2rIrkEMaBjkmuXoiBypeBIfwv0I/D3KsY5V5Oslr/+HpfMIRKoZmEFT2BbQll2+Xoe2ffZh9JWLWux+5HxRIKGSLEwF3vssWSwWkTsET9f1AhBKZpZzE0Tatzu1rruqbrKLU+VasqKwBISk6J3fbOmy9SVJjHoYePY+jww1iy+A/7YHU7Dqer3m07av+Lz97bZZsHjRhFtx59WDB/Ll9/+TEL589l8NARdZ8nYFr281RQENFGTGPHB8Oaaj9/ZR98m9QJWrdbwzVWkCkatBqoBLeriKzH5rBadQOL6Hb1Ag6igd9OglpLKNEgcFsgWjvAcCgmCWoQUwlzmGsF+UYKqyM5LIl0aPoXqEECByYeJUyK6osFkTUFqWLr2Cr215yJghodRm1nrRUs7A9aQ2iYQsHADqpCQifMtgxqpYgjJPQ6a1GC/f7ER+fyxSuh2Fw+XTFxKBYCO6APCR2/cFJkJrPOyMInXPXaqmkvsVbAGgtetwtiE9UA8UqoTobAECpFZlIsC1rzU2olxr7odQyytSq6OwpIVgPR+YYuyowECqwUNpoZNPVQVQ2TLLWSZNVPnBJGUwQB4aDETGB5pG2dA7tU1RsNWu0sa5ZaRZrqjfW1qQgBm8x05oa680eoMyGcZKpVHONeSBid+eEuLI50arL9tU4KW6wMsCBNqSZd97Ig3Jlfw93ophcw2r2Mvo6tLW40h9T8/JaTT/zD+S2cS45awbFxi/g6MIAKq+HCXn+Fgc5aIwcQdNSKESj8q9YIBycRUlUfaaqPdo4yPGoETbGIWBoBoeO33FSIeArNZNYbmTT+86/uPG539LvHGZ3HHYl+7vuEi/WRLBaEu7Ag3Ikz4ucdkCMQwkJjnZHFqkgbVkbasNVMQ6DECsr1cuRFjy/s7+ZtYep2lJpf0c8dse1moYDPclNoJvF1YCCCQTgw6KwX081R2OKCVUOorDWyWB5px7JwOwqtFMD+HhzmWk+q6qPITGRVpA2VIp5MtZIkNcj3wT78EOzNQa51jHUv2SfzroWADUYGv4VzWRDuXGek31+RGvTR2RchvUla2/dkcNoE4uLiycpuw2/zfooN662sKKNDhy4AzJv9AwcdMorSkkLWrF7OpRNuZuOGNQB0ze3JG688TUV5GSmpabz95gv07juQMeOOj7W/5M8/uOjy6+nRsx8L/phLeVkJlmWiqruuWrmj9nfUZm3PTXmAEYeNZdxRJ9KuXUfefuMFzjj7Yqa//BTFRQVkZuUw89svY1WCE5KS2bplA1nZbZj/26xYO6qmxQLYlqLOGq61jgFrAq+a4NMSij1c1XLs9hnHmjZqB68mdsBWQ1fM2HCZmoArjI7XchMQTn4J9WqCZ7u7FCLoRIROldj76+k2xEKlWnioFtGTBbv483FgkKAGaaeUEaeGcREduh09oDGxM8NBy66Mnb+TCqsKgvjoEGKBQlH0TCbUDKmtpr1WRj/nZpxYeC0HlVY8BVYKf4S7/uVh1o1lolFgpVJgpda5XcUiXakmVfNFl6KyCAqdMiue1ZGeGLU+9lUsUlQ/qaqPFNUXPdj0khq9nKr5cDci6+q3nPwe7sLcUHe2mmk4MBjo3EhbrYIV4Tb8Nzhwn70urUmZSKQskoiHEJ30YorNJF71jiVDrWKUewUHu9bsVtY7JHSqLA+VlocqK45K4Ylej6PK8uATLhKUECmqjxTVX+93nFz6ZqcMoeITLryWy55rbLnwCjde4cJnuWP3+aOvc7pWTaZaRYbmJUOtIl3z7vHQ+yXh9nzgOwSvcDPOvYQq012nEu/eo7DJtEdeZavldNZLiAiNSiuOChHPKiNnp8N192R/fuHGv10mtj6Bkwht1TKWRdrzWGU2Z8TPY7BzYxP2peUxhcJmM51VkRxWRdqwzsjCREPDpJNewlj3UhyKSX4kmTVGDstEUxaUE2SpVaSr1ThVk2IzsV6wamdWC+m0j4PVcjPODkYj7VgVaUMIBxomuXoRg5wbsVBZF8lkXii33t9rsZVMsZVMsuIlU/PyW6gLc0PdGODYxDjPEjrppU3e3xIzgd/CXfk91JUSKwkHBv2cm8nSqvBa9t++FR1eXbMmhdj+t4DaGfGaIdgC0FWFreoQGZwe6CZMvJPpLz/FjA/exOFwcN0t96M77OGQoVCIe267EsOIcMkVN5KYuG3obGpaBudddA2PP3wrlmXRrUcfRo0+pk7bJ556Di899yhOp4u09Cy65PakuKiA7Jx27MqO2ne53Lts86TT/o9XX5rMpx//E113cNFl15OcksYlV9zI00/8A8OIkJGZzaUTbgHg9DMv4J/Tn+OTj/7JgIHDYu30GzCED999jfi4PZsLsy/VzAsE6gStenSIb02QWXM4UOcsZJ1MR/05GwIFHROPGkbBPtCJoNlnymSS5C+JoFNuJVBOAuzi+1Cpk5UN4la3DeXWai0VEhQOVAQ99HzcqkHAsg/ICs0U/ox0xIq0zCGpFirFIpliI7nO7QoWqYqPdM1LnBLCrZoIIGjpeIWb9UYWC6y4ekNt45TQtmC1zo+XMDrzQt1YFO6EgUY7rYzjPQuoNN0siHTh93DuPnzmrVcAF2uMNihYdNaKCKMzw38Q/w4MYoRrNYe6VgFQFQ0yKy0PVSIuGoRGg0/hIdjAsmM6ZnQun30Swme5WBlpQ5Xw1Btl4CQSDVbrB677ewBrCJXF4Q6UWgl4awWesd/C1eDrW8Oelx+yR2UoAaqFhw3hjHrbJCs+MrVq0tVqMrVqMrRqMlT7p6E5bz7Lyac12VKtnGPdC/naP7DO9IF9pdBKpTCcuusH7hMKYZzkWWl4CBGnhHjDO4pFzvWcEffrfrMkjhBQZCXFhumuMXJiywu21co41LWKBDUULSiXzXfBfnuxNwpFVjJFVnL0miBLrSRd9eKIBqtfBQYhUHBg0EUvptteClZNobDOyGJZdLhuvmn/XaaqXoY415Ou+SgyE1httGFVsE2j2qwUCVQa9nrAbfUSVhltWFTVie56Pkd6ltBTz/9Ln31+y8nCcCd+Deey3shCQZCrF3KIaw1lZjyLIp1YEO6y5zuoxePSOFxrmixsc1CEaGiAaPMpLfViWXW7VFCwkZycbUPBmmOdUwBvIEJpVRBhNf4lq1nX84gxx+z6wZJUi7cqnymfbm3ubkgHCI8SJkXxkagGcCsRnNFh7wYaPstJlfBQbiXUW3fZrYQZ4txAuuplcaQDG4ymH8J8IMpRy4lXw6w3Mhucm6tjxArI1AztdqkGKgIThZBlzx/3WS6qsINaf/S9i1dCJCs+EtQQbiUcnWMvMIRCUDjxWi7KrfhdBrA1JypStLonLlpqYa6GmELh13Au3wQGUBYtJqRjEK+EYvPY49QQLgz7f0KxAHv+d0SoGEInKPRo1WY7oxqIFndLxE+65iU+Ot9SVwVBy0GV5abUStw2MiQqXgmSqVaToVWRoXrxqCG+C/TDK9yMdS+lwvTweyQX+f/VEEEbrZwiM5k4JcxZ8XPo79zS3J3abUJAuRXP2lrzRiujI5hSVS899HzSNR+Vln1CMd9MbTGjUrYFq9U4VEGxmRDrnx2sFpGjVcZqXQhUzNicSXuuZE3BQCt2X83cym1V7C0Uysx4QjjRMOmiF9NVL0SgsCGSyTozq0my+Q4idNRLKTaTqBJxtNdKGedeysBdFAerzRAqyyNt+T3UlcWRDphoZKmVDHZuwEJhUbhTLNhvSh6XxuUn9mVgt4wmb7spqKpCevqOE1atMnO6swBSkiRJ2n0BYRfAyrcazozUDAHOUSuIU0I4FINkLUjA0mLzS6WmU2ClgmVn2jrrJWRoXkwUApbDnnNnuagScRRayWwwd2dOH9uWv9lhIkMQr4Roo5bvMICtttwsj7SlWnjqHRzHK8E6wWssiI3+TlIDjT6421tMofBHuAtfBwZSaiXSQStlTNwy1kUyqRYevMJNlRVHvkjdo8JdAqginiqzoQyn/fp21Eqi6+OGcSAIo8UKAv1h2VME2mjlHOdexFf+gbEgRWqIQr6ZRhxBXIrBq96xDHeu5bS434hTw83duR2KCJXNRjobjMzYT837HKeE6K7n00avJGTpbDQz+D3ctYmHUTcdgUKhlRKb46kgyFYrSVOr0VWLYjOJTeEM1Gg5Hw2Bqlix62q0iGPNqgtqtNhjTcFKNVrYUkHQSSshXfNSYsaz2mjLmmBOkz+fCA7WGjmomHTRCqkScbzpO4L0QDVj3Us5yLW2wRNxNbUXfgvlMj/cGZ9wE68EGeFaTYrmZ3moLd8EB7SYkwotUavMnDaXPcmcStKekplTSZJaLkGCEiQlmoF1KRFciokSLVrjF057XqIVXy/bXlPJNV310tuxlaHO9aRpvn3Sa0sozA935r+BARRZydGhkatZFOrIajOHlpCVdCth0pVqkrQAihAsM9q3iH61HoK2ajkFVgqJSpBz4mfT25nX3J2KZUXtIDSDDUYmW8y0WLCZqnrprBWToVcjhEKekcpaI2u/OfGnIMhUq9AVq25WlJq1vBvOokZr2Td397GLg5UQQSffTCVBCTDKvZzDXCuJUyOUmfH8Hu7Kb6GuFFnJ6Jj0dWyms6OUjZE0lkY6ENlHOcHWnjmVwelukMGptC/J4FSSpNZMwyRZ9ZOsBIiLDm11qiYiOuy4zIpns5kRm3s1zLWOQY6Ne2W9QUvAonAnvgoMpNBKoY1WzqGuVSwJt2el0ZaWcfArNaV4ArhUgzIrkRGu1ZwS9ztuZd+tZRkWWq2sqB2M1hQZdGDQXi+jg1ZKnBqm0nKTb6axxUjDaKGZUamGoI1ajlMx2Whm4iJCG60iOoIFuuqF9HFspdz08Gekc73h+/tCaw9OW+WwXkmSJEmSWjYTjTIrkTISd/iYLLWSDK2aIjOJ932H8jEH08+5meHOdfRy5P3lIiqWgMWRjnwVGEi+mUq2WsFpcb+yPNyWGf6DkEHp/suHB58laKuVMjeUy4pIG/4eP5sejoIm35cQUGYlxILQ9UYmW8202JDwNLWabo5CMtRqTFSKjUQ2mhmsN7KavC/S3qaQb6UBkKlWkaQGCAoHR7n/REGwONKRLwNDmrmPrZsMTiVJkiRJahbbqn9adNSKiVfDrI7ksDDcmXglyGDnBoa71tFJK9mtSplCwJJIe74KDGKrmUamWslpcb+yKpzDJ/7hyKD0QKGQZ6aTqPgBeL56PIe7VnBi3PzdWqZpe9WWm01GOpuMDDaa9m9fdPkbBwYd9FIOd63Ao0SotNzkmWksCndssfNFpT1TbCVRbCURp4T4NtivwXXWpd0ng1NJkiRJkpqZaq+nadrVcnP1AjRFMDfUjV9CvchQqxjmWscw53oyteodtiIELIu046vAQDabGaSr1Zwa9xtrw1l86h8ui5AcoKpFHAhBO62UX0I9WR5py9/jZ5PrKNrltiGhs9lIZ5ORzkYzg01GRqyyc03Rnz6OLaSqfgxUSoxENpqZrDOy9/bTkloIv2i9y7a0RDI43U2vv/I0q5YvwTAMCgu20q69PRf26ONPY9SYYxvdznff/AuAceNP2iv93FMfv/86XXJ7MHT4SB6+70buum9Kc3dph8rLSnj1pcnccudjcskeSZKk/YSBzlrDrr6ZQIB2eiEhnPw3MJCvA4PorBUzzLWOwc4NJETXsxQCVhpt+Mo/iA1mJqmql1M8v7HByOAz/1CZ0ZAAha1mOsmKjwgaz1UfzSjXco6PWxCrumoIlXwzhY1GRiwrWmgmx/5+UlUvnbQShjjXg6JQYbjZaqXze7ir/BuTpCbSKoPTDz98i2Aw0OTtut0ezjzzvJ0+5qLLrkdYguKiAh6+7wYemfzKHu2rpQWlNf529kWxy8uXLmy+jjRCaloGt9z5WHN3Q5IkSdpLvHhYabQDIF2pIlurotRK4GP/wXziH04fx1Z6O7byR7gL64xsUlQfJ3t+Z7ORxucBGZRK9VWKeBAW7bQyfgj1YVmkHT0d+Wwy0tlqbitIFK8E6aiV0suVh0s1qTTd5JupLI20JxJplYfPktQqtMr/rr0RmP7VdvPzNvPatCn4vNW4XG7Ou/gacrv1YtrUSTidTtatXUkg4OeU08/lsFHjmfHhGwCcfuaFzP75Oz6b8TaKAl1ze3HJlTeh69vemiWL5/P+W9NQFIX4+ESuvv5uEpOS+fnHb/j63x8jLEGXrj244NLrcDqdDbb3+Sdvx/YHcP1V53DXfU+xfOlCfv7xG6qrKhk8bAQVZaX07juQDetXA3DvHVcx5sgTWLZkAVdddxcAMz58A6fTxYmnnBPro7e6ildefIL8vM3ouoP/u2ACffsPYcEfc/jovekIIcjKbsPFl99AckpabP+ZWTksW7qQTz58k7vvf4qH7r2B3G69WLl8MVVVFVxwybUMHHwwJcUFvPz841RVVeB0urh0ws14PPE8fN8NPP3Ce3v8vkmSJEmtQ6lIotRIQsEenpmsBtlkpLMk0oFkxc+Jnj/IN5L5IjBkj9YmlQ4kKlvNdFIULyHhYF4ol/Z6GYe6VhGnRvCaDgpNew3h5dGTI5Ik7Rvy07uJvPjsIxx93Kk8+uSrnHvhVTz75H1EIvbCz4WFedz78FTuuPdJ3n1rGhXlZbHtykqLefvNF7jtnseZ9NTrWJbJwvlz67T9+Yy3uejyG3hw0kv0GziUDetXsWXzemZ++2/ufWgqj0x+haTkFP7zxQeNam97ZaXFPPzEy5z190tjt51/8bUA3P/oCxxy6BiWLJ5PIGAXFJjzy/ccdsRRddr4+IPXyc5px+NPv8GEa+/go/emU1lZzvRpU7jh1gd59MlX6d6zL2++9uwuX0vDiHDfI1M598Kr+Oi96QC8/sozDD/kCB6bMp3TzryAz2a8vct2JEmSpP2PiA7PXBZph1e46K7n0U0v4N+Bwfwe6SYDU6nRKkQCVcJNB62UUjORn0K9+TowgF/CvVlttq23Rq8kSXtfq8yctjTBQIDCgjyGH3wEAN169CEhIYn8vM0AHDHmGHRdJz09kx49+7JqxeLYtmtWLaNHz76kp9vrI02YeGe99ocMG8HTT/yDYcNHMmT4SPoPHMY3X31KYf4W7rvzagAMw6Bz1+47bG/jhjU77H/nLt3RtB1XkHN7PAwcfBC/zfuZrOw2ZGW3JTWt7tpJK5YtimVWO3Tqyn2PTGXBH3Po2q0XmVn23KGxR57AF5/uOss5YNBBALTv0AWftzrW/tXX3w3AoCGHMGjIIRQXNX05eEmSJKn1sNBYbbRt7m5IrZhAZa2Z09zdkCQpSganTcAS9ddhE0JgmvYE+9qBnxACtdZ1TddRatXHr6qsACApOSV227EnnMHgoYeycP4c3n/7ZdatWYHL7eHgQ0fHMpzBQADTMlmxbFGD7SmKgrBE7HbT2FZC3encdZWxUWOP5fMZb5OV1YYjRh9d735N01FqVUHM27oJy6r7ugjAtOzXREFBCFGvLwAOp3Nbn7Efo9Ua5iyEIG/LRpwu9y77LUmSJEmSJElS6yDHvjSBuLh4srLb8Nu8nwA7G1pZUUaHDl0AmDf7B4QQlBQXsGb1cnr27h/btmtuT9asWhYb6vv2my/wx++z6rR/7x1XEQz6Oeb4v3HMCX9jw/rV9O47iN9//YXKynKEELz+ylN8/e+Pd9heYmIyWzZvAGDt6uVUVJSxK6qqxgLsXr0HUFZazLKlCxl60Mh6j+3ZewCzZ30P2IHp4w/fRrfuvVmzenkswznz2y/p03cQAAlJyWzdYvdn/m+z6rW3vV69BzB31kwAlvz5B69Na7lVhCVJkiRJkiRJ2n0yc9pEJky8k+kvP8WMD97E4XBw3S33ozscAIRCIe657UoMI8IlV9xIYmJybLvUtAzOu+gaHn/4VizLoluPPowaXXc5lDP+fikvPz8JVdVwuz1cOuFmctq059QzzufR+2/CsgSdOudy4il/x+l0Ntiez+/lt7k/cev1F9Gla3c6de62y+c0dPhI7rz5Uh6cNA2n08mwgw/HW12Fw1F/DsbpZ13Iay9N5s6bL0VVNa689g6SU9K45IobefqJf2AYETIys7l0wi3248+8gH9Of45PPvonAwYO22VfLrhkIq++NJlv//s5LpeLS6+8eZfbSJIkSZIkSZLUeiiiZmxlC1Fa6sWy6napoGAjOTmdYtebaykZbyBCaVWwzvDYXdkf1t8UQmAYER578BbOvfBqunTt0dxdOiB4q/KZ8unW5u6GJEmSJEmS1Ep4XBqXn9iXgd0ydv3gZqCqCunpCTu8v1VmTne1FqnUtCoqyrjthosYM+54GZhKkiRJkiRJkrRXtMrgtDW54prbmrsLf1lqajovv/Gv5u6GJEmSJEmSJEn7MVkQSZIkSZIkSZIkSWp2MjiVJEmSJEmSJEmSmp0MTiVJkiRJkiRJkqRmJ4NTSZIkSZIkSZIkqdnJ4FSSJEmSJEmSJElqdq2yWu+WxU9iGb4mb1fV42nf/6adPub1V55m1fIlGIZBYcFW2rW31189+vjTGDXm2Ebv67tv7Oq348aftOcdbkLnnjGWtz/6vrm7IUmSJEmSJEnSAapVBqd7IzBtbLsXXXY9whIUFxXw8H038MjkV/ZoXy0lKJUkSZIkSZIkSWoJWmVw2hLl523mtWlT8HmrcbncnHfxNeR268W0qZNwOp2sW7uSQMDPKaefy2GjxjPjwzcAOP3MC5n983d8NuNtFAW65vbikitvQte3vTUzPnyDkuJC8rZuorqqkrFHncAJJ5+NZZq8+9Y0li9biLAsDh99NMeecAYAn3/yDrN++hZVVek/cBjnnHs5paXFTJl0N23adWDr5g1kZGYz4do7SUhMiu0rGAjwxmvPsGXTeizL4oRTzubQw8bVea4/zfyaBX/MpaqqnIryMgYPHcH/XTABRVH416fvMm/ODwjLov/A4Zx97uUoisKH777G0iXz8XmrSUlN59ob7iE5Ja3B525EIrvsgyRJkiRJkiRJ+xcZnDaRF599hBNPPYfhBx/BmlXLePbJ+5j87D8BKCzM496Hp1JZWc49t11JvwHDYtuVlRbz9psv8OCkl0hPz+TFZx9h4fy5DDvosDrtb1i3mnsfeg7LMrn7tivp238I69asAODhx18mEgkz6aHb6JLbk1AwwPzfZ/PgpBfRdQfPTL6X7/73BYOGHMLmTes47+Jr6NN3EO+8+SKffPQm5198bWw/n814iy5de3DlNbfj9/t44O5r6da9N1nZbev0Z9WKxTwy+RXi4xN5+L4b+P3Xn3E63WxYt4oHH30RFIWXnnuUWT9/S7fufcjP28S9Dz2Hqqqx2w85dEyDz33NqmWN6oMkSZIkSZIkSfsPGZw2gWAgQGFBHsMPPgKAbj36kJCQRH7eZgCOGHMMuq6Tnp5Jj559WbVicWzbNauW0aNnX9LTMwGYMPHOBvcx4rCxuD0eAIYMO5RlSxawZtVyNm5Yw7IlC+x+BANs3rSe4sI8Rowci8vlBmDU2GP5+Yf/MmjIIeS0aU+fvoMAOHz0eF545uE6+1m6eD6hUIgfv/8KgFAwyJbNG+oFhkOGjyQ5JQ2AQ0aOZdniBTicTtauXsHdt10JQDgcIj0ji8OOOIq/nz+BH777D/l5m1m9ahlZOW13+Nw/n/F2o/ogSZIkSZIkSdL+QwanTcASVr3bhBCYpgmApml1bldrXdd0HUVRYterKisASEpOqdOeqm4rrCyEhaZpWJbJOeddHguKq6sqcbk9fPjuq43riyVQVa3OYy3LYsLEO+jStQcAlRVlxCcksb26z8lC1TQsy+Lo40/nuBPtocU+nxdN1Vi/dhXPP/MQx57wNw465AhUVUUIscPn3tg+SJIkSZIkSZK0/5BLyTSBuLh4srLb8Nu8nwA7G1pZUUaHDl0AmDf7B4QQlBQXsGb1cnr27h/btmtuT9asWkZFeRkAb7/5An/8PqvePv749RcikTA+bzUL/phD/wHD6Nt/MDO//Q+GYRAMBHjgnomsWb2Mvv0HM2fW94RDIUzT5KeZX9On3yDAnhu7cf0aAH6c+TUDBx9UZz99+g2OVRIuLy/ljpsvo7SksF5//lz4K36fl3A4zJxfvmfg4IPp028ws376H8FAANM0eerxe/h17o8sX7aI3n0GMm78SeS0bc+CP+ZiWdYOn3tj+yBJkiRJkiRJ0v5DZk6byISJdzL95aeY8cGbOBwOrrvlfnSHA4BQKMQ9t12JYUS45IobSUxMjm2XmpbBeRddw+MP34plWXTr0YdRo4+p177D6eLBe64jEPBz4ql/p12HzmS3aU9B/lbuvvVyTNPkiDHHxIbsbtywhntuvxLTNOk/cBjjjz2NstJiEhKSmPHhGxQWbKVDp65ceuXNdfZz6hnn88arz3D7jRdjWRZnn3s52Tnt6vUnMSmFJx69g+qqSkYecRQDBg0HYNPGtdx751VYlsWAQcM5fPTRlJeV8Mzke7n9xksA6Jrbg+Kigh0+92Ao2Kg+SJIkSZIkSZK0/1CEEKK5O1FbaakXy6rbpYKCjeTkdIpdb651Tr2BCKVVQYTV+Jds2tRJ9O47kCPG1A84G6t2Zd+/omb5m6dfeO8vtfPTzK9ZvnQRV1xz219qR9o5b1U+Uz7d2tzdkCRJkiRJkloJj0vj8hP7MrBbRnN3pUGqqpCenrDD+1tl5nRnAaQkSZIkSZIkSZLU+rTKzGlz2ZPMqSTtKZk5lSRJkiRJknZHa8+cyoJIkiRJkiRJkiRJUrOTwakkSZIkSZIkSZLU7GRwKkmSJEmSJEmSJDU7GZxKkiRJkiRJkiRJzU4Gp5IkSZIkSZIkSVKza5VLyTyyYB1ew2zydhN0jTsHd93pY15/5WlWLV+CYRgUFmylXXu7ivDRx5/GqDHHNnpf333zLwDGjT9pt/t5/VXncNd9T5GZlbPb2zakKdZibUpFhfl8PuNtLrvqlmZpu7yshFdfmswtdz622+3vaA3YplpjVpIkSZIkSZL2V60yON0bgWlj273osusRlogFG49MfmWP9rUnQemBoqSkkMLCvGZrOzUtY48CU0mSJEmSJEmS9lyrDE5bovy8zbw2bQo+bzUul5vzLr6G3G69mDZ1Ek6nk3VrVxII+Dnl9HM5bNR4Znz4BgCnn3khs3/+js9mvI2iQNfcXlxy5U3o+ra3xltdxYvPPUJpSTHt2nciEg4DYJkm7741jeXLFiIsi8NHH82xJ5zB00/8g0MPP5KDDjkCgLtvvYJLJ9yM2x3HG688TbW3EpfTzfmXXEvnLt3rPI8fZ37FV198BCh06dqDCy6ZiNvj4apLTmPYwYezeuVS+/rEu8jMyuH6q85hxMixLPnzD1RN45TTz+OrLz6koGArfz9/AoccOprKijKmv/wUpSVFKKrKWX+/lH4DhjLjwzcoLyuhIH8rpcWFjB53HCeffi5vTZ9KUWEeb7z6DBdeel2sb6Wlxbz4zMOEQkEUReH8i6+lW48+dTLJy5Yu5JMP3+Tu+5/iP198xC8//hdFUenarReXXHFjnbbPu+gaXn/lKbZs2kBlZTkdO3Xl6uvuprKyPJblnDZ1Ep64eDasW0V5WQmnnHEeo8YcS1lpMa+8OBm/30tFWSmHjz6av519EQCFBVt58B/X4fNWM2jIIZz1f5fVeY139HpIkiRJkiRJ0oFMzjltIi8++whHH3cqjz75KudeeBXPPnkfkYgdRBYW5nHvw1O5494nefetaVSUl8W2Kyst5u03X+C2ex5n0lOvY1kmC+fPrdP2jA9ep3OX7jw25TWOOuZkKivLAZj53b8BePjxl7n/0Rf447fZrFj+JyOPOIo5v3wPQEH+FiKRMJ27dGfa849x9nmX8/DjL3PJFTcy9akH6+xn88Z1fD7jHe66/ykem/IaLrebTz56E4Cqqgq69ejDo0++yoiRY/jn9Odi2yWnpPHgpJdo264jX3z2Hrfd8wQTrr2TLz59F4C3Xn+eI8Ycw0OPT+PG2x5i+stTCAT8sX3efvfj3Pfo83zx2Xv4fF7Ou/gauuT2rBOYAvz4/X8YNPQQHpz0EqefdRErVyze4fthmSZffPouDzz2Eg9OegnDiFBWWlyn7dUrl6DrDu57ZCpPPvcWPp+XhQvm1WurrLSIex58hhtve4j3/jkNgDmzvmfEYWO5/5HnefTJV/n6PzOorqoEoLgon+tuup+HJk1j1YolzP99dp32dvZ6SJIkSZIkSdKBSmZOm0AwEKCwII/hB9uZym49+pCQkER+3mYAjhhzDLquk56eSY+efVlVK6has2oZPXr2JT09E4AJE++s1/7yZYu4+rq7AejVZyBZ2W0AWPLnfDZuWMOyJQvsfgQDbN60njFjj+P/27v34CirPI3j3066A0IETEgMl3AREDCQBIxyGRNuYpQkqEQFxQFEC4ZRmAV2IQhCjYAUKLcZHVZqB5HVmgIKFgQkMiCgJIwCoigCAhqjIXcRyD3dffaPwKuNZAza0g08nyr+eN8+73nPeXlC8etzOr3q73+hvLyMvXve5XcJA6koL+fLE8dY/soCq9/KinLOnTvjcZ/ucb248cbGAPS7O5nlf5sPgMMRRHyfewC4q08iq9/8H+u6mG53AtA07GZCQpoSGBhI07CbKS09VzPOTw9wKiebdatXAuByuijIq9la2zkqFrvDQePGN9Ew+EbKy0pqfc5dut7Okpdm8fVXJ4i9vScD732w1rYBgYF06BjFzLRx3H7H7xiU8gghoWHk5eVYbTrdFkPwjY34Z/oGTuVkk5/7LRUV5T/pq2tMHDabjZat2lJSchaApMFD+fyzg2x5azXfZmfhclZTWVlzbfe43jRq3ASAHr37cuTwx7Rq3c7qr7bn0bpt+1rnIyIiIiJyrVNx6gVu4/7JOWMMLlfNZ1gDAwM9zgf86DjQbsdms1nHZ898D2AVNzVsGIx1FBBQc73b7eLR34+xiuJzZ89Qr/4N2B0OusX15qP9mXywdxf/Oe0F3G4XjqAgj8/IFhcXEhzcyGNsF00Ct8t9/p4B1jiNMR5zstsdP4ztR+et5+N28+yshQTfWHOv06eLadyoCfv37cERFPTDLG02Lh7Cj93aqQvzF6/g4IF/8a+Mnby/8x3SZr6IDZs1dpfTabWfOGU2J774nE8OfsiCuVP544TpHv0d2JfButUrSUwaQkK/e2sK9UsMwOEIssZ3wZuvL6Mg/xS94wdw+x138dmnB6xLAzz+vt0EBnr+mNX2PERERERErmfa1usFDRo0JPzmZuz74D2gZjX0zPffERnZFoAPMndhjKGoMI8Tx4/QsXNX69pb2nXkxBefW1t933j9bxzYn+HRf5fo7uzZ/U8ATp44Sv75Vceort3Yuf1tnE4nFeXlPP/cBE4c/xyAuxIG8vamNQQHN6JpWAQNGgYTEdGCPe/V9PPpJ/uZM/M/PO7TOSqGj/ZnUnKuZnVw544tdI6KBaCyssLanvrezq1En18trYuoLt3Yvm0jADnfZDFt0pNUVlXW2j4wIBC366e/nOof//sqGe9tJ6FvIiOfnEDWV8cBCG7UmJxvswD4aF/Nszt75numTnyCyFa38NCwJ+gaE8c32Sc9+j786Uf06N2XPv3uo2HDYI589jFu90/faLiUzw4dIOn+ofTo1ZfcU99w+rsi3O6afj85+AGlpSVUVVWxN2MnXaK7/6rnISIiIiJyPbgqV06D7YG/2VfJ/FLjJjzLiuWLWbf6dRwOB3/6rz9jd9SsKFZWVvLc1D/gdFbz5NhJ1rZZqPnNsL9/4hkWzJ2C2+2m/a230aev51e6DHlkFMtfmc/UiU/QvEUra1tv/4GDycvNYcaUMbhcLhL63ctt54vJWzt1obyslLt/9FuB//in6axYvpgtG1djt9t5ZuJzHquBrVq3I+XBx5gzayIul5O2t9zKE2MmWq9/uHc3a/+xgiY3hfKHi74q5d8ZMXo8f391EdMmP4UxhnETpnHDDQ1qbd+8ZSvKykpY9pcXPLY533Pfg7yydC7v7UwnICCAsc+kAZD6yEhWrfgr69euIjomDqhZee53dxIzp40jKKgezZpHktDvPqqrq6y+k+8fxitL57J3z7vY7XZu7dSFwoLcOs0p5cFH+e+/ziMoqB4hoeG0bdeRwoI8AJo1b8VLL6RRWlpK7/j+dI25w3rtlzwPEREREZHrgc38ZC+nbxUXl+B2ew4pL+9rIiJa+2hEPygpr6b4bAXGXfdH5m/fIfpLPf5wf95Y+66vh3FdKTmby6L/y/n5hiIiIiIiwA31AhmTEkVM+6a+HsolBQTYCA0Nrv31KzgWERERERERkUu6Krf1Xk3GXsb2V3+mVVMREREREfktaeVUREREREREfO6qKU797KOxIr8tY/7t1+qIiIiIiFxrrori1G4PorT0rApUuT4Yg9tVTsH3Vb4eiYiIiIjIFXNVfOb0ppvCOH26kJKS7306jsoqFyUVThXJ8psyBgq+r2LD3mJfD0VERERE5Ir5TYrTTZs2sWzZMpxOJyNHjmT48OG/qr/AQDtNmzbz0uh+uQ8+z2fVO8cor3T6eigiIiIiIiLXFK8Xp/n5+SxevJj169cTFBTEsGHD6NGjB+3bt/f2rUREREREROQa4fXPnGZmZtKzZ0+aNGlCgwYNSExMJD093du3ERERERERkWuI11dOCwoKCAsLs47Dw8M5dOhQna8PCLB5e0hec0M9O2FNbqCiStt6RURERETEv9QPCqR+Pbvf1lQ/Ny6vF6dutxub7YebGmM8jn/OTTc19PaQvGZAaDADerbx9TBERERERESuOV7f1hsREUFhYaF1XFhYSHh4uLdvIyIiIiIiItcQrxenvXv3Zu/evXz33XeUl5ezbds2EhISvH0bERERERERuYZ4fVvvzTffzMSJExkxYgTV1dU89NBDREdHe/s2IiIiIiIicg2xGWOMrwchIiIiIiIi1zevb+sVERERERERuVwqTkVERERERMTnVJyKiIiIiIiIz6k4FREREREREZ9TcSoiIiIiIiI+p+L0vJKSEpKTk/n2228BWL9+PYMGDSIlJYU5c+bgdDo5cuQI999/v/UnPj6e5ORkAE6dOsXw4cO59957GTduHKWlpb6cjlzn6pJngMOHD5OamsrgwYMZO3YsZ8+eBZRn8S91zfPu3btJSUkhJSWFyZMnW7lVnsWfvPzyyyQlJZGUlMSCBQsAyMzMJCUlhXvuuYfFixdbbY8cOcKQIUNITExk+vTpVtaVafEXl5PnC6ZMmcL69eutY+VZPBgxH3/8sUlOTjZRUVHmm2++MSdPnjTx8fEmPz/fGGPMrFmzzIoVKzyuKSsrM0lJSWbfvn3GGGPGjBljNm/ebIwx5uWXXzYLFiy4spMQOe9y8vzoo4+aXbt2GWOMmTdvnlm0aJExRnkW/1HXPJ85c8b07NnTHD9+3BhjzPLly83s2bONMcqz+I+MjAwzdOhQU1lZaaqqqsyIESPMpk2bTJ8+fUx2draprq42o0ePtv5dTkpKMgcPHjTGGDNt2jTz5ptvGmOUafEPl5vnvLw8M3bsWBMdHW3WrVtn9aM8y49p5RRYs2YNs2bNIjw8HIBjx44RGxtrHffr14/t27d7XPPqq69yxx13EBcXR3V1Nfv27SMxMRGAIUOGkJ6efmUnIXLe5eTZ7XZb71CWl5dTv3595Vn8Sl3znJWVRfPmzWnfvr3HeeVZ/ElYWBhpaWkEBQXhcDho164dWVlZtG7dmsjISOx2OykpKaSnp5OTk0NFRQWxsbHAD9lVpsVfXE6eATZt2sSAAQO47777rD6UZ7mYilNg7ty5xMXFWcedOnXik08+ITc3F5fLRXp6OkVFRdbr586dY82aNTzzzDMAnD59muDgYOx2O1Dzw5qfn39lJyFy3uXkOS0tjRkzZnDXXXeRmZnJsGHDlGfxK3XNc5s2bcjLy+Po0aMAbN26laKiIuVZ/EqHDh2sYjMrK4utW7dis9kICwuz2oSHh5Ofn09BQYHH+QvZVabFX1xOngGeeuopHn74YY8+lGe5mIrTS2jbti2TJ09m3LhxDB8+nI4dO+JwOKzX33rrLe6++25CQ0MBMMZgs9k8+rj4WMRXastzRUUF06dPZ+XKlezZs4fHHnuMqVOnKs/i12rLc6NGjZg/fz7PPfccqamphIeH43A4lGfxS8ePH2f06NFMmTKFyMhIj0xeyKzb7b7keWVa/E1d8lwb5VkuZvf1APxRZWUl0dHRbNiwAah5Bz4yMtJ6ffv27YwdO9Y6DgkJ4dy5c7hcLgIDAyksLLS2nIn4Wm15/uKLL6hXrx7R0dEADB06lKVLlyrP4tdqy7PL5SIiIoK1a9cCcOjQISIjI5Vn8TsHDhxgwoQJPPvssyQlJfHhhx9SWFhovX4hoxERER7ni4qKCA8PV6bFr9Q1z7VRnuViWjm9hLKyMkaNGkVJSQlVVVW88cYbDBo0CKh5h+fw4cN069bNau9wOIiLi+Ptt98GYMOGDSQkJPhk7CIXqy3PrVu3Ji8vjy+//BKAHTt20LVrV+VZ/FptebbZbIwePZr8/HyMMaxcuZJBgwYpz+JXcnNzefrpp3nppZdISkoCICYmhq+++oqvv/4al8vF5s2bSUhIoEWLFtSrV48DBw4AsHHjRhISEpRp8RuXk+faKM9yMZsxxvh6EP6if//+rFq1ipYtW7J27VpWrlyJ0+kkOTmZ8ePHA1BcXMzgwYPJyMjwuDYnJ4e0tDSKi4tp1qwZixYtonHjxr6YhghQtzzv3r2bhQsXYowhNDSU2bNnExkZqTyL36lLnnft2sXChQupqqqiV69eTJ8+HYfDoTyL35gzZw7r1q2jVatW1rlhw4bRpk0b5s2bR2VlJX369GHatGnYbDaOHj3KjBkzKCkpISoqinnz5hEUFKRMi1+43DxfkJaWxp133smQIUMA/R9aPKk4FREREREREZ/Ttl4RERERERHxORWnIiIiIiIi4nMqTkVERERERMTnVJyKiIiIiIiIz6k4FREREREREZ9TcSoiIiIiIiI+p+JUREREREREfM7u6wGIiIhcS2bMmEFoaCgTJ04EYOPGjWzbto3U1FSWLVtGdXU19evXZ+rUqXTr1o2ioiJmzpxJcXExhYWFtGjRgiVLlhAaGkr//v2Jjo7m2LFjTJo0iYEDB/p4diIiIr8drZyKiIh40fDhw1m3bh1OpxOANWvWEB8fz+LFi1m+fDkbNmxg9uzZjB8/nrKyMrZs2UJsbCyrV69mx44d1K9fn40bN1r9dejQga1bt6owFRGRa55WTkVERLyoc+fOtGzZkl27dtG2bVsKCgpwuVwUFBQwatQoq53NZiM7O5uRI0eyf/9+XnvtNbKysjh+/DgxMTFWu7i4OB/MQkRE5MpTcSoiIuJlF1ZP27RpwyOPPILb7aZXr14sWbLEapObm0t4eDgvvvgihw4dIjU1lR49euB0OjHGWO0aNGjggxmIiIhcedrWKyIi4mWJiYkcOXKEd955h9TUVHr16kVGRgYnT54EYPfu3QwePJiKigr27NnDyJEjeeCBBwgNDSUzMxOXy+XjGYiIiFx5WjkVERHxsqCgIBITEykqKiIkJISQkBCef/55Jk2ahDEGu93OsmXLaNiwIU8//TQLFixg6dKlOBwOunfvTnZ2tq+nICIicsXZzI/3DomIiMivVlZWxuOPP87MmTOJjY319XBERESuCtrWKyIi4kXvv/8+ffv2JT4+XoWpiIjIZdDKqYiIiIiIiPicVk5FRERERETE51ScioiIiIiIiM+pOBURERERERGfU3EqIiIiIiIiPqfiVERERERERHxOxamIiIiIiIj43P8DpKpYmPsqV/AAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 1152x648 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_year.plot.area(figsize=(16,9)); "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fdb88e86",
   "metadata": {},
   "source": [
    "* Probably the largest growth group has been \"sustainable development\" \n",
    "* The largest reduction in relative terms seems to be close between South Africa and peace \n",
    "\n",
    "* Likely need to do some more filtering, to get some more meaningful topics but a good overview "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ee4a3234",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "09d56e6e",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "51a11a89",
   "metadata": {},
   "source": [
    "# Nonnegative Matrix Factorisation (NMF)\n",
    "* Re-try with bi-grams"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "a9c780bc",
   "metadata": {},
   "outputs": [],
   "source": [
    "## start with just unigrams \n",
    "tfidf_para_vectorizer_bigrams = TfidfVectorizer(stop_words=stopwords, min_df=5, max_df=0.7, smooth_idf = True, \n",
    "                                       ngram_range=(1,2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "d8d6056a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 30.8 s\n"
     ]
    }
   ],
   "source": [
    "%%time \n",
    "\n",
    "# tfidf_text_vectors = tfidf_text_vectorizer_bigrams.fit_transform(df['text'])\n",
    "\n",
    "tfidf_para_vectors_bigrams = tfidf_para_vectorizer_bigrams.fit_transform(paragraph_df[\"text\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "4fbe0114",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(282210, 337296)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# substantially bigger with bi-grams and uni-grams now \n",
    "tfidf_para_vectors_bigrams.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d8eafcf7",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "1f2447c6",
   "metadata": {},
   "outputs": [],
   "source": [
    "# reinitialise base model\n",
    "nmf_para_model = NMF(n_components=10, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "d817d7ac",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\jason\\anaconda3\\lib\\site-packages\\sklearn\\decomposition\\_nmf.py:312: FutureWarning: The 'init' value, when 'init=None' and n_components is less than n_samples and n_features, will be changed from 'nndsvd' to 'nndsvda' in 1.1 (renaming of 0.26).\n",
      "  warnings.warn((\"The 'init' value, when 'init=None' and \"\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 2min\n"
     ]
    }
   ],
   "source": [
    "%%time \n",
    "\n",
    "# define matrixes from para model + the afore-defined para vectors variable \n",
    "W_para_matrix = nmf_para_model.fit_transform(tfidf_para_vectors_bigrams)\n",
    "H_para_matrix = nmf_para_model.components_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9f8f4833",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "ce23c4ad",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Topic 00\n",
      "  nations (2.30)\n",
      "  united nations (2.27)\n",
      "  united (2.27)\n",
      "  organization (0.45)\n",
      "  charter (0.36)\n",
      "  role (0.31)\n",
      "  member (0.31)\n",
      "  states (0.30)\n",
      "\n",
      "Topic 01\n",
      "  general (1.22)\n",
      "  assembly (1.18)\n",
      "  session (1.16)\n",
      "  general assembly (1.05)\n",
      "  mr (0.81)\n",
      "  president (0.71)\n",
      "  election (0.62)\n",
      "  session general (0.57)\n",
      "\n",
      "Topic 02\n",
      "  international (0.51)\n",
      "  world (0.39)\n",
      "  peace (0.29)\n",
      "  community (0.27)\n",
      "  international community (0.23)\n",
      "  new (0.21)\n",
      "  us (0.20)\n",
      "  states (0.18)\n",
      "\n",
      "Topic 03\n",
      "  east (0.85)\n",
      "  middle (0.81)\n",
      "  middle east (0.81)\n",
      "  palestinian (0.75)\n",
      "  israel (0.64)\n",
      "  people (0.58)\n",
      "  peace (0.54)\n",
      "  arab (0.52)\n",
      "\n",
      "Topic 04\n",
      "  rights (2.76)\n",
      "  human (2.62)\n",
      "  human rights (2.50)\n",
      "  respect (0.42)\n",
      "  respect human (0.36)\n",
      "  universal (0.35)\n",
      "  fundamental (0.33)\n",
      "  protection (0.32)\n",
      "\n",
      "Topic 05\n",
      "  nuclear (2.20)\n",
      "  weapons (1.45)\n",
      "  nuclear weapons (0.90)\n",
      "  disarmament (0.77)\n",
      "  treaty (0.77)\n",
      "  proliferation (0.67)\n",
      "  non (0.54)\n",
      "  non proliferation (0.52)\n",
      "\n",
      "Topic 06\n",
      "  africa (1.59)\n",
      "  south (1.48)\n",
      "  south africa (1.16)\n",
      "  african (0.64)\n",
      "  namibia (0.60)\n",
      "  apartheid (0.56)\n",
      "  regime (0.40)\n",
      "  people (0.39)\n",
      "\n",
      "Topic 07\n",
      "  council (2.26)\n",
      "  security (2.18)\n",
      "  security council (1.87)\n",
      "  permanent (0.57)\n",
      "  peace security (0.55)\n",
      "  reform (0.49)\n",
      "  resolution (0.43)\n",
      "  peace (0.42)\n",
      "\n",
      "Topic 08\n",
      "  countries (1.78)\n",
      "  developing (1.12)\n",
      "  developing countries (1.02)\n",
      "  economic (0.72)\n",
      "  developed (0.60)\n",
      "  developed countries (0.45)\n",
      "  trade (0.41)\n",
      "  international economic (0.28)\n",
      "\n",
      "Topic 09\n",
      "  development (1.55)\n",
      "  sustainable (0.48)\n",
      "  sustainable development (0.41)\n",
      "  goals (0.37)\n",
      "  social (0.34)\n",
      "  economic (0.34)\n",
      "  global (0.30)\n",
      "  development goals (0.30)\n"
     ]
    }
   ],
   "source": [
    "display_topics(nmf_para_model, tfidf_para_vectorizer_bigrams.get_feature_names(),\n",
    "              top_words=8)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5fc05f9e",
   "metadata": {},
   "source": [
    "* The bigrams concept looks good although we need to revisit our stop-words now to be a little more aggressive \n",
    "    * too many topics are clogged up with overlaps, E.G. topic 8 = countries, developing, developing countries, developed countries etc. \n",
    "* But it has shown up some additional info, E.G. topic 5 with nuclear disarment, we now see terms like \"non proliferation\" which gives a clearer idea how nuclear disarment is being approached "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7287a97a",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "e7ac5740",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Topic 0 = 11.01%\n",
      "Topic 1 = 10.09%\n",
      "Topic 2 = 18.94%\n",
      "Topic 3 = 8.25%\n",
      "Topic 4 = 6.59%\n",
      "Topic 5 = 5.69%\n",
      "Topic 6 = 8.78%\n",
      "Topic 7 = 8.32%\n",
      "Topic 8 = 9.86%\n",
      "Topic 9 = 12.46%\n"
     ]
    }
   ],
   "source": [
    "## see individual contributions again: \n",
    "for i, val in enumerate(W_para_matrix.sum(axis=0) / W_para_matrix.sum() * 100.0 ):\n",
    "    print(\"Topic {} = {:.2f}%\".format(i, val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e2a3dd9a",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3d24cd09",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "2766ec25",
   "metadata": {},
   "source": [
    "## FUTURE WORKS\n",
    "* further tune the bi-gram NMF model with lemmatisation/stopword removal\n",
    "* add tuned Latent Dirichlet Allocation \n",
    "* add singular value decomposition (SVD) comparison "
   ]
  }
 ],
 "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": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}