{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# Abstract\n", "\n", "Thirty-three months of daily reading rate data collected by a second-language learner of Mandarin Chinese is analyzed to evaluate growth of mean reading rate as a function of total time spent reading. Two models for this function are fit to the data: (1) a linear model and (2) an exponential decay model where mean reading asymptotically approaches that of a native speaker. Confidence regions for the parameters of these models are plotted, and these confidence regions are then used to (1) evaluate how much growth in mean reading rate has already been achieved and (2) estimate how much further reading will be necessary to reach a reading rate that is comparable to a native speaker. Despite high variance in measured reading rates, it is concluded that after 270 hours of reading the mean reading rate has increased between 52 to 122 percent and anywhere between 700 and 3000 hours of further reading will be necessary to achieve a reading rate comparable to that of a native speaker." ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": [ "# Import Python libraries needed for analysis.\n", "import csv\n", "import matplotlib.pyplot as plt\n", "import numpy as np\n", "import scipy.optimize\n", "import scipy.stats" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# 1. Introduction and Project Objectives\n", "\n", "I study Mandarin Chinese as a hobby. My Chinese reading rate is still slow compared to that of a native speaker, so I've been spending about 20 minutes a day reading in Chinese for a couple years. Every time I've sat down to read I've measured (1) how long I spent reading and (2) how many Chinese characters I've read, thus allowing me to track my reading rate as a function of total time spent reading. The focus of this project will be to analyze this data in an attempt to answer the following questions:\n", "\n", "1. How much can I confidently say my reading rate has improved after this practice?\n", "2. What model of reading rate as a function of total time spent reading appears to fit my reading data best? A linear model may fit the data well, but this would imply that my reading rate increase has remained constant over time. A possibly more reasonable alternative would be an exponential decay model where my reading rate asymptotically approaches that of a native speaker. This model would better capture diminishing returns: Increases in reading rate can only be accomplished via exposure to rarer and rarer vocabulary, and the only way to achieve this exposure is via greater and greater amounts of time spent reading. \n", "3. Assuming that the previously determined best-fit model is correct, when can I expect to reach a reading rate comparable to that of a native speaker?\n", "\n", "This project report is presented in the format of a [Jupyter Notebook](https://jupyter.org) with blocks of Python code interspersed throughout, but programming experience is not necessary to understand the conclusions I have made.\n", "\n", "# 2. Literature Review\n", "\n", "## 2.1 Second-Language Reading Rate Increases\n", "No studies were found that assessed reading rates increases of second-language learners continuously over time. (Beglar and Hunt 2014) summarized several studies that examined reading rate increases of English as a Second-Language (ESL) students who undertook intensive reading programs. Of the ten studies described, however, all of them only assessed student reading rates at the beginning and end of the reading program.\n", "\n", "## 2.2 Chinese Native Speaker Reading Rates\n", "(Trauzettel-Klosinski and Dietz 2012) found that Chinese native speakers read aloud at an average rate of 255 Chinese characters per minute (CPM) with a standard deviation of 29 CPM. This rate is presumably significantly slower than if the test participants read silently, but it is a good estimate of the reading rate a non-native speaker could achieve with extensive practice. \n", "\n", "## 2.3 Zipf's Law\n", "\n", "Zipf's law is an empirical observation that the probability of encountering any given word is inversely proportional to its rank in the word frequency for table in that language (Murphy 2012). If the log of word rank is plotted versus the log of word frequency, a straight line relationship should be observed if Zipf's law holds for the language. The plot below ([Sergio Jimenez/Wikimedia, CC BY-SA 4.0](https://commons.wikimedia.org/wiki/File:Zipf_30wiki_en_labels.png)) shows such a relationship holds for many languages when Wikipedia is used as a corpus.\n", "\n", "![](zipf_law.png)\n", "\n", "# 3. Methods\n", "\n", "## 3.1 Data Collection\n", "\n", "I collected the data used in this project while reading material from three Mandarin language sources:\n", "\n", "1. The [*Remembrance of Earth's Past*](https://en.wikipedia.org/wiki/Remembrance_of_Earth%27s_Past)《地球往事》trilogy by Liu Cixin (often referred to as the *Three Body Problem* trilogy, after the name of the first book). Started March 2016 and finished October 2017.\n", "2. [*Wolf Totem*](https://en.wikipedia.org/wiki/Wolf_Totem)《狼图腾》by Jiang Rong. Started November 2017 and finished September 2018.\n", "3. [\"InTouch Today\"](https://view.news.qq.com/)（今日话题）, a daily news essay published by [Tencent](https://en.wikipedia.org/wiki/Tencent) that discusses various topics relevant to current events and society in China. Started reading in September 2018, continuing until the present (December 2018).\n", "\n", "I read only on my Android phone, and the time I spent reading every day was measured by an app ([QualityTime](http://www.qualitytimeapp.com/)) that tracks and reports phone usage data. I attempted to read as quickly as possible while comprehending the material enough so that I could hypothetically produce an English translation. When I encountered a sentence I could not understand, I would continue reading to the end of the sentence, attempting the infer the meaning of any unknown words. If I was still unable to infer the meaning of any unknown words, I would look them up in a Chinese-English dictionary app ([Pleco](https://www.pleco.com/)). Time spent looking up words in the dictionary app was not counted towards the daily time spent reading and thus did not affect the reading rate measurements. In the rare instances I was still unable to understand a sentence after exhausting the information in my dictionary app, I would continue on to the next sentence.\n", "\n", "## 3.2 Linear Model\n", "\n", "The reading rate for any particular reading session (number of characters read divided by time spent reading, [CPM]) is dependent on internal factors (e.g., the size of your vocabulary and your reading comprehension ability) as well as external factors (e.g., fatigue and environmental distractions). The reading rate will vary from session to session due to variance in these internal and external factors, but a mean reading rate can be calculated and tracked over time.\n", "\n", "The first of two proposed models for mean reading rate increase as a function of time spent reading is a simple linear model.\n", "\n", "$$R(t) = R_0 + m t$$\n", "\n", "Where:\n", "\n", "* $R$ is the mean reading rate [CPM].\n", "* $t$ is the total time spent reading [min].\n", "* $R_0$ is the initial mean reading rate [CPM].\n", "* $m$ is a linear coefficient [CPM/min].\n", "\n", "The primary assumption inherent with the linear model is that on all time spent reading is equally effective towards increasing the mean reading rate. That is, if spending an hour reading increases your mean reading rate by 0.1 CPM at the start of a study program, then an hour spent reading after a year of practicing every day for an hour will still increase your mean reading rate by 0.1 CPM.\n", "\n", "## 3.3 Exponential Decay Model\n", "\n", "The linear model may work well for second-language learners at the start of an extensive reading practice program, but it may not work well for learners who have already spent hundreds of hours reading. Advanced learners experience diminishing returns, in which greater amounts of reading time are required for inducing an equivalent increase in the mean reading rate. For such learners an exponential model may be more appropriate for describing increases in the mean reading rate brought about by time spent reading.\n", "\n", "$$R(t) = R_0 + (R_{max}-R_0)(1-e^{-kt})$$\n", "\n", "Where:\n", "\n", "* $R$ is the mean reading rate [CPM].\n", "* $t$ is the total time spent reading [min].\n", "* $R_0$ is the initial mean reading rate [CPM].\n", "* $R_{max}$ is the maximum reading rate that the given second-language learner will be able to achieve [CPM]. As described in the literature review section, a value of 255 CPM is a good estimate for this parameter.\n", "* $k$ is an exponential decay coefficient [min${}^{-1}$].\n", "\n", "In the exponential decay model, the mean reading rate $R$ will asymptotically approach the maximum reading rate $R_{max}$ as the total reading practice time $t$ increases. A higher value for the exponential decay coefficient $k$ implies that $R$ will approach $R_{max}$ more quickly.\n", "\n", "The exponential decay model above may be more appropriate than the linear model because of [Zipf's Law](https://en.wikipedia.org/wiki/Zipf%27s_law). If one's reading rate is proportional to the size of one's vocabulary, and a word only becomes part of one's vocabulary when one has read it several times, then more and more reading will be required to learn rarer and rarer words. The simulation below demonstrates this phenomenon." ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [ { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgIAAAFSCAYAAACAI7emAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4xLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvDW2N/gAAIABJREFUeJzs3XeYFeXZx/HvTe8dlqWDgHQQELGvXawxsYs9msREU16NGpNoNEZjjNHEqLF3DfaGBZG1glJEeln6wrKUpS2w/X7/mCEe1y2z9Wz5fa7rXHum3/Occ3bueeaZZ8zdERERkfqpQbwDEBERkfhRIiAiIlKPKREQERGpx5QIiIiI1GNKBEREROoxJQIiIiL1mBKBGsjMks3s/mrYTh8zczMbW9XbKmb715rZ6nhsu7Yws9Vmdm2844gXM+sUfkeT4h2LVIyZ3W9myTHDT5rZ23EM6X9Ki8XMLjGzzKjDtY0SgWpmZp3N7IHwH3y2maWb2VQzOy5mth8CN8YrxpJU9483PAicWV3bq4EOBB6oyArMrIWZ/cXMUswsy8y2mNnnZnZeJcUYV2aWFH5POsU7lprMzG4Jy8nNrMDMNpjZc2bWM04h/RKYWNUbMbOGZna9mS02sz1mts3MZpnZNWVYzX+BflUVY7w1incA9dArQAvgciAF6AIcCXTcN4O7Z8QntPrHzBoB+V5De9Zy982VsJqHgEMJ/vEuADoAB4V/pX5ZCiQRnATuB/wbmAQcXN2BuPuOatrUzcBVwC+Ar4BWwAFAr6grcPe9wN4qia4mcHe9qukFtAMcOLaU+ZKB+2OGVwN/BJ4EdgHrgHPC9b0IZALLgeNjlkkKt9UpZlyfcNzYYoYbAo8Bqwi+9MuB3wINwum3hPPHvpLCad3DWLaFr3eAAYX267fAxjDep8P1rS6lLBw4s9DwlcBLwG5gJTCx0DLdgOeArcAeYC5wVMw+LAAuAVYA+QT/GCyMb0W47/OLWO+dBP9I94afyV1As5jpPYE3gIxwu0uAc2Oml1pGRez/auDasux/EevYDvy4lHmi7H+x5RpO/wlBcpsT/r2iiM+ytM/uQGA2kAV8DZwc+z0rJvYkCn3Xi1jnB8AWYCfwGXBwOWI7CJgTE9tJfPc38L04KONvLJynEfCPmO/JP4AHgeSyfF5FlMMtwIJC464O42sTM64J8FcgNSyLmcAJMdOj7END4O6Yfbi3iH14Eni70P+9B4C/hJ/VpnAdsetNAN4Mt7sGuJTg93xLCfs9F/hzKWVTOJaRQBpwezh8CZAZM73E4dr20qWB6pUZvk4zs2ZlXPZXBNnsaIIM/ingeWAyMAr4BHi2HOuN1QBYD5wNDAZuAn5H8GOD4Ec5CfgQSAxfX5hZC2AawT/IIwnOLtKAD8NpmNnZwJ8JsvPRBAfU35Qzzj8SHHBHElTZPW5mvcPttAQ+JvgHfAYwHLi10PJ9gfOBs8J1ZIWxXQ78HBgC3AH8x8xOjlluN3BZWDZXAeeGZbTPAwS1PUcBQwk+s+1hXKWWUWXsfzE2AieaWdsS5ilx/0srVzM7A7if4B/+MOA+4AEzOzVq7OE23iE4CI8FbiD4zlVUa+AZ4HBgHMGBYXIRlxJKiq0V8DZBcjeG4MD3t3LEUtpvDOBaggPLj4Hx4TLnF1pPlO9ricysK8FlyPzwtc8TBN/R8wk+56eAt8xsZBn24f+AKwiSw4MJEoMLIoR1AZAHHEJwBv8rgpOefZ4CegNHA6cTXFoo6bsPwfc/ycwSImwfMzuc4Ld6l7vfVNr8dUK8M5H69gJ+RHDGmAVMJ/hHd1CheZL5fo3ACzHDrQiy+H/GjOvDd888kihjjUAx8d4JfBgz/CQxmXM47jKCswKLGdeQ4Mzx7HD4C+CRQst9SPlqBO6IGW5EcHY6MRy+gqDWpLizw1uAXCAhZlxLgjOMwwvNey8wuYTYfgqkxAzPA24uZt5Sy6iY5Vbz/RqBYve/mHUcQVCLlEtwRns/cFxZ9j9CuX4OPF5o3JPAZ2X47K4kSJxaxcwzkQrWCBQxvxEkYRPLENtPCH63zWPmOZ8y1ghE/I2lATcUincJ4dl0Bb6vtxAc8DPDffPwdV/MPPsBBUCvQsu+DjxQhn3YANwUM9wAWEbpNQLTC613CvBo+H7/MN7xMdN7hvt0SwmxDQEWhfu1EHiUIAGywrEApxDUGl1UaB2XUIdrBNRGoJq5+ytm9g7B2cnBwInA/5nZTe7+lxIWnRezjkwz20NQHbhPevi3S0XiM7OfEpyJ9AaaA40JquBKMobgLHuXmcWOb0HwjwWCM4dHCy03HehfjjBjyyLPzDbz7X4fAMxz9y0lLJ/q7ukxw0OAZsB7ZuYx4xsTHIgBCBst/iqMuRXBgbxhzPz3AQ+Z2YnAVOA1d58dTotSRlGVtP/f4+6fmFk/grPLQwnOpj4ws4fd/SdE2//SynUw8HihcZ8Bp5Uh9sHhNmJbX0+PXdjMFvLtGeCn7j6hmHhil+kC3EZQU5NA8Jk15/vXiEuKbRBBtXrsdeIvS9t2MfEU+xsLa226EtT+7YvFzWwmwUEPIn5fi7GC4JJGU4Iz6h8RnM3vM5og8VhU6HvaFPioDPuQSMxn5+4FZvZlzD4UZ16h4Q189zMoAGbFrHedmW0oaYXuvsjMhhH8Bg8jSIwnEfwGTnH3gnDWMcBrwPnu/lIpcdYpSgTiwN2zCDLdKcCtZvYocIuZ3e3uOcUsllt4NYXG7fuHsO9yz74vd+yvuXFJcZnZOQRnFdcSnMHvJKh6PKOk5cJtziWoKi+sKho+FlUW+/bbKN3uQsP7lj0VWFvUtsxsPMH1/T8BvyY4cz2NmKprd3/MzN4n+Ed7LMFlkzvc/RYqt4xK2v8iuXsu8Gn4utPMfg/cZmZ3EGH/iVauHmFcRT+7k/j2exy18dZTBAnArwkOlNkEiVqTMsZW1P7FKvU3V4bfWEnbivJ5FSfH3VPC9wvNbABBg8FLYtbtBO0qCq9rbxn3oTwq+v0oUniwnxm+/mFmEwkuFx1BUBMBQZuHTcBlZvamu2eXd3u1jRKBmmERwWfRjKChVWXY19o8Meb9qFKWOQz40t3/14eBmRU+W83hu2fBEFQ3nwdscfftxax7McEZaexZ4/hS4imPOcBEM+tUSq1ArEUEB4fe7v5RMfMcCqx399v2jSjqury7pwIPAw+b2fUELfVvIVoZVadF4d9WRNv/0sp1McH3J/bzPSxmO1FjutjMWrr7vmTtO98Rdy+tdqoohwHXuPs7AOG14sQyrmMxcJGZNY+pFRhXaJ4ov7kSf2PuvsPMNobrnhZON4ID88ZwtiifV1S3AUvN7F9h7dXXBAfcru4+rZhlouxDGsFn91HMPowjuOxRXosJkoIxhLUxZtaDoBFrWcV+//fJIEjupwKvmdkZ9SUZUGPBamRmHc3sIzObaGYjzKyvmZ1F0PBoqrvvrMTNpRBcF77FzAaa2fHA70tZZhkw2swmmNkAM/sDQaOhWKuBYWa2vwWdvTQmaEmeDrxhZkeG+3WEmf09POOAoNr8YjO7Ilz3jQStsCvb8wRZ/etmdngYy2lmdlRxC7j7LoIz+7vN7DIz629mo8zsp2Z2ZTjbMqC7mV1gZv3M7GcEB/b/MbP7zOzEcPoogss++/7hRCmjKmFBB1U/MbMxFnQidRJBy+ylwOKI+19auf4NuNDMfh5+vlcTNPy6qwyhPk/QUOxxMxtqQd8aZWmsNSyMO/a179r0RDMbYmYHEtTslDXhfo7gWvQj4XqO5dsq9X1n71F+c1F+Y/cBvzWzM8xsf+DvBMlF0Jgh2ucVibuvJGiFf1s4vCzc1yfN7MzwuzzWgs6/fliOfTgz3Id7KXvyVTjWpcD7BJffxoe/sSf4tr1DkczsZTP7tZkdZGa9Leic6t8E3+cvCm1jC3AM0AN41cyaViTm2kKJQPXKBGYQnCV+TNBw5S8E/wDPKWG5Mgurgs8l6ATjG4Iq7d+VuBD8h+Da2fMEVWh9CP4JxXqEIDOfRXDWc6i77yGoYltJcPvVEoLq2PYEtw7h7v8lODO+neCsYzhwT/n3sGjhmeSRBK2a3yIo4z9RerXuH8L4rg2XmUJw/XRVuN63CA529xJcxzyOoJV5rAbAvwgO/lMIDvwXh8uXWkZV6H3gwvDvEoK7Gz4laDC4r7V4aftfYrm6++sEt6L9mmD/fwlcFZZbJGHbgFOAAQQ1EHcD15dhP6cRfLdiXy0IGmq2Irgt8UWCWovVZVjvvthOJbgb5GuC78It4eSscJ4ov7kov7G7CaqtnyD4fwHBteusmHlK/LzK6O/ABDM7JBy+NNz2XQTfl7cJvrv7amOi7MPfw3U8SnD23oAgwaioSwhua0wmSGCeIzigZxW/CO8T3Ib6JkES8wzBvhztRfTZEiYDRxO0Z3ilPiQDFrZ4FBGRMjCz0wkO0F3KcBmqvNuaA3zu7ldX5XZqGwtuAd0AnOfur8Q7ntpKbQRERCIws4sJanTWEfSVcC/wVmUnARa0PTmBoNawEcFtlSPDv/WamR1N0C/EfIK7CW4n6HzovXjGVdspERARiSaBoLo/kaDh3juU7dJFVAXARQSXHxoQXGqZ4O6zSlyqfmhM0JlSP4K2AV8CR8Q0LpVy0KUBERGRekyNBUVEROoxJQIiIiL1WL1oI9CpUyfv06dPpa5z9+7dtGzZslLXWd+oDCtOZVhxKsOKUxlWXFWU4ezZs7e4e+fS5qsXiUCfPn2YNaty29kkJyeTlJRUqeusb1SGFacyrDiVYcWpDCuuKsrQzCL1xKlLAyIiIvWYEgEREZF6TImAiIhIPaZEQEREpB5TIiAiIlKPKREQERGpx5QIiIiI1GNKBEREROoxJQIiIiL1mBIBERGROFq5OZP3V+eSk1cQl+3Xiy6GRUREaoq8/AJmrt7G1MXpTF2yiVVbdgNw5vrtjOndodrjUSIgIiJSxXbsySV52SamLt5E8tJN7MzKo0nDBhy8X0cuO7QPLbavjEsSAEoEREREqsSqLbuZujidDxenM3P1NvILnI4tm3D80K4cO7gLhw3oTKumwWE4OXl13OJUIiAiIlIJ8vILmL1mG1OXbOLDxems3BxU+Q/q2pqfHtmPYwYnMLJHOxo2sDhH+l1KBERERMppZ1YuHy/dzNTF6SQv28z2Pbk0bmiM79eRiw/uw9GDutCzQ4t4h1miak0EzGw1sAvIB/LcfayZdQD+C/QBVgNnu/s2MzPgPuAkYA9wibvPCddzMfD7cLV/dvenqnM/RESk/lqzdTcfLt7E1MXpfLUqg7wCp0PLJhwzKIFjB3fh8IHfVvnXBvGI9Ch33xIzfAMw1d3vNLMbwuHrgQnAgPB1EPAgcFCYONwMjAUcmG1mb7r7turcCRERqR/yC5w5a7fx4eJ0pi7eRMqmTAAGJrTiiiP6cezgLozq2b7GVflHVRNSltOBpPD9U0AyQSJwOvC0uzsww8zamVliOO8Ud88AMLMpwInAC9UbtoiI1FW7snL5ZNkWpi5OZ9rSTWwLq/wP6tuRCw7qxTGDEujVsWZX+UdV3YmAAx+YmQP/cfeHgQR3TwNw9zQz6xLO2x1YF7NsajiuuPEiIiLlti5jDx+Grfy/WpVBbr7TvkVjjtq/C8cMTuCIgZ1o3axxvMOsdNWdCBzq7hvCg/0UM1tSwrxF1bF4CeO/u7DZlcCVAAkJCSQnJ5cj3OJlZmZW+jrrG5VhxakMK05lWHG1tQwL3FmxvYCvN+XzzeY81mcGh5JuLY3jejViVJeG9G/XgAa2HTK2M3vG0iqLJZ5lWK2JgLtvCP9uMrPXgHFAupklhrUBicCmcPZUoGfM4j2ADeH4pELjk4vY1sPAwwBjx471pKSkwrNUSHJyMpW9zvpGZVhxKsOKUxlWXG0qw11ZuXy6fAsfLk4neelmMnbn0KiBMa5vBy5LChr79e7YstrjimcZVlsiYGYtgQbuvit8fzxwK/AmcDFwZ/j3jXCRN4FfmNmLBI0Fd4TJwvvAX8ysfTjf8cCN1bUfIiJSu6zL2PO/7nxnrNxKbr7Ttnljjtq/M8cMTuDI/TvTpg5W+UdVnTUCCcBrwV2BNAKed/f3zGwmMMnMLgfWAmeF808muHUwheD2wUsB3D3DzG4DZobz3bqv4aCIiEh+gTN33fbg4L94E0vTdwHQr3NLLj20L8cM6sKY3u1p1FDP3YNqTATcfSUwsojxW4FjihjvwM+LWdfjwOOVHaOIiNROefkFfLUqg8kL0nhvQTpbMrNp2MAY16cDvz95MMcMTqBvp+qv8q8NasLtgyIiImWWm1/AFyu28u78ND5YlE7G7hyaN27I0YO7cPyQBJIGdqFti/pb5R+VEgEREak1svPy+TxlC5Pnb2TKonR27M2lVdNGHDO4CxOGJXLkwM40b9Iw3mHWKkoERESkRsvKzeeTZZt5d8FGPlyUzq7sPFo3a8RxQxI4aVgihw3oRLPGOviXlxIBERGpcfbm5JO8dBOTF2zko8Xp7M7Jp12LxkwY3pUJwxM5dL9ONGmkxn6VQYmAiIjUCLuz8/hoySbeXZDGtCWb2ZubT4eWTThtVHdOGt6V8f060lgt/SudEgEREYmbnVm5fLR4E5Pnp/Hxss1k5xXQuXVTzhzTgwnDuzKuTwfd5lfFlAiIiEi12rEnlymL03l3fhqfLt9CTn4BXds047xxvThpeCJjetfeJ/nVRkoERESkymXszmHKoo1Mnr+Rz1O2kFfgdG/XnIsO7s2E4Ykc0LMdDXTwjwslAiIiUiW2ZGbz/sKNvDt/I9NXbiW/wOnVoQWXH96Xk4YlMqJHW8LeZiWOlAiIiEil2bQzi/cWbmTy/DS+WpVBgUPfTi356ZH9mDAskaHd2ujgX8MoERARkQrZureAxz9bxbsL0pi1ZhvuMKBLK35x9ABOGt6V/RNa6+BfgykREBGRMluXsYf3Fmxk8oI0vl67F1jEoK6t+fWxA5kwrCsDElrHO0SJSImAiIhEkrptD5Pnp/HOvDS+Sd0BwLDubThzQGOuOu0Q+nVuFecIpTyUCIiISLHWb9/Lu/PTeHteGnPXbQdgRI+23DBhECcNS6RXxxYkJycrCajFlAiIiMh3pO3Yy+T5G3ln3gbmrA0O/sO6t+H6Ewdx8vDg4C91hxIBEREhfWfW/6r9Z63ZBsCQxDZcd8L+nDw8kT6dWsY5QqkqSgREROqpTTuzeHfBRt6Zl8bMNRm4w6Curbn2+IGcNDxR1f31hBIBEZF6ZPOubN5bEFzz/2p1cPDfPyFo7X/S8ET6d9HBv75RIiAiUsdtyczmvfDM/8tVWylw6N+lFb88ZgAnD0/UrX71nBIBEZE6aGtmNu8vTOed+RuYviI4+Pfr3JJfHD2AU0YkMlAHfwkpERARqSO27c7h/YUbeWd+Gl+sCPr279upJT8/qj8nj0hUD39SJCUCIiK12PY9OXywMJ2356fxecoW8guc3h1b8NMj+3Hy8G4MTtTBX0qmREBEpJbZsSeXDxYFZ/6fLQ8e6durQwuuPKIfJw/Xg32kbJQIiIjUAjuzcpmyMJ135qfx6fLN5OY7Pdo35/LD+3LK8G4M666Dv5RP5ETAzHoChwNdgAax09z9nkqOS0Sk3isocL5YsZVJs9bx/sKNZOcV0L1dcy49tC8nD09kRI+2OvhLhUVKBMzsAuBxIA/YDHjMZAeUCIiIVJI1W3fzyuxUXpmznvXb99KmWSPOHtuTM0Z354Ce7XTwl0oVtUbgVuDvwB/cPb8K4xERqZd2Z+cxeX4aL81O5atVGZjB4QM6c8OEQRw3JIFmjRvGO0Spo6ImAgnAo0oCREQqj7sza802Js1cx+T5aezOyadPxxZcd8L+/HB0dxLbNo93iFIPRE0EJgMHASurMBYRkXohbcdeXpmdysuzU1m9dQ8tmzTk5BGJnDW2J2N7t1fVv1SrqInAFOCvZjYUmA/kxk5091crOzARkbokKzefKYvSmTRrHZ+lbMEdDurbgV8cPYAJw7rSsqlu4pL4iPrN+0/493dFTHNAF69ERApxd+av38FLs1J5Y+56dmbl0a1tM64+qj8/GtOD3h31aF+Jv0iJgLs3KH0uERGB4CE/r3+9npdmpbI0fRdNGzXgxGFdOXNMDw7ZrxMNG6jqX2qOqLcPNlRDQRGR4uXmFzBtySZemp3KtCWbyCtwRvZsx+1nDOOUEd1o27xxvEMUKVLUSwM7zOxzIDl8faXEQEQElmzcycuzUnl97nq2ZObQqVVTLjusL2eN6aHH+0qtEDUROAM4EjgZuAXIMbMvCBMDd59eJdGJiNRAO/bk8uY363lpdirzUnfQqIFxzOAunDWmJ0fu35nGDXU1VWqPqG0EphDcOYCZNQcOBS4g6GioAWosKCJ1XEGB82nKFl6atY4PFqWTk1fAoK6t+cMpQ/jBqG50bNU03iGKlEtZnjWQACSFr6OBXsDnwLSqCExEpCbYsSeXl2av4+npa1ibsYd2LRpz3oE9OWtsTz3lT+qEqI0FFwJ9gK8ILgf8BJju7tlVFpmISBwtTtvJ09NX89rX68nKLeDAPu259oT9OWFoAk0bqRJU6o6oNQJtgXxgD7Ab2AXkVFVQIiLxkJtfwAcL03lq+mq+WpVBs8YN+MGo7lx4cG+Gdmsb7/BEqkTUNgI9zKw/314a+CXQysw+Baa5+z+ibtDMGgKzgPXufoqZ9QVeBDoAc4AL3T3HzJoCTwNjgK3AOe6+OlzHjcDlBMnJNe7+ftTti4gUtmlXFi9+tY7nvlxD+s5senZozu9OGsTZY3vSrkWTeIcnUqUitxFw9xQgxcyeAMYBVwITCe4kiJwIECQRi4E24fBfgX+4+4tm9hDBAf7B8O82d+9vZueG851jZkOAc4GhQDfgQzMbqNsZRaQs3J2v123nP99kMWvKR+TmO0cM7MxfzuhN0v5d1OmP1BtR2wgcCBwVvg4FmgJfEzyaOHJjQTPrQZA43A78xoJWNkcD54ezPEVwe+KDwOnhe4CXgfvD+U8HXgzbJ6wysxSCxES3MIpIqbJy83nrmw08PX0N89fvoHkjmDi+DxeO702/zq3iHZ5ItYtaI/A5MBv4GLgP+NTdd5dje/cCvwX29bLREdju7nnhcCrQPXzfHVgH4O55ZrYjnL87MCNmnbHLiIgUKXXbHp6dsZb/zlzLtj25DOjSitt+MIxOmSuZcOzQeIcnEjdRE4H25Tzw/4+ZnQJscvfZZpa0b3QRs3op00paJnZ7VxJcviAhIYHk5OSyhlyizMzMSl9nfaMyrDiVYcncncUZBXy4JpevNwVXD0cnNOTYoc0Y1KEAy1pFZtZulWEF6XtYcfEsw6iNBXcDmNnRwBCCA+8idy9LHwKHAqeZ2UlAM4I2AvcC7cysUVgr0APYEM6fCvQEUs2sEcGdCxkx4/eJXSY25oeBhwHGjh3rSUlJZQi1dMnJyVT2OusblWHFqQyLlpmdx6tzUnnqi9Ws2JxFh5ZNuOqoPpx/UG+6t2v+nXlVhhWnMqy4eJZh1DYC3YHXgdF8e9DtZmazgDPc/XsH4sLc/UbgxnB9ScC17n6Bmb0EnElw58DFwBvhIm+Gw9PD6R+5u5vZm8DzZnYPQWPBAQT9G4hIPZeyKZNnpq/mlTnryczOY2SPttxz9khOGp5Is8a691+kKFEvDfwTyAP6u/sqADPrBzwbTjuzAjFcD7xoZn8maID4WDj+MeCZsDFgBsGdArj7QjObBCwKY/q57hgQqb/yC5yPlmzi6emr+XT5Fpo0bMApIxK56JA+jOrZLt7hidR4UROB44CkfUkAgLuvNLNrgKll3ai7JxP0UIi7ryRo9V94nizgrGKWv53gzgMRqacyducwadY6npm+hvXb95LYthnXnbA/5xzYk07q918kssj9CBSjoFKiEBGJwN2Zs3Ybz85Yyzvz08jJK2B8vw784ZTBHDs4gUZ66p9ImUVNBKYC/zSz89x9HYCZ9SK4lbDMNQIiImWRmZ3H61+v59kZa1iycRetmjbivAN7csH43gxMaF36CkSkWFETgWsIGvGtNLMNBHcNdAfmhdNERCrd0o27eHbGGl77Omj8NySxDX85Yzinj+pGy6YVrdAUEYh+++A6YLSZHQcMIriXf5G7f1iVwYlI/ZOdl897Czby7Iw1zFy9jSaNgsZ/E8f35oCe7fTYX5FKVqaU2t2nAFOqKBYRqcfWZezh+a/WMmnmOrbuzqF3xxb87qRBnDWmJ+1b6sE/IlUlciJgZgcBxwBdgO+0yHF3XR4QkTLLL3CSl27i2RlrSF62GQOOHZzAxPG9Oax/JxrowT8iVS5qh0LXAncBKQQdCsV26fu97n1FREqyJTOb/85cx/NfrmX99r10bt2Uq4/qz7njetGtUM9/IlK1otYI/BK4xt3vr8pgRKTucne+WpXBs1+u5b0FaeTmO4fs15GbTh7McUMSaKxb/0TiImoi0AaYXJWBiEjdtCsrl9fCW/+WpWfSulkjJo7vzQUH9aZ/Fz32VyTeoiYCLwAnAg9UYSwiUocs2rCTZ2as4Y2569mTk8/w7m2560cjOHVkN5o3Ub//IjVF1ERgHfAnMzuUoO+A3NiJ7n5PZQcmIrWPu5O8bDOPfLKSL1ZspWmjBpw2shsTx/dmpPr9F6mRoiYCPwYygUPCVywHlAiI1GM5eQW8MXc9j3y6kmXpmXRt04wbJwzinAN70q6Fbv0TqcmidijUt7hpZqYWPiL11I69uTz/5Vqe/GIV6TuzGdS1NX8/aySnjuxGk0b61yBSG5SYCJjZye7+TgnTGwBPAxMrOzARqbnWb9/L45+t4sWv1rI7J5/D+nfirjNHcsSATur5T6SWKa1G4CUzO9bdvyg8wYJf+xPAhCqJTERqnIUbdvDIJyt5a14aAKeOSOTHh/djWPe2cY5MRMqrtETgbuBtMzvS3ecXmvYYcDpwXJVEJiI1grvz6fItPPzJSj5L2ULLJg259JA+XHpYX7qr8x+RWq/ERMDd/2hmnYD3zewQd18NYGaPAGcCJ7j7zKoPU0SqW05eAW/P28DDn6xkycZdJLRpyg0TBnHeuF60bd443uGJSCWJ0ljw50AH4EN9FexVAAAgAElEQVQzOwy4BTgXmODu06swNhGJg51Zubzw5Vqe+Hw1G3dmsX9Ca+4+aySnqQGgSJ1UaiLg7m5mE4G3gUVAE+Bkd/+sqoMTkeqTtmMvT3y+mue/XEtmdh6H7NeRO340nKSBndUAUKQOK+2ugd/EDH4BJAHvA2PMbMy+CepQSKT2Wpy2k0c+Wcmb32zAgZOHJ3LF4f0Y3kMNAEXqg9JqBK4uNJwGjAhf+6hDIZFa6KtVGTyQnELy0s20aNKQiw7uw6WH9qFnhxbxDk1EqlFpjQWL7UhIRGofd+ejJZt4MHkFs9Zso2PLJlx3wv5MPKg3bVuoAaBIfRS1i2ERqcXy8gt4Z34aDyavYMnGXXRv15w/nTaUs8f21AOAROo5JQIidVhWbj4vz07l4U9WsjZjDwO6tOKes4MugBs31B0AIqJEQKRO2pWVy3NfruWxz1axeVc2o3q24/cnD+bYwQk0aKA7AETkW0oEROqQrZnZPPH5ap6evpqdWXkcPqAT9507ioP7ddQtgCJSJCUCInVA6rY9PPrpKl6cuZbsvAJOHNqVnyXtx4ge7eIdmojUcJESATPrDODum8Ph4cA5wEJ3f6HqwhORkqzPLOA3k+by5twNAJxxQHd+cuR+9O/SKs6RiUhtEbVGYBLwDPB4+OyBT4ANwNVm1s3d/15VAYrI981dt50HpqXwwaK9NG+8kQsP7s0Vh/ejmx4CJCJlFDURGAHMCN+fCaS4+4FmdjrwN0CJgEg1WLpxF39+ZxGfLt9Cm2aNOG2/xtxyfhIdWjaJd2giUktFTQSaA5nh+2OBN8P3c4CelR2UiHxXZnYe9324jMc/X02bZo343UmDOP+g3sya/pmSABGpkKiJwHLgh2b2CnA8QS0AQAKwvSoCE5GgJ8C356Xx53cWsWlXNuce2JPfnjCI9jr4i0gliZoI/Al4geASwFR3/zIcfwLwdVUEJlLfpWzK5OY3F/B5ylaGdW/DQxPHcECv9vEOS0TqmEiJgLu/ama9gG7ANzGTPgReqYrAROqrPTl5/OujFB79dCXNGzfkth8M4/xxvWiojoBEpApE7kfA3dOB9ELjvixmdhEpI3fn/YUbufWtRWzYkcWZY3pww4RBdGrVNN6hiUgdVmwiYGaPR12Ju19WOeGI1E+rtuzmljcX8vGyzQzq2pr7zjuAA/t0iHdYIlIPlFQj0LnQ8BFAATA/HB4GNCDoU0BEyiErN58HpqXw0McradKoAX88ZQgXHdybRnogkIhUk2ITAXc/dd97M7sR2Atc6u67w3Etgcf4NjEQkTKYujidW95ayLqMvZw+qhs3nTSYLm2axTssEalnorYRuAY4Zl8SAODuu83sNmAqcHtVBCdSF63L2MOf3lrIh4s30b9LK56/4iAO2a9TvMMSkXoqaiLQiuCOgUWFxicCLaKswMyaEVxGaBpu92V3v9nM+gIvAh0IOii60N1zzKwp8DQwBtgKnOPuq8N13QhcDuQD17j7+xH3QyRusvPyefjjldw/LYWGDYwbJwzi0kP70qSRLgOISPxETQReAZ4ws+v4tqvh8cBfgVcjriMbONrdM82sMfCZmb0L/Ab4h7u/aGYPERzgHwz/bnP3/mZ2britc8xsCHAuMJQgOfnQzAa6e37EOESqVXZePi/NSuWBaSls2JHFScO78odThpDYVs8FEJH4i5oI/IygM6EngcbhuDyCNgLXRlmBuzvfdlPcOHw5cDRwfjj+KeAWgkTg9PA9wMvA/RY8UP104EV3zwZWmVkKMA6YHnFfRKpFdl4+k8IEIG1HFqN7teNvZ43k0P66DCAiNUfURMCAq4HrgP3C4ZTYNgORVmLWEJgN9Af+DawAtrt7XjhLKtA9fN8dWAfg7nlmtgPoGI6fEbPa2GVE4i47L59JM9fxQPIK0nZkMaZ3e+46cwSH9e9EkMuKiNQcpSYC4cF7BzDS3RcB88q7sbD6fpSZtQNeAwYXNdu+TRczrbjx32FmVwJXAiQkJJCcnFyekIuVmZlZ6eusb+paGeYWOJ+k5vHOylwyspwB7Rpw3dhmDOmYTf76hXy8vvK3WdfKMB5UhhWnMqy4eJZhqYmAu+eb2Rqg0p5y4u7bzSyZoJ1BOzNrFNYK9AA2hLOlEjzZMNXMGgFtgYyY8fvELhO7jYeBhwHGjh3rSUlJlRU+AMnJyVT2OuubulKGWbn5TJq1jgemrWDjzhzG9m7Pv44byCH7dazyGoC6UobxpDKsOJVhxcWzDKNeGrgNuNPMJrr7lvJsyMw6A7lhEtCc4HHGfwWmAWcS3DlwMfBGuMib4fD0cPpH7u5m9ibwvJndQ9BYcADwVXliEqmIrNx8/jtzHQ8kp5C+M5sD+7Tn72ePrJYEQESkskRNBK4F+gLrzSwV+E7bAHcfEWEdicBT4aWGBsAkd3/bzBYBL5rZnwmeZPhYOP9jwDNhY8AMgjsFcPeFZjaJ4FbGPODnumNAqlPhBGBcnw784+xRHKwEQERqoaiJwMsV3ZC7zwMOKGL8SoJW/4XHZwFnFbOu21EnRlLNsvPyeeHLtTz48YogAejbgX+cM4qD+ykBEJHaK+pjiP9U1YGI1FQFBc5b8zbwt/eXkrptL+P6duDecw7g4P06xjs0EZEKi/wYYgAzOxoYQtBKf6G7J1dFUCI1xecpW7jj3cUsWL+TIYlteOby4Rw+oPDzuEREaq9IiYCZdSe43W8M37bQ72Zms4Az3P17rfZFarPFaTu5890lfLxsM93bNefec0Zx2shuNGigSwAiUrdErRH4J0G//v3dfRWAmfUDng2nnVk14YlUrw3b93LPlGW8MieVNs0ac9NJg7nw4N40a9ww3qGJiFSJqInAcUDSviQAgkZ+ZnYNwdMHRWq1HXtzeTB5BU98vgoHrjy8H1cl9adti8alLisiUpuVqY1AEQoqJQqROMnOy+eZ6Wu4f1oKO/bmcsao7vzm+IH0aB/poZoiIrVe1ERgKvBPMzvP3dcBmFkv4D5UIyC1UOE7AQ4f0InrTxzEsO5t4x2aiEi1ipoIXEPQ499KM9tAcNdAd4LnDlxTRbGJVIkvUrZwx7tLmL9+B4MT2/D0ZcM5YqDuBBCR+ilqPwLrgNFmdhwwiODBP4vc/cOqDE6kMi3csIO/vreUT5ZtplvbZtxz9kh+MKq77gQQkXqtxETAzPq4++p9w+4+BZhS1UGJVKZ1GXu4Z8oyXp+7XncCiIgUUlqNwMrwyYMfETwc6CP1GSC1RcbuHO7/KIVnZ6zBDH5yxH78LGk/2jbXnQAiIvuUlgicABwJHAVcCDQ0s+WESQGQ7O6bqzZEkbLZm5PP45+v4qHkFezOyeOsMT351XEDSGzbPN6hiYjUOCUmArGXAsysJXA4QVJwJPBjoIGZLXD3kVUdqEhp8vILeGl2Kv+YsoxNu7I5dnACvz1xfwYmtI53aCIiNVbkfgTcfTfwnpl9ABwInA78HBhWRbGJROLuvL8wnbveX8LKzbsZ07s9/75gNAf26RDv0EREarxSEwELnq86mqAm4CjgMGAX8DHwWyC5CuMTKdGSjTv54xsL+WpVBvt1bsl/LhzD8UMS9FhgEZGISrtr4A3gCIID/yfA68Cv3H15NcQmUqydWbn8Y8oynp6+hjbNGnH7GcM4Z2xPGjVsEO/QRERqldJqBE4F1gJPEZz5f+Hu2VUdlEhx3J1X56znjneXsHV3NueP68W1x+9P+5ZN4h2aiEitVFoi0JNvLwk8BiSa2VcESUEySgykGi3asJM/vrGAWWu2MapnO5645ECG91CXwCIiFVHaXQPrCR41/CwEHQwBSeHrSaCLmX3p7klVF6LUdzv25nLPB0t5ZsYa2rVowl0/GsGZY3qoR0ARkUpQpqcPuvtqM/uIb586eAbBLYUila6gwHl5Tip/fXcJ2/bkMHF8b/7vuP31aGARkUoU5a6Bbnx7eeAooA+QA0wH7iboXEikUi3ZuJPfvTqfOWu3M7pXO566bJyeDCgiUgVKu2tgGbAfkAd8BTxHcOBX2wCpMq/OSeXGV+fTqmkj/nbmCH40WpcBRESqSmk1Aq8QdCX8mbvvrYZ4pB7LySvgtrcX8cyMNYzv14F/nTeazq2bxjssEZE6rbTGgjdWVyBSv23ckcVVz81mztrtXHF4X64/cZD6BBARqQZlaiwoUhVmrNzKL56fw56cfP59/mhOHpEY75BEROoNJQISN+7Oo5+u5I53l9C7QwteuGI8A/SAIBGRaqVEQOJid3YeD36TzVcbF3PC0ATuPmskrZvptkARkepW7EVYM3vczFqH748wMyUNUilWbM7kB//+nJkb87n+xEE8NHGMkgARkTgpqTXWRKBl+H4aoGe6SoW4O69/vZ7T7/+crbtzuHZsM36WtJ+eFCgiEkclneWvBq42sw8AAw42s21Fzejun1RBbFKHpO3Yy02vLeCjJZs4oFc77j9/NMvnfhnvsERE6r2SEoHrgEeAGwEHXitmPgcaVnJcUkcUFDgvzFzLHZOXkFdQwO9PHsylh/alYQNDz7IWEYm/YhMBd38DeMPM2gEZwFBgU3UFJrXf6i27ueHVecxYmcEh+3Xkzh+OoFfHFvEOS0REYpTaANDdt5vZUcByd8+rhpiklssvcB7/bBV/n7KUxg0acOcPh3POgT3VFkBEpAaKdCeAu39sZk3N7CJgCMHlgEXA83rmgMRaunEXv335G75J3cGxg7vw5x8Mp2vbZvEOS0REihEpETCzIcC7QFtgfjj6CuAWMzvR3RdXUXxSS+TmF/Bg8gr+9dFyWjdrzD/PO4BTRySqFkBEpIaL2jfAfcBc4EJ33wlgZm2AZ4F7gROqJjypDRZt2Ml1L3/Dwg07OXVkN245dQgdW+lhQSIitUHUROBQ4MB9SQCAu+80s5uAGVUSmdR4OXkFPJCcwv0fpdCuRRMemjiGE4d1jXdYIiJSBlETgSygXRHj24bTpJ5ZuGEH1740j8VpO/nBqG7cfOpQ2rdsEu+wRESkjKImAm8Bj5jZFXxbA3Aw8B/gzaoITGqmnLwC7p+WwgPTUmjfsgkPXziG44eqFkBEpLaK+sD3XwLLgU8JagCygI+BZcCvoqzAzHqa2TQzW2xmC83sl+H4DmY2xcyWh3/bh+PNzP5pZilmNs/MRses6+Jw/uVmdnH03ZWKWLE5k9P//Tn/nLqc00Z2Y8qvj1ASICJSy0W9fXA7cLqZ9QcGE3Q5vMjdU8qwrTzg/9x9Tvgwo9lmNgW4BJjq7nea2Q3ADcD1wARgQPg6CHgQOMjMOgA3A2MJbmOcbWZvunuR3R9L5Xhj7npufHU+zRo35NGLxnLskIR4hyQiIpWgTE8UDA/8ZTn4xy6bBqSF73eZ2WKgO3A6kBTO9hSQTJAInA487e4OzDCzdmaWGM47xd0zAMJk4kTghfLEJSXLys3nT28t5IWv1jGuTwf+ed4B6hdARKQOicujhc2sD3AA8CWQECYJuHuamXUJZ+sOrItZLDUcV9x4qWQrN2fy8+e/ZnHaTn6WtB//d9xAGjWMejVJRERqg2pPBMysFfAK8KvwFsRiZy1inJcwvvB2rgSuBEhISCA5Oblc8RYnMzOz0tdZk8xIy+PJBdk0agC/GdOUEc028tmnGyt1G3W9DKuDyrDiVIYVpzKsuHiWYbUmAmbWmCAJeM7dXw1Hp5tZYlgbkMi3DzZKBXrGLN4D2BCOTyo0Prnwttz9YeBhgLFjx3pSUlLhWSokOTmZyl5nTZCbX8Ctby3imW/WMKZ3e/513gF0a9e8SrZVV8uwOqkMK05lWHEqw4qLZxmWWs9rZo3M7Coz61aRDVlw6v8YsNjd74mZ9Cawr+X/xcAbMeMvCu8eGA/sCC8hvA8cb2btwzsMjg/HSQXtzMrl0idm8syMNVx5RD9evHJ8lSUBIiJSM0R5+mCemf0NeKeC2zoUuBCYb2Zzw3G/A+4EJpnZ5cBa4Kxw2mTgJILGiXuAS8N4MszsNmBmON+t+xoOSvmlbtvDpU/MZNWW3fztzBGcNbZn6QuJiEitF/XSwAxgNLCmvBty988o+vo+wDFFzO/Az4tZ1+PA4+WNRb7rm3XbufypWWTn5fP0ZeM4pH+neIckIiLVJGoi8AjwdzPrDcwGdsdOdPc5lR2YVI/3FmzkV//9mk6tmvLilQfRv0vreIckIiLVKGoi8Hz4954ipjnQsHLCker06KcruX3yYkb2aMejF4+lk54YKCJS70RNBPpWaRRS7e7/aDl3f7CMCcO68o9zRtGssXI5EZH6KGoXw+VuGyA1z8OfrODuD5bxwwO687ezRtKwQbF9OYiISB0XuZs4M5tgZm+b2SIz6xmO+7GZfa+hn9RcT3y+ir9MXsLJIxK568wRSgJEROq5SImAmV0ATCJ4AmFfoHE4qSHw26oJTSrbc1+u4U9vLeKEoQnce84odRcsIiKRawR+C1zh7r8meIrgPjOAUZUelVS6SbPWcdNrCzh6UBf+dd5oGisJEBERoicCA4DpRYzPBNpUXjhSFV7/ej3XvzKPwwd04oELRtOkkZIAEREJRD0ibAAGFjH+CGBF5YUjle2tbzbwm0lzGd+3Iw9fOFZ3B4iIyHdETQQeBv5pZoeGwz3N7GLgLuDBKolMKmzy/DR+9d+5jO3TgccuGUvzJkoCRETku6LePniXmbUFpgDNgGlANnC3u/+7CuOTcnp/4UaueeFrDujZjscvOZAWTar9idMiIlILRD46uPtNZnY7MISgJmGRu2dWWWRSbh8uSucXz89heI+2PHHpgbRqqiRARESKVtYjhANZ4fv8So5FKsFHS9K56rk5DElsw1OXjaN1s8alLyQiIvVW1H4EmprZvUAG8A0wD8gws/vMrFlVBijRTZ6fxk+emc3+XVvz9GUH0UZJgIiIlCJqjcCDwPHAj/n2NsKDgTuA1sBllR+alMUrs1O57uVvOKBXe5649EAlASIiEknUROAs4IfuPiVm3Eoz2wS8ghKBuHruyzXc9NoCDu3fkUcuGquGgSIiElnUI8ZuYH0R49cDeysvHCmrxz9bxa1vL+LoQV144ILR6idARETKJGo/Av8Cbjaz5vtGhO//EE6TOHh5diq3vr2ICcO68tDEMUoCRESkzIqtETCzNwuNSgLWm9m8cHh4uHzLqglNSjJt6Sauf2Ueh/XvxH3nHqBug0VEpFxKujSwtdDwK4WGV1VyLBLR3HXbuerZOQxObM1DF45REiAiIuVWbCLg7pdWZyASzcrNmVz25Ew6t27KE5eMU2dBIiJSITqVrEU27sjiwse+woCnLxtH59ZN4x2SiIjUcpFOJ82sPXALcBTQhUIJhLt3qfTI5Du27c7hwse+ZMfeXF64Yjx9OqlphoiIVFzUeuWngaHAU0A6QVfDUk12Z+dx6ZMzWZOxh6cuHcfwHm3jHZKIiNQRUROBJOBId59ThbFIEbLz8vnps7OZv34HD14wmoP36xjvkEREpA6J2kZgRRnmlUqSm1/AL57/mk+Xb+HOHw7n+KFd4x2SiIjUMVEP7r8E7jCzkWamXmuqQX6B85tJ3zBlUTq3nj6Us8b2jHdIIiJSB0W9NJACNAfmAJjZdya6u5KDSlRQ4Nz46jze+mYDN0wYxEUH94l3SCIiUkdFTQReANoC16DGglXuqemrmTQrlWuOGcBPj9wv3uGIiEgdFjURGAuMc/cFVRmMwIrNmdz57hKOHtSFXx87IN7hiIhIHRe1jcAioE1VBiKQl1/AtS99Q/MmDbnzh8O/dwlGRESkskVNBH4P3GNmx5pZgpl1iH1VZYD1yX8+WcnXa7dz2+nD6NKmWbzDERGReiDqpYHJ4d8P+G77AAuH1Viwguan7uDeD5dx8vBETh3ZLd7hiIhIPRE1ETiqSqOo5zKz87j6hTl0atWU288YFu9wRESkHomUCLj7x1UdSH128xsLWZuxhxeuGE+7Fk3iHY6IiNQjUR86NLqk6ep6uPze+mYDr8xJ5ZfHDOCgfuo+WEREqlfUSwOzCNoCxDZjj20roDYC5ZCxO4eb31zIyB5tufro/vEOR0RE6qGoiUDfQsONgQOAm4AbKzWieuTWtxayKyuXu84cT6OGepSDiIhUv6htBNYUMTrFzHYANwPvVmpU9cC0JZt4fe4GfnnMAPbv2jre4YiISD1V0dPQVcCoKDOa2eNmtsnMFsSM62BmU8xsefi3fTjezOyfZpZiZvNi2yiY2cXh/MvN7OIKxh8XWzKzue7leeyf0JqrjlIXwiIiEj+REoHCHQiZWUczGwbcASyNuK0ngRMLjbsBmOruA4Cp4TDABGBA+LoSeHBfHAQ1EAcB44Cb9yUPtYW7c91L37AzK5f7zhtF00ZqXiEiIvETtY3AFr7/oCED1gHnRFmBu39iZn0KjT4dSArfPwUkA9eH4592dwdmmFk7M0sM553i7hkAZjaFILl4IeJ+xN3zX61l2tLN/Om0oQzqql6bRUQkvsrboVABsBlIcfe8Cmw/wd3TANw9zcy6hOO7EyQZ+6SG44ob/z1mdiVBbQIJCQkkJydXIMzvy8zMLPM6s/Ocv36yl4HtG9ArexXJyasrNabapjxlKN+lMqw4lWHFqQwrLp5lWFM7FCrqaTuFb1+MHf/9ke4PAw8DjB071pOSkiotOIDk5GTKus5/T0thZ85Snrh8PGN66xEN5SlD+S6VYcWpDCtOZVhx8SzDEhOBqA8U2ldVXw7pZpYY1gYkApvC8alAz5j5egAbwvFJhcYnl3Pb1Spjdw4PfbyCYwd3URIgIiI1RmmNBbcQXAIo6bWp2KVL9yawr+X/xcAbMeMvCu8eGA/sCC8hvA8cb2btw0aCx4fjarR9DQSzcwu47oRB8Q5HRETkf0q7NFDSw4ZOBH4JRGojYGYvEJzNdzKzVILW/3cCk8zscmAtcFY4+2TgJCAF2ANcCkHNg5ndBswM57u1ArUR1eaxz1Yxdckmbjl1iPoMEBGRGqXERKCotgHhPf1/BY4A/gPcFmVD7n5eMZOOKWJeB35ezHoeBx6Pss2aYOGGHfz1vSUcNySBiw/pE+9wREREviNyh0Jm1tfMnge+BDKAIe5+jbtvrrLoarnsvHx++eJcOrRswl0/GoFZUW0dRURE4qfURCDsPOg+YAnQFTjY3c9x9xVVHl0t98KXa0nZlMmdPxpB+5Z6vLCIiNQ8JSYCZvY7YAVwJHC6ux/t7rOqJbJabk9OHvdPW8H4fh1IGtg53uGIiIgUqbTGgn8G9hLctneVmV1V1EzuflplB1bbPfH5arZkZvPgxNG6JCAiIjVWaYnA0xTTYY8Ub2tmNg8mr+DYwQkc2Ed9BoiISM1V2l0Dl1RTHHXKA8kr2JOTxw0T9o93KCIiIiWq6GOIpZCNO7J4ZsYafjS6B/27qM8AERGp2ZQIVLInvlhFXn4B1xwzIN6hiIiIlEqJQCXKys3nvzPXcfyQrvTs0CLe4YiIiJRKiUAl+mBROtv35HLhwb3jHYqIiEgkSgQq0VvfbCChTVPG9+sY71BEREQiUSJQSXbsyeXjpZs5ZUQ3GjZQvwEiIlI7KBGoJG/P30BOfgFnHNA93qGIiIhEpkSgkrz+9Xr6d2nF0G5t4h2KiIhIZEoEKsH67XuZuXobPxjVTd0Ji4hIraJEoBK8MXc9AKeN1GUBERGpXZQIVJC78/rX6xnTuz29OqrvABERqV2UCFTQorSdLEvP5AdqJCgiIrWQEoEKenPuBho1ME4enhjvUERERMpMiUAFfbxsM+P7daRDyybxDkVERKTMlAhUwLbdOSxN38VBfTvEOxQREZFyUSJQAZ8s34w7HD6wc7xDERERKRclAhXw9drtNG/ckOHd28Y7FBERkXJRIlABM1ZuZUSPtnq2gIiI1FpKBMppW1YBSzbu4uhBXeIdioiISLkpESinpdsKADi0f6c4RyIiIlJ+SgTKaf2uAho2MAYmtI53KCIiIuWmRKCcUjML6N2hBU0aqQhFRKT20lGsHNydVTsKGNFDdwuIiEjtpkSgHNZm7GF7tjOmd/t4hyIiIlIhSgTKYVl6JgBD1X+AiIjUckoEymFtxh4A+nRsGedIREREKkaJQDmsy9hDs4bQvkXjeIciIiJSIUoEymHzrmzaNTXM1KOgiIjUbkoEymHH3lxaNlYSICIitZ8SgXJI35lFm6ZKBEREpPZTIlAOmzOzaa9EQERE6gAlAmXk7mRm5dG8kRIBERGp/WptImBmJ5rZUjNLMbMbqmu7WbkF5BU4umFARETqglqZCJhZQ+DfwARgCHCemQ2pjm3vzskDoGlD1QiIiEjtVysTAWAckOLuK909B3gROL06NpyxOweAxrW15ERERGLU1sNZd2BdzHBqOK7K5eQVALAnrzq2JiIiUrXM3eMdQ5mZ2VnACe7+43D4QmCcu18dM8+VwJUACQkJY1588cVK2XZugbNpt9OGPbRu3apS1llfZWZm0qqVyrAiVIYVpzKsOJVhxVVFGR511FGz3X1safM1qtStVp9UoOf/t3fuUXZUVR7+fiFEUMNDAwqCJECADrgIEBGBJY1hkME1iTOIhhERiMzwDM4oPvAVATHgAApLxSAxyqwhasYlbSaaNZpEEIgmSIIkIIQQHmIkQRMIeUDI9o99WiqVe7ur+3bfR9f+1qrVVadO1dn167pV++xz6pzM9j7AM9kMZjYVmAowZswYa29v71MD5s+fT1+fs2yEhrUTGtZOaFg7oWHtNFLDVm0aWAiMlDRC0hBgAtDRYJuCIAiCoOVoyYiAmW2RdDEwB9gBmGZmSxtsVhAEQRC0HC3pCACY2WxgdqPtCIIgCIJWplWbBoIgCIIg6APCEQiCIAiCEhOOQBAEQRCUmHAEgiAIgqDEhCMQBEEQBCUmHIEgCIIgKDHhCARBEARBiWnJuQZ6iqTVwBN9fNphwJo+PmfZCA1rJzSsndCwdkx6rvIAAAyLSURBVELD2ukPDfczsz26y1QKR6A/kLSoyGQOQXVCw9oJDWsnNKyd0LB2GqlhNA0EQRAEQYkJRyAIgiAISkw4Ar1naqMNGACEhrUTGtZOaFg7oWHtNEzD6CMQBEEQBCUmIgJBEARBUGLCEegCSadI+oOk5ZI+XWH/ayT9IO3/jaTh9beyuSmg4X9KWibpAUm/lLRfI+xsdrrTMZPv/ZJMUvTgzlFEQ0kfSPfjUkn/U28bm50Cv+e3Spon6f70mz61EXY2M5KmSXpW0oNV9kvSjUnjByQd2e9GmVksFRZgB+AxYH9gCLAEGJXLcyFwc1qfAPyg0XY301JQwxOB16b1C0LD3umY8g0F7gQWAGMabXczLQXvxZHA/cDuaXvPRtvdTEtBDacCF6T1UcDKRtvdbAvwLuBI4MEq+08FfgYIOAb4TX/bFBGB6hwNLDezFWb2EjADGJ/LMx74XlqfCYyVpDra2Ox0q6GZzTOzDWlzAbBPnW1sBYrciwBXAtcCm+ppXItQRMPzgG+Y2V8BzOzZOtvY7BTR0IBd0vquwDN1tK8lMLM7gb90kWU88H1zFgC7SdqrP20KR6A6bwGeymw/ndIq5jGzLcA64I11sa41KKJhlom4JxxsS7c6SjoC2NfMZtXTsBaiyL14EHCQpLslLZB0St2saw2KaDgZOFPS08Bs4JL6mDag6Olzs2YG9+fJW5xKNfv8JxZF8pSZwvpIOhMYA5zQrxa1Jl3qKGkQcANwdr0MakGK3IuD8eaBdjwydZekw8xsbT/b1ioU0fAMYLqZXSfpncBtScOt/W/egKHu75WICFTnaWDfzPY+bB/m+nseSYPxUFhXIZ+yUURDJJ0EfBYYZ2ab62RbK9GdjkOBw4D5klbi7Yod0WFwG4r+nu8ws5fN7HHgD7hjEDhFNJwI/BDAzO4FdsLH0A+KU+i52ZeEI1CdhcBISSMkDcE7A3bk8nQAH0nr7wfmWurtEQAFNEwh7W/jTkC0yVamSx3NbJ2ZDTOz4WY2HO9rMc7MFjXG3KakyO/5J3jnVSQNw5sKVtTVyuamiIZPAmMBJLXhjsDqulrZ+nQAZ6WvB44B1pnZn/qzwGgaqIKZbZF0MTAH7y07zcyWSroCWGRmHcCteOhrOR4JmNA4i5uPghp+FXg98KPUz/JJMxvXMKObkII6Bl1QUMM5wMmSlgGvAJeZ2XONs7q5KKjhx4FbJP0HHs4+OypH2yLpdrz5aVjqS/FFYEcAM7sZ71txKrAc2ACc0+82xf8oCIIgCMpLNA0EQRAEQYkJRyAIgiAISkw4AkEQBEFQYsIRCIIgCIISE45AEARBEDQR3U1MlMtb80RP4QgEQR8h6XxJaxptR3dIWpU+A2saJF2dHnwmqWk/w5W0pZnt6wpJxyd939xoW4JumQ4UHeL6c8APzewI/BP2b/a0sHAEggFDesh1tUwveJ4Zkmb2k42rki1vz6VPkVTKAYDSCIifwYdI3gsf2CefZ7Gkm3Jp5yUtL8qlXyfpkX40uVdIOiR3P66VdE/MaRDkqTQxkaQDJP1c0n2S7pJ0SGd2apzoKRyBYCCxV2Y5r0LapQ2yK88mfJbAAYOkQZJ26OXhBwIvmdlsM1tlZpVmT5xLGvUvQzs+kl2l9Hm9tIU0al5/0o7fj+8EHgB+Iumgfi4zaH2mApeY2VHAJ3i15j+ZGid6CkcgGDCkl8gqM1sFrM2nmdk68GGNJc2XtFHSc5K+I2lo2jcF+CBwWqbmdkzad72kR9Nxj0v6ci9fGjcDx3XVllcpQpBveujMk2rGT0paL+nbkgZL+pikP0paI+kaabvpsXdNkY8XJT0jaVKurDdIulXSaknPS5oraXTeFknj00h8LwEjqlxLd3rfDgxJWlebQnkecKikPTNp7cBXgPbO65O0GzAadxw6yx8raaGkzZL+JOlaSTtm9i+Q9PW0rAF+mdIPSTWvTZKWSTo5d12SdGXSfnPS8TtV7M/yXLofHwI+D7yG3GRbkv5F3ua7SdIKSZNzNp+TaoYvpCjTDOVC/pL+SdIjSfd5wP4FbAuaEEmvB47FR2BdjA/L3jk1cedET/vgIxLeJp+IrDDhCASlQtIu+BCpzwJvB04H3o2/nAGuAu4AZvFqJOG+tG8dcBbQBkzCh/68rBdmPJbKm9LTH2wFDgZOwtsTP5jsm5VsHAtciNce8k7HJ4HfAUcAVwPXdTomqWb/c3xK7X8EjgIWAXMl7ZE5x1D8+icChwLbjYdeUO+Lgc241vtVuc478WF/29N5DwZ2A76b9r8t/T0Bf67NT/mGA/+Hz79wOHAB/n+bnDv/ufhwrscC5yUNOpJd7wD+HXc6sv+vfwUuwqNPI/F55O+jIMmJ/GjafDmTPg6Yhs8oOQr4N+DD+FC0newIXJ6u6X34xDS3Zc5xAPC/wE9xx+gWYEpR24KmYxCw1sxGZ5a2tK/2iZ7MLJZYBtyCTwJlFdIvAdYAO2fSTgG2Avum7RnAzAJlfAx4MLN9PrCmm2NW4S++PYDn8bHYwR/SizL5ttmudP6U5wXgdZm0WXgb4eBM2gLgv3I2/DR37v8GfpHWT8XbJ4fk8jwMTMrYYsCh3VxvEb3PBDYV0Hsh8K1M+f+f1n8MXJrWv5b7n1wHLCUNp545dgOwY0afhbmyxuFRjr0yaSela56Qti8Hfg/sUPCePCQdvwFYnzQw4BFgl0y+3+LzHGSPnQD8pYtzj07nGpa2rwd+n8tzVcrz5nr8BmOpbQGG5+7le4DT07qAw9P6zzLPkbb0+1dPyoqIQFA22oD7zWxjJu3X+A+rrfIhjqQz5J27Vklaj7+I39obI8xsNT7h0hWSdurNORIrzOzFzPafgYfMbEsubc9tD+PeCtuj0vpRpCm1U3PD+nS9BwIHZI7ZaGZLu7Gv13pXYB6v9gc4kVTrT3+z6XMzx7QB91h6SmbK35ltmzLyHTXbgJW27axvec1mAG8AHpd0i6TTsuH7LjgNj8S8D48OnW1mz4M3N6R9V+S0nwbsLmn3lO9oSbNSs8QLwN3p3J33Y1sFe/PbQZMin5joXuBgSU9Lmgh8CJgoaQnu3I5P2T+OR7GW4M1sZ+fu926J2QeDsiG8VlSJqj8eSSfgodfPAb/AmwlOB75Qgy3X46H7Sp17tiZbs1R6ybyc27YqaT1x+gfhc6KPrbBvXWZ9Q4Fz9UrvKswFLpO0N94EcGNK/xUwOTVbvI1tw/5Fy38xty+v/fYHm62QdCDwD7hWNwKflXSsVe7w2MlTZvYo8KikzcBMSaPMbG0qdxB+n91R4djnUz+IOXjY/0P4NL9vwe/Lzj4r3dofNC9mdkaVXdt9YWJmy4DjaikvIgJB2VgGHClp50za8fhL4eG0/RI+zSq5PI+Z2RQzW5Qe5MNrMSTV5Cfjn87tntu9Gsh/7z2avuOYCtsPpfXfAXsDm81seW7p6dzyRfQuyq+BLbjzNBQPoYP3vDf8qxDj1UhBZ/nH5TpLHg9sBFZ2Y/fwXAe8vGaY2UYz6zCzS/H+BUcARxe9IDObg3/5cHna3gosBg6qoP1yM3sF74+xG/ApM7vLzB4G3lTB/kr/4yDYjnAEgrLxPby2PV3SYZJOBL4B3G5mT6U8K4HDJY2UNEzSYLwdd4SkD8i/552Eh3hr5Va8I91ZufS5wN6SLkvlnY+3W/cVJ0j6RLrGC/E26BvSvtm4M9Ah6WRJwyUdK+kqSe/oYTlF9C6Ema3H+wlMwsP9L6d0A+5K6UvM7K+Zw27CmzO+nr4CGA9cCdzQeXwVZgNPAN+XdLik44FryEQR5F9rnJOuawTwEdyJfKwn14X3Y7go43R8CThX0hckHSqpLd13V6f9j+NRn0mS9k+dC/ORqW8CbZK+Kulg+SBI5/bQrqAkhCMQlIrUFvsevAa1EJiJtz2fn8n2Lfxhez9eMx+T8t2EP2AX47XKL/WBPVvwiMBOufQl+IttErAklXdNreVluBavIS7GXyKfNLNZqexXgJPxzknTcSdoBt6mvqonhRTUuyfMw6MB83Pp81N6tn8AZrYSeC9eW1+Cf3b1Xbb/aiBv9xa8DfZ1eOThVjxcvzWTbS1+HXfjnQbfC4w3sz/28Jp+jPfj6IwKdKSyT8H7LizAv/x4Iu1/Bn+pT8Br/p/B24mz9i/Hm67+OV33hZ3nD4I86mGfgiAIgiAIBhAREQiCIAiCEhOOQBAEQRCUmHAEgiAIgqDEhCMQBEEQBCUmHIEgCIIgKDHhCARBEARBiQlHIAiCIAhKTDgCQRAEQVBiwhEIgiAIghLzN+GNcnYvbBS9AAAAAElFTkSuQmCC\n", "text/plain": [ "