model_new.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The project is intended to create new horoscopes using RNN.\n",
    "First of all we need raw materials!\n",
    "Even if we'll use a pre-trained NLP model, we need as much horoscopes as possible.\n",
    "The best site we found was https://mystarshoroscopes.com/ \n",
    "It contains about 10000 horoscopes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-10T03:59:10.124763Z",
     "start_time": "2019-09-10T03:59:07.557375Z"
    }
   },
   "outputs": [],
   "source": [
    "from fastai.text import * \n",
    "from bs4 import BeautifulSoup\n",
    "import requests\n",
    "import pandas as pd\n",
    "import re\n",
    "import csv\n",
    "from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from scipy import stats\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "error_msg = \"The page you are looking for no longer exists. Perhaps you can return back to the homepage and see if you can find what you are looking for. Or, you can try finding it by using the search form below.\"\n",
    "\n",
    "signs = [\"aries\", \"taurus\", \"gemini\", \"cancer\", \"leo\", \"virgo\", \"libra\", \n",
    "         \"scorpio\", \"sagittarius\", \"capricorn\", \"aquarius\", \"pisces\"]\n",
    "months = [\"january\", \"february\", \"march\", \"april\", \"may\", \"june\", \"july\", \n",
    "          \"august\", \"september\", \"october\", \"november\", \"december\"]\n",
    "\n",
    "days = [str(day) for day in range(1, 32)]\n",
    "    \n",
    "years = [\"2019\", \"2018\", \"2017\", \"2016\"]\n",
    "\n",
    "voc = \"qwertyuiopasdfghjklzxcvbnmèòàùQWERTYUIOPASDFGHJKLZXCVBNM.,;?! \""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We start by using requests module to request the html of the pages we are looking for.\n",
    "With BeautifulSoup library and some basic knowledge of html it can be found the horoscope's text on each page."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "datas = []\n",
    "\n",
    "for sign, month, day, year in zip(signs, months, days, years):\n",
    "    try: \n",
    "        url = \"https://mystarshoroscopes.com/\" + sign + \"-daily-horoscope-for-\" \\\n",
    "                + month + \"-\" + day + \"-\" + year + \"/\"\n",
    "        requests.get(url, verify=True)\n",
    "        html = requests.get(url).text\n",
    "        soup = BeautifulSoup(html, 'html5lib').find(class_ = \"entry-content\")\n",
    "        if soup and soup.text != error_msg:\n",
    "            datas.append((soup.text, sign, day + \" \" + month + \" \" + year))            \n",
    "    except:\n",
    "        pass"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We store the datas using pandas' DataFrame and we finally convert it into csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.DataFrame(datas, columns = ['text', 'sign ', 'date']) \n",
    "export_csv = df.to_csv (r'C:\\Users\\Giuliano\\Documents\\GitHub\\horoscope\\horoscopes.csv', index = None, header=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Cleaning the text ..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hor_csv = open(\"horoscopes.csv\", encoding=\"utf8\")\n",
    "hor = pd.read_csv(hor_csv)\n",
    "hor = hor.drop(columns = \"date\")\n",
    "index = [i for i in range(0, len(hor))]\n",
    "\n",
    "for i in index:\n",
    "    text = hor[\"text\"][i].strip()\n",
    "    text_new = \"\"\n",
    "    for char in text:\n",
    "        if char in voc:\n",
    "            text_new = text_new + char\n",
    "    hor.xs(i)['text'] = text_new\n",
    "\n",
    "export_csv = hor.to_csv (r'C:\\Users\\Giuliano\\Documents\\GitHub\\horoscope\\text_and_sign.csv', index = None, header=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hor_csv = open(\"horoscopes.csv\", encoding=\"utf8\")\n",
    "hor = pd.read_csv(hor_csv)\n",
    "hor = hor.drop(columns = \"date\")\n",
    "hor = hor.drop(columns = \"sign\")\n",
    "index = [i for i in range(0, len(hor))]\n",
    "\n",
    "for i in index:\n",
    "    text = hor[\"text\"][i].strip()\n",
    "    text_new = \"\"\n",
    "    for char in text:\n",
    "        if char in voc:\n",
    "            text_new = text_new + char\n",
    "    hor.xs(i)['text'] = text_new\n",
    "\n",
    "export_csv = hor.to_csv (r'C:\\Users\\Giuliano\\Documents\\GitHub\\horoscope\\textonly.csv', index = None, header=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And we are ready to try the fastai library!\n",
    "The text module of the fastai library contains a pre-trained model for text-generation.\n",
    "We can try it without fine-tuned the model and it produces well sentences, but not horoscopes... \n",
    "So, we need to train the model with our datas.."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-10T03:59:52.395122Z",
     "start_time": "2019-09-10T03:59:47.406824Z"
    }
   },
   "outputs": [],
   "source": [
    "data_lm = TextLMDataBunch.from_csv('.', 'textonly.csv')\n",
    "data_lm.save('data_lm_export.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-10T03:59:59.121113Z",
     "start_time": "2019-09-10T03:59:58.330769Z"
    }
   },
   "outputs": [
    {
     "ename": "RuntimeError",
     "evalue": "Error(s) in loading state_dict for AWD_LSTM:\n\tsize mismatch for encoder.weight: copying a param with shape torch.Size([7304, 400]) from checkpoint, the shape in current model is torch.Size([7320, 400]).\n\tsize mismatch for encoder_dp.emb.weight: copying a param with shape torch.Size([7304, 400]) from checkpoint, the shape in current model is torch.Size([7320, 400]).",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-98-abbb50ec7bbc>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[0mdata_lm\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mload_data\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'data_lm_export.pkl'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[0mlearn2\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mlanguage_model_learner\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdata_lm\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mAWD_LSTM\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdrop_mult\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m0.5\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 5\u001b[1;33m \u001b[0mlearn2\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mload_encoder\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'language_model'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      6\u001b[0m \u001b[0mlearn2\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0munfreeze\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\lib\\site-packages\\fastai\\text\\learner.py\u001b[0m in \u001b[0;36mload_encoder\u001b[1;34m(self, name, device)\u001b[0m\n\u001b[0;32m     67\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mdevice\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mdevice\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdata\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdevice\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     68\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mhasattr\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mencoder\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'module'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mencoder\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mencoder\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmodule\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 69\u001b[1;33m         \u001b[0mencoder\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mload_state_dict\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mload\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m/\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmodel_dir\u001b[0m\u001b[1;33m/\u001b[0m\u001b[1;34mf'{name}.pth'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmap_location\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mdevice\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     70\u001b[0m         \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfreeze\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     71\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36mload_state_dict\u001b[1;34m(self, state_dict, strict)\u001b[0m\n\u001b[0;32m    775\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0merror_msgs\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m>\u001b[0m \u001b[1;36m0\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    776\u001b[0m             raise RuntimeError('Error(s) in loading state_dict for {}:\\n\\t{}'.format(\n\u001b[1;32m--> 777\u001b[1;33m                                self.__class__.__name__, \"\\n\\t\".join(error_msgs)))\n\u001b[0m\u001b[0;32m    778\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0m_IncompatibleKeys\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmissing_keys\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0munexpected_keys\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    779\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mRuntimeError\u001b[0m: Error(s) in loading state_dict for AWD_LSTM:\n\tsize mismatch for encoder.weight: copying a param with shape torch.Size([7304, 400]) from checkpoint, the shape in current model is torch.Size([7320, 400]).\n\tsize mismatch for encoder_dp.emb.weight: copying a param with shape torch.Size([7304, 400]) from checkpoint, the shape in current model is torch.Size([7320, 400])."
     ]
    }
   ],
   "source": [
    "# Load previous iteration of the model\n",
    "path = 'C:/Users/Giuliano/Documents/GitHub/horoscope'\n",
    "data_lm = load_data(path, 'data_lm_export.pkl')                                 \n",
    "learn2 = language_model_learner(data_lm, AWD_LSTM, drop_mult=0.5)\n",
    "learn2.load_encoder('language_model')\n",
    "learn2.unfreeze()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-10T15:06:41.430833Z",
     "start_time": "2019-09-10T15:06:40.561734Z"
    }
   },
   "outputs": [],
   "source": [
    "# See if it's working\n",
    "learn2.predict(\"Fabio is \", n_words=100)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-10T11:03:33.867516Z",
     "start_time": "2019-09-10T04:00:38.093827Z"
    }
   },
   "outputs": [],
   "source": [
    "learn2.fit(1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-09-10T11:03:34.119240Z",
     "start_time": "2019-09-10T11:03:33.870675Z"
    }
   },
   "outputs": [],
   "source": [
    "data_lm.save('data_lm_export.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Sometimes you are lucky with horoscopes and sometimes not.\n",
    "Now, we want to predict if the horoscope brings you bad or good luck.\n",
    "Even if it is easy to implement a quick-model from scratch, we use vaderSentiment library. \n",
    "This library uses a bag of words approach (a lookup table of positive and negative words) with some simple heuristics.\n",
    "The advantage of this approach is that sentences containing negated positive words (e.g. “not happy”, “not good”) will still receive a negative sentence sentiment (thanks to the heuristics to flip the sentiment of the word following a negation).\n",
    "Instead, a quick-model will just take the average of the sentiments of the words and would miss details like this."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "analyser = SentimentIntensityAnalyzer()\n",
    "\n",
    "def sentiment_analyzer_scores(sentence):\n",
    "    score = analyser.polarity_scores(sentence)\n",
    "    if score['compound'] > 0.3:\n",
    "        print(\"That's great!\")\n",
    "    elif score['compound'] < - 0.3:\n",
    "        print(\"That's a shame\")\n",
    "    else:\n",
    "        print(\"Could be better...\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "You will n’t do do n’t do do do do do n’t n’t do know % n’t to do feel do buttons do schemes so scenes la curious scenes scenes electronic to scenes crushing fronts scenes zone ’ll feel scenes scenes % to attention to system scenarios big scenes scenes scenes hustle to to to feel to been to much hustle to statements to cabinets to map takes wo to big attention to to to to multi bubble scenes lifetime with in to buttons to dedication scenes so reflecting scenes dedication with light to fore takes number par hustle light doors closet \n",
      "\n",
      "Could be better...\n"
     ]
    }
   ],
   "source": [
    "example = learn.predict(\"You will\", n_words=100)\n",
    "\n",
    "print(hor, \"\\n\")\n",
    "\n",
    "sentiment_analyzer_scores(hor)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2e227d9cb70>"
      ]
     },
     "execution_count": 105,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "hor_csv = open(r'C:\\Users\\Giuliano\\Documents\\GitHub\\horoscope\\textonly.csv', encoding=\"utf8\")\n",
    "\n",
    "hor = pd.read_csv(hor_csv)\n",
    "sent_arr = []\n",
    "\n",
    "for i in index:\n",
    "    score = analyser.polarity_scores(learn.predict(hor.iloc[i]['text']))\n",
    "    sent_arr.append(score['compound'])\n",
    "\n",
    "sns.distplot(sent_arr)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Using seaborn (a data visualization library) we plot an histogram of the horoscopes' luck.\n",
    "As can be seen from the graph above, most of them are positive.\n",
    "Now, we expect that our \"new\" horoscopes are positive, as well... And fortunately, that's true :)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.4877, -0.2263, 0.9338, 0.7003, 0.8074, 0.7589, 0.7893, -0.8126, 0.7579, -0.25, 0.6486, 0.9521, 0.7506, -0.3446, 0.8671, 0.128, 0.4019, 0.7498, -0.1027, 0.784, 0.2263, 0.7399, 0.34, 0.128, 0.1027, 0.128, 0.8042, 0.5994, 0.7096, 0.9118, 0.7351, 0.34, 0.3818, -0.4588, 0.0, 0.5719, -0.5106, 0.8074, -0.4939, 0.4939, 0.7635, 0.5267, 0.9634, -0.3182, 0.5254, 0.891, 0.8151, 0.3612, 0.25, 0.6705, 0.5574, 0.5859, 0.5859, 0.4939, 0.7402, 0.4019, 0.765, 0.6705, 0.9464, -0.5574, 0.2648, 0.7269, 0.6124, -0.5715, 0.8225, 0.8934, 0.5106, 0.2212, 0.9403, 0.7399, 0.6344, 0.836, 0.6249, 0.1372, 0.8323, 0.3744, 0.6249, 0.2732, 0.743, 0.8934, 0.0516, 0.2177, 0.4404, 0.3182, 0.128, 0.7076, 0.7003, 0.8519, 0.8316, 0.5777, 0.8348, 0.6457, 0.603, -0.2263, 0.7717, 0.3535, 0.0772, 0.25, 0.9203, 0.8402]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2e228023b00>"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD4CAYAAAD8Zh1EAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nO3dd3xc5Z3v8c9PvXfJclFztzG4N8iCTSAxsEvZkFCTsJeEJbukbHLvLkn2koRkS7LZTdmQAllCEgKEEAKGNesQSggYV9xtbMtNzbZk9WKV0Tz3D8lcYWRrLI10pnzfr5denplzdM7veEbf15nnPM9zzDmHiIiEvxivCxARkeBQoIuIRAgFuohIhFCgi4hECAW6iEiEiPNqx3l5ea60tNSr3YuIhKUtW7acdM7lD7bMs0AvLS1l8+bNXu1eRCQsmdnRsy1Tk4uISIRQoIuIRAgFuohIhFCgi4hECAW6iEiEUKCLiEQIBbqISIRQoIuIRAgFuohIhPBspKiIhLfHNlSM+j5uXVo86vuIJDpDFxGJEAp0EZEIMWSgm9nDZlZrZruGWG+xmfWa2Y3BK09ERAIVyBn6I8Cqc61gZrHAN4G1QahJRESGYchAd869BjQMsdqngd8CtcEoSkREzt+I29DNbCJwA/DjANa9y8w2m9nmurq6ke5aREQGCMZF0e8C/+Cc6x1qRefcg865Rc65Rfn5g95wQ0REhikY/dAXAU+YGUAecLWZ+ZxzzwRh2yIiEqARB7pzruz0YzN7BHheYS4iMvaGDHQzexxYAeSZWRXwFSAewDk3ZLu5iIiMjSED3Tl3S6Abc87dMaJqRERk2DRSVEQkQijQRUQihAJdRCRCKNBFRCKEAl1EJEIo0EVEIoQCXUQkQijQRUQihAJdRCRCKNBFRCKEAl1EJEIo0EVEIoQCXUQkQijQRUQihAJdRCRCKNBFRCKEAl1EJEIo0EVEIoQCXUQkQgwZ6Gb2sJnVmtmusyy/zcx29P+sM7O5wS9TRESGEsgZ+iPAqnMsPwxc5py7CPg68GAQ6hIRkfMUN9QKzrnXzKz0HMvXDXi6Hpg08rJEROR8BbsN/U7ghbMtNLO7zGyzmW2uq6sL8q5FRKLbkGfogTKzlfQF+vvOto5z7kH6m2QWLVrkgrVvERl9Pb1+Nh9p5NDJNqobT9HW5WP6uHRizLwuTfoFJdDN7CLgp8BVzrn6YGxTRELH3mMtfP7J7ew91gKAGTgHuakJXD6zgPnF2R5XKBCEQDezYuBp4KPOuf0jL0lEQsnDrx/mX17YS2ZyAt+7eR6LS3PITUvgvmd38/qBk/xmSxWdPj/LJ+d6XWrUGzLQzexxYAWQZ2ZVwFeAeADn3I+B+4Bc4IfW99XL55xbNFoFi8jY+fWmCu5/fg9XzBrHt268iJzUhHeWzZ2UxZwJmTy2sYLnttcQF2MsLs3xsFoJpJfLLUMs/wTwiaBVJCJB8diGihH9/oETrfz8zSNMK0jjsun5/M+u4+9ZJzbGuGVxEY9uOMozW6vJSU1gSn7aiPYrw6eRoiLyHifbunhsYwUF6UncsqSY2JizX/iMi43h1iUlZKXEs3pbDT6/fwwrlYEU6CLyLn7n+N3WaszgY8tLSIqPHfJ3EuJiuHbuBOraunjjwMkxqFIGo0AXkXfZfKSRwyfbuXrOeLJSEob+hX4zCjOYPT6Dl/fV0tjRPYoVytko0EXkHS2nenhh1zEm56WysOT8uyL++UXjAVi7+73t7TL6FOgi8o41u47R63fcMH8iNowBQ1kpCSwty2VXdTNNOksfcwp0EQGguukUO6qaed+0PHLTEoe9nYun9PVHX3dQYwzHmgJdRAD4/e7jJMfHcum0/BFtJyslgTkTM9l0pIHOnt4gVSeBUKCLCAfr2jhQ28aKGfkB9WoZyvum5tHl87P5aGMQqpNAKdBFopxzjt/vPk5mcjzLgjR8f1J2CqW5qaw7eBK/0zx8Y0WBLhLlymvbqGw8xeUzCoiPDV4kLJ+SS1NHD4fq2oO2TTk3BbpIlPvj/joykuKYX5wV1O3OLEwnMS6GbZVNQd2unJ0CXSSKVTZ0cOhkO++bmkdcEM/OAeJjY5gzMZPdNc10+zQdwFhQoItEsT/uryM5PnbUZkmcV5RFl8/P28dbRmX78m4KdJEoVdvSyZ5jLSybnEtiEHq2DKYsL5WMpDg1u4wRBbpIlHq9/CRxMcbyKaN3Y4oYM+YWZbH/RCvtXb5R24/0UaCLRKH2Lh/bKpuYX5xNWmLQbi08qHlFWfgd7KppHtX9iAJdJCptOtKAz+/eGaY/mgozkshNTXjnfqQyehToIlGm1+9Yf6ieqQVpjMtIGvX9mRkzC9M5WNdOl6YCGFUKdJEos6u6mZZOH5eMwdn5abPGZ9DrdxyobRuzfUYjBbpIlFl38CR5aQlMG5c+ZvssyU0lOT5W3RdH2ZCBbmYPm1mtme06y3Izs++bWbmZ7TCzBcEvU0SCoaqxg8rGUyybnEvMMOY7H67YGGNGYTpvH2/V3C6jKJAz9EeAVedYfhUwrf/nLuBHIy9LREbD+kMNJMTGsKD4/O9GNFIzC9Pp6O6lsqFjzPcdLYYMdOfca0DDOVa5DviF67MeyDKz8cEqUESCo6PLx46qJuYVZwVlitzzNX1cOjGGeruMomC0oU8EKgc8r+p/7T3M7C4z22xmm+vq6oKwaxEJ1Oajjfj8LmhT5J6vpPhYyvJS2Xu81ZP9R4NgBPpgDXGDNpI55x50zi1yzi3Kzx/ZXVFEJHB+59hwuJ7S3FQKx6Cr4tlMH5dOXWsXzad6PKshkgUj0KuAogHPJwE1QdiuiATJ/hOtNHb0sGzy6EzCFagp+WkAHFT3xVERjEBfDXysv7fLMqDZOXcsCNsVkSBZf6ie9KQ4LpiQ6WkdhZlJpCbEUl6nQB8NQ07iYGaPAyuAPDOrAr4CxAM4534MrAGuBsqBDuCvRqtYETl/9W1dHDjRxsqZBcTGjF1XxcHEmDGlII3y2jacc9gYdp2MBkMGunPuliGWO+Bvg1aRiATVhsMNmMGSUZrz/HxNK0hjR1Uzx1s6GZ+Z7HU5EUUjRUUiWLfPz5ajjcyekElGcrzX5QAwtaBvhGq52tGDToEuEsF2VDVxqqfX84uhA2Umx5OflqhAHwUKdJEI5ZzjzUP1jMtIpCw31ety3mVqQRpH6tvx9epeo8GkQBeJUEfqOzjW3MnFk/NC7uLj1II0enodFZoGIKgU6CIRat3BkyTHxzK3KMvrUt6jNDcVAw6fbPe6lIiiQBeJQNVNp9hT08Li0hwS4kLvzzw5IZbCzCQFepCF3jstIiP2izePYEZIXQw9U1leKhUNHWpHDyIFukiEOdXdyxMbK5k9PoOslASvyzmrsrxUfH5HVeMpr0uJGAp0kQjzu63VNJ/q4eIpeV6Xck6l/T1vDter2SVYFOgiEcQ5xyPrDnPBhAxKclO8LuecUhPjGJeRqHb0IFKgi0SQdQfr2X+ijTsuLg25roqDKctLpaK+g16/bksXDAp0kQjyszeOkJuawF/MneB1KQEpy0uju9dPdZPa0YNBgS4SIY7Wt/PS2ye4dWmxJ7eYG47S/mYhNbsEhwJdJEL87I0jxMUYty8r8bqUgKUnxZOXlsgRBXpQKNBFIkBDezdPbKrgunkTGefhLeaGoyQ3hYqGDvxO7egjpUAXiQA/X3eEzh4/d1822etSzltJTgqneno52dbldSlhT4EuEuY6un38/M0jXDFr3DtzjYeT4v529Ip6TdQ1Ugp0kTD3602VNHX08KkV4Xd2DpCXlkhyfCxHNfPiiCnQRcJYt8/PQ68dYnFpNgtLQnfelnOJMetrR9cZ+ogp0EXC2G+2VFLT3MmnL5/mdSkjUpyTQl1bFx1dPq9LCWsBBbqZrTKzfWZWbmb3DrK82MxeMbOtZrbDzK4OfqkiMlC3z88DL5czvziLP5sW2vO2DKWkf14X3fBiZIYMdDOLBR4ArgJmA7eY2ewzVvtH4Enn3HzgZuCHwS5URN7t9Nn5566YHhbD/M9lYlYyMYba0UcokDP0JUC5c+6Qc64beAK47ox1HJDR/zgTqAleiSJypm6fnx++cpD5xVlcGuZn5wAJcTFMyErmqNrRRySQQJ8IVA54XtX/2kBfBW43sypgDfDpwTZkZneZ2WYz21xXVzeMckUE4PGNFVQ3nYqIs/PTSnJSqGrURF0jEUigD/ZpOfN//BbgEefcJOBq4Jdm9p5tO+cedM4tcs4tys/PP/9qRYTWzh6+/9IBlk3OiYiz89OKc/tueFGjibqGLZBArwKKBjyfxHubVO4EngRwzr0JJAGR80kTCSEP/ekw9e3dfPGqWRFzdg59Z+igC6MjEUigbwKmmVmZmSXQd9Fz9RnrVADvBzCzWfQFutpURIKstrWTn/7pENdcNJ65RVlelxNUGcnxZKXE68LoCAwZ6M45H3APsBbYS19vlt1mdr+ZXdu/2heAT5rZduBx4A7nNNOOSLB958UDdPv8/J8PzPC6lFFRkpNCRX07io/hiQtkJefcGvoudg587b4Bj/cAlwS3NBEZaGdVM09squCvLi6jNC/V63JGRXFuKturmmk61UN2CN/gOlRppKhIGPD7HV9ZvYvc1EQ+d2V4jwo9l9Pt6Oq+ODwKdJEw8PTWat6qaOLeq2aSkRTvdTmjZlxGEglxMVQ06IYXw6FAFwlxTR3d/OsLe1lQnMVfzj9zCEhkiY0xirM1UddwKdBFQtw3/nsvTR09fP36OcTERE43xbMpzk3hWHMnXT29XpcSdhToIiHsj/vreGpLFXdfNoULJmR6Xc6YKM5JwQGVjRpgdL4U6CIhqq3Lx5ee3smU/FTuuXyq1+WMmeKcFAw4qnb08xZQt0URGXv/9N97qWk+xVN3LycpPtbrcsZMUnwsBRmJVGqA0XnTGbpICFq7+ziPb6zg7sumhO2diEaiOCeFioYO/Jqo67wo0EVCzImWTu797Q4unJjJ310x3etyPFGck0Jnj59DJ9u8LiWsKNBFQkiv3/H5J7fR2ePnuzfPIyEuOv9Ei3P6RsK+dbTJ40rCS3R+WkRC1Pf+sJ83yuv56rWzmZKf5nU5nslLSyA5PpYtRxu9LiWsKNBFQsQr+2r5/svlfHjhJG5aXOx1OZ4yM4pzUnirQoF+PhToIiGgor6Dv/v1NmaNz+Dr18/xupyQUJSTwoHaNppP9XhdSthQoIt4rLmjhzse2QjAj25bEFVdFM+luH+irm2VakcPlAJdxEPdPj9//ehmqhpO8eBHF0XstLjDUZSdTIzBW2pHD5gGFol4xDnHF5/eyfpDDXz3pnksKYu+/ubnkhgfy4zCDLWjnwedoYt45Acvl/Pbt6r4uyumc32Ez6I4XAuKs9hW0aQBRgFSoIt44Nlt1fz7i/v5y/kT+cz7o2eelvO1oDib1i4fB2o1wCgQanIRGcRjGypGbdtHTrbzX28cpjQ3lXlFWTy+sXLU9hXuFpZkA/BWRSMzCtM9rib0BXSGbmarzGyfmZWb2b1nWecjZrbHzHab2WPBLVMkMtS3dfHohqNkp8Rz+7Ji4mL1JflcSnJTyElN0ACjAA15hm5mscADwJVAFbDJzFb33xj69DrTgC8ClzjnGs2sYLQKFglXHV0+Hll3BICPLy8lJUFfkIdiZiwoztKF0QAFcnqwBCh3zh1yznUDTwDXnbHOJ4EHnHONAM652uCWKRLefL1+Ht1QQdOpHj66rITctESvSwob84uzOVTXTlNHt9elhLxAAn0iMLCRr6r/tYGmA9PN7A0zW29mq4JVoEi4c87x9NZqjtS3c+OCSZTkqq/5+Tjdjr61QgOMhhJIoA92E8Mz+xDFAdOAFcAtwE/NLOs9GzK7y8w2m9nmurq6861VJCy9vK+WbZVNXDFrHHOL3vNnIUO4aFImsTGmdvQABBLoVUDRgOeTgJpB1nnWOdfjnDsM7KMv4N/FOfegc26Rc25Rfn7+cGsWCRtbKxp5aW8tC4qzWDlDn/nhSEmIY9b4dLWjByCQQN8ETDOzMjNLAG4GVp+xzjPASgAzy6OvCeZQMAsVCTeHT7bz9NZqyvJSuX7+RMwG+7IrgVhQnM32yiZ6NcDonIYMdOecD7gHWAvsBZ50zu02s/vN7Nr+1dYC9Wa2B3gF+D/OufrRKlok1NW3dfHo+qNkpyRw+9IS4mLUPXEkFpZk097dy9vHW7wuJaQF1G/KObcGWHPGa/cNeOyAz/f/iES1Ll8vv1x/FICPLy8hOUGzJ47UguK+C6NbjjZywYRMj6sJXTptEAkiv3M8taWKutYubllSrO6JQTIpO5nCjCQ2H1E7+rko0EWC6NV9deyuaeGqOYVMLYjeW8gFm5mxsDSbzUcavC4lpCnQRYJk77EW/rD3BPOKsrhkap7X5UScxSXZ1DR3Ut10yutSQpYCXSQIals7eXJzJROzkrlBPVpGxaLSvvnidZZ+dgp0kRHq8vXy6PoK4mJjuG1pMfGacGtUzCxMJzUhVgOMzkGfPJERWr2thvq2Lm5ZXERWSoLX5USsuNgY5hdns0kXRs9KgS4yAlsrGtla2cTKmQVMztdF0NG2sCSbfcdbaOns8bqUkKRAFxmmk61dPLuthtLcVFbO0IzRY2FxaQ5+p4m6zkaBLjIMvl4/T2yqIDbGuGlxEbExugg6FuYVZxFjujB6Ngp0kWH4n93HqWnu5MaFk8hMjve6nKiRlhjHBRMy2XhYgT4YBbrIeXr7WAvrDtazfEous8ZneF1O1FlSlsPWyia6fL1elxJyFOgi56G9y8dv36pifGYSV11Q6HU5UWlpWQ7dPj/bK5u9LiXkKNBFzsPq7TV09vj58MIi3eDZI4v7BxhtPKwJXc+kT6RIgHZVN7OzupmVMwsozEzyupyolZ2awMzCdDaoHf09FOgiAWjv8vHs9homZCVx2XTdechrS8py2HK0kZ5ev9elhBQFukgAnttRQ2d3Lx9aMEldFEPA0rJcOrp72VWtdvSBFOgiQ9hd08yOqmZWzsxnfGay1+UIsLis74YX6r74bgp0kXPo6PLxzLYaJmQmcdl0jQYNFQXpSUzOT1Wgn0GBLnIOz+2o4VS3jw8tVFNLqFlalsPGIw26cfQACnSRs9hT08z2qr5eLWpqCT3LJufS2ulTO/oAAQW6ma0ys31mVm5m955jvRvNzJnZouCVKDL2Tje1jM9MYoWaWkLSxVP67gr1xsGTHlcSOoYMdDOLBR4ArgJmA7eY2exB1ksHPgNsCHaRImPt+Z3H6Oj2qVdLCMtPT2TGuHTWlWuA0WmBnKEvAcqdc4ecc93AE8B1g6z3deBbQGcQ6xMZc7/ffZxtlU2snFHAhCw1tYSyS6bmselIA509mtcFAgv0iUDlgOdV/a+9w8zmA0XOuefPtSEzu8vMNpvZ5rq6uvMuVmS0NXV08+VndjE+M4nLZmgAUai7ZGouXT4/b1XoLkYQWKAP9n3zncvKZhYDfAf4wlAbcs496Jxb5JxblJ+vPxYJPV97bg+N7d18aMEk4mLUZyDULSnLITbG1OzSL5BPbBVQNOD5JKBmwPN0YA7wqpkdAZYBq3VhVMLNi3tO8Lut1fzNyqlqagkT6UnxzJ2UqQuj/QIJ9E3ANDMrM7ME4GZg9emFzrlm51yec67UOVcKrAeudc5tHpWKRUZBU0c3X/rdTmYWpnPPyqlelyPn4ZKpeeyoaqZV9xkdOtCdcz7gHmAtsBd40jm328zuN7NrR7tAkbFw/3N7aGjv5tsfnktCnJpawsnFU/Lo9TvWH9Ko0bhAVnLOrQHWnPHafWdZd8XIyxIZO3/Yc4Knt1bzmcunMmdiptflyHlaUJJFSkIsr+2v48rZ47wux1M6FZGoVt/WxRdPN7VcPs3rcmQYEuNiuXhKLq/ur8W56J4GQIEuUcs5x71P76S5o4fv3DRPTS1h7LIZBVQ2nOLQyXavS/GUPsEStX69qZIX95zg71fN0M2ew9yK/puOvPJ2rceVeEuBLlHp8Ml2vvbcHi6Zmsv/uqTM63JkhIpyUphakMYf90f3gEUFukSdnl4/n/v1NhLiYvj2h+cSo7laIsKK6flsONRAR7fP61I8o0CXqPOfLx1ge2UT/3zDhZoWN4KsmFFAd6+fNw9G76hRBbpElc1HGvjBK+X85YKJXHPReK/LkSBaXJZNSkIsr+6L3mYXBbpEjZNtXdzz2FYmZifztWsv8LocCbK+7ot5vPx29HZfVKBLVPD1+vn0Y1tp7OjmR7ctJD0p3uuSZBR84IJxVDedYld1i9eleEKBLlHh27/fz5uH6vmnGy7UaNAIdsWsccQYrN193OtSPKFAl4j31JYqfvzHg9y2tJgbF07yuhwZRTmpCSwty+V/FOgikWfdwZN88ekdXDI1l6+q3TwqfPCCcZTXtnGwrs3rUsacAl0i1v4Trdz9yy2U5Kbyw9sWEh+rj3s0+MAFhUB0NrvoEy4R6WBdG7c+tIHE+Fh+dsdiMpN1ETRaTMhKZu6kTNbuUqCLhL2j9e3c+tB6wPH4J5dSlJPidUkyxj5wQSHbq5qpaTrldSljSoEuEWVPTQsf/vGbdPv8PPqJpUwtSPe6JPHANRf2DRp7bnvNEGtGFgW6RIzXD5zkIz95k9gY44m7ljOzUDMoRqvSvFTmF2fxu63VXpcyphToEvb8fsdDrx3ijp9tZFJ2Mk//zcXMKNSZebS7Yf5E3j7eyt5j0TPISIEuYe1kWxd3/nwT/7RmL1fMGseTdy/XhFsC9DW7xMYYz2yLnrP0gO4pKpHjsQ0VXpcwYrcuLcbX6+dXGyr499/vo7PHz9evu4Dbl5VgpqlwpU9uWiKXTc/n2a01/MMHZ0bFNMkBnaGb2Soz22dm5WZ27yDLP29me8xsh5m9ZGYlwS9VBHr9jme2VnPN91/nK6t3c+GkTNZ89n18dHmpwlze4/r5Ezne0sn6w9Expe6QZ+hmFgs8AFwJVAGbzGy1c27PgNW2Aouccx1m9ingW8BNo1GwRKfalk52VDfzVkUjTR09TCtI44FbF3D1hYUKcjmrK2eNIy0xjqe2VHHxlDyvyxl1gTS5LAHKnXOHAMzsCeA64J1Ad869MmD99cDtwSxSoktnTy91rV3UtnZytL6DwyfbqW/vxoCy/FS+feNcLp9ZEBVfoWVkkhNiuW7eBJ7aUsX/vWY22akJXpc0qgIJ9IlA5YDnVcDSc6x/J/DCYAvM7C7gLoDi4uIAS5RQ1dPrp7XTR1tnD61dPlo7fbR3+ejpdfT6/fj8Dp/f4fc7HODc6X/5/88ddPf66fb56eju20aXz//OPpLjYynJTWH5lFzmTMwkIymeK2aP8+qQJQzdvqyEX22o4KktVXzy0slelzOqAgn0wU6DBp093sxuBxYBlw223Dn3IPAgwKJFi6JzBvow1dblo7Khg8qGDmqaT1Hb2kVTR8+g68aaERtrxMX0/cTGGGaGAX2tI4ZZ3wcrxoz4WCMhLobxmclMGxdHRlI8+WmJ5KcnkpuWQIyaVGQEZo3PYHFpNr/acJQ731cW0d/sAgn0KqBowPNJwHuGX5nZFcCXgcucc13BKU+84neOyoYO9hxroby2jWPNnQDEGIzLSKI4J4WFJYlkJSeQnhRHWmIcaUlxpCbEERvBfzASnm5fVsJnn9jG6+UnuXR6vtfljJpAAn0TMM3MyoBq4Gbg1oErmNl84CfAKudcbdCrlDFzoqWTLUcb2VHVREunj1gzinNT+MDscZTlpTIhK1mzFkrYWTWnkNzUBH65/mh0B7pzzmdm9wBrgVjgYefcbjO7H9jsnFsN/BuQBvymv8dBhXPu2lGsW4Ko1+/YVd3MuoMnqWw8RawZ0wvTWTUxk5mF6STFx3pdosiIJMbFcvOSIn706kEOn2ynLC/V65JGRUADi5xza4A1Z7x234DHVwS5LhkDPb1+Nh5u4E8H6mjp9JGbmsDVcwqZV5xNWqLGnElkuePiMn76p8P86NVyvnXjXK/LGRX6q41CPb1+Nh9p4NX9dbR2+ijLS+X6+ROZPi5dFyAlYuWnJ3LLkmIeXX+Uz14xnYlZkTdFhAI9ivT0+tlwuJ5X99XRfKqH0txUblpcxOS8NK9LExkTd106mV9tOMpP/niQ+6+b43U5QadAjxKv7KvlG8/v4WBdO8U5Kdy4cBKT81I1ylKiyoSsZD60YBJPbKrknpVTKchI8rqkoFKgR7jy2ja+8d97eHVfHWV5qXx0WQkzC9PDOsgjYYIx8c6nVkzhN1uq+O5LB/jnGy70upygUv+zCNXU0c1XV+9m1XdfY8vRRv7xmlms/dylzBqfEdZhLjJSJbl9JzZPbKyIuLnSdYYeYXy9fh7bWMF/vLifllM93LykmC9cOZ3ctESvSxMJGZ+7YhrPbKvm/uf28Ngnl0bMSY4CPYK8tr+Orz+/hwO1bSyfnMt9fzGbWeN1GzaRM2WlJPD5K6dz37O7Wbv7BKvmFHpdUlAo0CPAgROt/OsLb/PS27UU56Twk48u5AOzx0XMWYfIaLi1vwvj/c/tZvmUXDKT470uacTUhh7Gals7+eLTO/ngd19j4+EG7r1qJi9+/lI+eIHmCBcZSlxsDN+6cS4nWrv42urdXpcTFDpDD0Md3T4eeu0wP3ntIN0+Px9bXspn3j+NnAif61kk2OYVZXHPyql876UDvH/WOK65aLzXJY2IAj2MtHX5+OWbR3noT4doaO/mqjmF/P2qmRE7L4XIWLjn8qm8uq+WLz+zk7lFmUzKTvG6pGFToIeBls4efrHuCD99/TBNHT1cNj2fz14xjQXF2V6XJhL24mNj+M5N87jugTe485HN/OZTy8lICs/2dAV6CKtq7OBXGyr41fqjtHT6eP/MAj79/mnMK8ryujSRiDI5P40f3baQO362kXse28rDH19EXBhOE61ADzHOOd4or+fnbx7hpb0nAPjA7EL+duVULpyU6W1xIhHsfdPy+Mb1c7j36Z38799s59sfnht2oa5ADxGVDR2s3l7D029VcbCunZzUBD61Ygq3Li2JyFnhRELRzUuKqW/v5t/W7qO9u5f/vGV+WN0PQIHuoZNtXazZeYxnt9Ww5W1Sa3sAAAefSURBVGgjAItKsvmPj0zl6gvHh9UHSSRS/O3KqaQnxXHfs7u542cb+c9bFpCfHh4jrRXoY8g5x+6aFl55u5aX99WyrbIJ52BmYTp/v2oGzg/ZqQl09vh5+q1qr8sViVofW15KRlI8//DbHVz1vT/xnZvm8mfTQv/WdQr0UeSco6rxFBsON7DhUD2vHajjREvf/bMvmpTJZy6fxlUXFjKzsG94vmYRFAkd18+fyKzxGdzz2Ft87OGNfGRhEV/44HQK0kN3yl0FehB19vTy9vFWdlY18VZFExsO1VPT3AlAVko8yyfnsnJmAStm5If0h0JE+swoTGf1Pe/j33+/j0fWHeH5HTV84s8mc/uykpBshlGgD4NzjtrWLg6caONAbSv7jreys7qZfcdb8fkdAHlpiSwty+Gvy3JYOjmH6QXpxMRoOL5IuElOiOUf/3w2ty0r4V/W7OV7Lx3gR68e5JqLxnPt3AlcPDWXxLjQuN4VUKCb2Srge0As8FPn3L+esTwR+AWwEKgHbnLOHQluqWPHOUd7dy91rV1UN56iqrGDysYOqhpPUdHQQXltG62dvnfWz0qJ58KJmdx16WQumpTJhZOymJCZpPlURCJIWV4qD35sEQfr2vjFuiM8/VY1v9taTXpiHMum5LK4NJsFxdlML0z3bGDSkIFuZrHAA8CVQBWwycxWO+f2DFjtTqDROTfVzG4GvgncNBoFt3T2cKK5EzMwMwyIMcOMd25w7PM7un1+enr9dPf66fH1/9vrp62rl9bOHto6fbR2+mjt7KG100dLp4+mjm7q2rqobeniVE/vu/YbG2NMyEpiUlYK18+byLRxaUwtSGNaQTp5aQkKb5EoMSU/ja9dN4cvXTOLdeX1rN19nPWH6nlxz4l31inMSGJCVhLjMvp+CjISyU9LJD0pnvSkOIpzUijKCf4UA4GcoS8Byp1zhwDM7AngOmBgoF8HfLX/8VPAD8zMnHMuiLUCfXN+3/PY1qBsK8YgLTHunf/k7JQE5hVlkZ+WSH56InlpiUzISqYoJ5nCjKSwG2QgIqMnMS6WlTMLWDmzAIDalk52VDWzv7aV8to2jjd3sv9EK6+Xn3zXN3qAv75sMl+8albQawok0CcClQOeVwFLz7aOc85nZs1ALnBy4EpmdhdwV//TNjPbN5yiQ0QeZxxfhInk44vkY4MIOr7b3vtSRBzbl74JXxp8USDHV3K2BYEE+mBtCWeeeQeyDs65B4EHA9hnyDOzzc65RV7XMVoi+fgi+dggso8vko8NRn58gbQhVAFFA55PAmrOto6ZxQGZQMNwixIRkfMXSKBvAqaZWZmZJQA3A6vPWGc18PH+xzcCL49G+7mIiJzdkE0u/W3i9wBr6eu2+LBzbreZ3Q9sds6tBv4L+KWZldN3Zn7zaBYdIiKi6egcIvn4IvnYILKPL5KPDUZ4fKYTaRGRyKB+eCIiEUKBLiISIRToATKzD5vZbjPzm9lZuxWZ2Soz22dm5WZ271jWOBJmlmNmL5rZgf5/B71hqZn1mtm2/p8zL46HlKHeCzNLNLNf9y/fYGalY1/l8ARwbHeYWd2A9+oTXtQ5HGb2sJnVmtmusyw3M/t+/7HvMLMFY13jSARwfCvMrHnAe3dfwBt3zukngB9gFjADeBVYdJZ1YoGDwGQgAdgOzPa69gCP71vAvf2P7wW+eZb12ryuNcDjGfK9AP4G+HH/45uBX3tddxCP7Q7gB17XOszjuxRYAOw6y/KrgRfoG/+yDNjgdc1BPr4VwPPD2bbO0APknNvrnBtqZOs70yQ457qB09MkhIPrgJ/3P/45cL2HtQRDIO/FwGN+Cni/hcekPOH8ORuSc+41zj2O5TrgF67PeiDLzMaPTXUjF8DxDZsCPbgGmyZhoke1nK9xzrljAP3/FpxlvSQz22xm680slEM/kPfiXVNWAKenrAh1gX7OPtTfJPGUmRUNsjxchfPfWaCWm9l2M3vBzC4I9Jc0H/oAZvYHoHCQRV92zj0byCYGeS1k+oWe6/jOYzPFzrkaM5sMvGxmO51zB4NTYVAFbcqKEBRI3c8Bjzvnuszsbvq+iVw+6pWNjXB93wL1FlDinGszs6uBZ4BpgfyiAn0A59wVI9xEINMkeOZcx2dmJ8xsvHPuWP/X19qzbKOm/99DZvYqMJ++9txQcz5TVlSF2ZQVQx6bc65+wNOH6JvSOlKE9N/ZSDnnWgY8XmNmPzSzPOfckJOSqckluAKZJiFUDZy+4ePAe76RmFl2/81MMLM84BLePY1yKInkKSuGPLYz2pSvBfaOYX2jbTXwsf7eLsuA5tPNhZHAzApPX8sxsyX05XT9uX+rn9dXfMPlB7iBvjODLuAEsLb/9QnAmgHrXQ3sp++s9cte130ex5cLvAQc6P83p//1RfTdpQrgYmAnfb0qdgJ3el33EMf0nvcCuB+4tv9xEvAboBzYCEz2uuYgHtu/ALv736tXgJle13wex/Y4cAzo6f+buxO4G7i7f7nRd9Odg/2fw0F7nYXqTwDHd8+A9249cHGg29bQfxGRCKEmFxGRCKFAFxGJEAp0EZEIoUAXEYkQCnQRkQihQBcRiRAKdBGRCPH/AOSJANIT0CCwAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sent_arr = []\n",
    "for _ in range(0, 100):\n",
    "    score = analyser.polarity_scores(learn.predict(\"You will\", n_words=100))\n",
    "    sent_arr.append(score['compound'])\n",
    "\n",
    "sns.distplot(sent_arr)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, we want to classify our \"new\" horoscopes by their signs.\n",
    "We implement a naïve Bayes classifier from scratch.\n",
    "It is a simple \"probabilistic classifiers\" based on applying Bayes' theorem with strong (naïve) independence assumptions between the features.\n",
    "We use our datas to train it and then we can predict our new horoscopes. ( I should divide the training data into 2 parts to evaluate the accuracy. Buuut it can be low, soo maybe I won't show it in the real doc :( )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "aries\n"
     ]
    }
   ],
   "source": [
    "def prob_word_sign(sign, word):\n",
    "    num_in = 0\n",
    "    num_in_tot = 0\n",
    "    for i in index:\n",
    "        if hor.loc[i]['sign'] == sign:\n",
    "            num_in_tot = num_in_tot + 1\n",
    "            if word.lower() in hor.iloc[i]['text'].lower().split():\n",
    "                num_in = num_in + 1\n",
    "    p = (num_in + 1) / (num_in_tot + 2)\n",
    "    return math.log(p)\n",
    "\n",
    "def prob_sent(sent, sign):\n",
    "    sum = 0\n",
    "    for word in sent.split():\n",
    "        sum = sum + prob_word_sign(sign, word)\n",
    "    return sum\n",
    "\n",
    "def sign_sent(sent):\n",
    "    tuple = max((prob_sent(sent, sign), sign) for sign in signs)\n",
    "    return tuple[1]\n",
    "\n",
    "\n",
    "hor_csv = open(r'C:\\Users\\Giuliano\\Documents\\GitHub\\horoscope\\text_and_sign.csv', encoding=\"utf8\")\n",
    "hor = pd.read_csv(hor_csv)\n",
    "\n",
    "\n",
    "print(sign_sent(\"Ggrrrr!!!! Feel like taking a risk in love? Or maybe you want to add a bit of uh, spice,  to your relationship. Then you’ll be pleased to know that amorous Venus enters your house of free-spirited thinking today. However, this isn’t just about being kinky, so stop thinking 50 Shades. This is more about taking exotic vacations together, dating a brand new type or going to new venues on date night.\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}