lang_toxic_scraping/lang_identification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "20af37f8-1993-4f12-9bc8-e89b5b46226d",
   "metadata": {},
   "source": [
    "# Scraping YouTube for Language Identification and Toxicity Detection Tasks\n",
    "## Assignment #2\n",
    "###  Practical Data Science course, MSc in Data Science (2023/2024)\n",
    "\n",
    "---\n",
    "\n",
    "> Dimitris Tsirmpas <br>\n",
    "> MSc in Data Science f3352315 <br>\n",
    "> Athens University of Economics and Business"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "30e8cf04-0f13-49e2-850c-cf99acd707f6",
   "metadata": {},
   "source": [
    "In this project we attempt to achieve the following goals:\n",
    "\n",
    "- Creating a language dataset including Greeklish\n",
    "- Crawling YouTube videos which include both Greek and Greeklish comments\n",
    "- Training a language identification classifier\n",
    "- Training a LLM-based toxicity classifier\n",
    "- Using the LLM classifier to produce data for, and train a traditional ML toxicity classifier\n",
    "- Applying our language identification and toxicity classifiers on the crawled YouTube videos and identifying interesting facts and trends"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6e573599-9ae5-4594-956f-35514b0dbe4c",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true
   },
   "source": [
    "## Directory Structure\n",
    "\n",
    "The project is structured as follows:\n",
    "\n",
    "Main files:\n",
    "- lang_identification.ipynb: is the main Jupyter Notebook containing the project code\n",
    "- prompts.pdf: Supplemental material containing the prompts used for the toxicity LLM classifier\n",
    "- report.pdf: Supplemental material containing Figures, Tables and analysis on the results of the project\n",
    "  \n",
    "Directories:\n",
    "- src: a library of general functions for Data Science tasks\n",
    "- tasks: task-specific modules\n",
    "- data: the input data\n",
    "- output: the output data (.csv)\n",
    "- results: Graphs, Tables and Figures produced in the project"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "81287552-bc4b-454e-9a59-77b5f34461ff",
   "metadata": {},
   "source": [
    "## Disclaimers\n",
    "\n",
    "- Most documentation was generated by ChatGPT, and was manually corrected / augmented where necessary\n",
    "- In each case where code has been obtained from outside sources, is clearly listed either in comments or in the markdown explaining the code block\n",
    "- Most code implementation is \"hidden\" in the `src` and `tasks` modules, as stated in the section above, which contain all the documentation and implementation details"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "61c64785-61ad-4212-a91e-141f1cfd19b5",
   "metadata": {},
   "outputs": [],
   "source": [
    "from time import time\n",
    "\n",
    "start = time()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "613dd683-1740-4ee8-9e6c-46a061641b23",
   "metadata": {},
   "source": [
    "Before we begin, we need to have a baseline rule-based model which we will originally use to classify our input data. We thus create a model which predicts a text's language only using regex rules.\n",
    "This classifier obviously does not need to be trained (fitted), and thus we will only use it for predictions. The implementation of our model was inspired by [this excellent LinkedIn post](https://www.linkedin.com/pulse/regex-one-pattern-rule-them-all-find-bring-darkness-bind-carranza/).\n",
    "\n",
    "We will define our language identification task as identifiying one of three languages: Greek, English and Greeklish. We will also include an \"other\" category for all other languages."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "26f6978d-5009-4a94-89e1-ee8ccb01d081",
   "metadata": {},
   "source": [
    "## Regex classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "59b239c8-6826-4835-8b63-5874804b468a",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from sklearn.base import BaseEstimator, RegressorMixin\n",
    "import numpy as np\n",
    "import re\n",
    "\n",
    "\n",
    "class RegexClassifier(BaseEstimator, RegressorMixin):\n",
    "    \"\"\"\n",
    "    Language Classifier using Regular Expressions\n",
    "    \"\"\"\n",
    "\n",
    "    language_regex_dict = {\n",
    "        'el': r'([\\u0370-\\u03ff\\u1f00-\\u1fff]+)\\s?',  # Greek\n",
    "        'en': r'([a-zA-Z]+)\\s?',  # English\n",
    "    }\n",
    "\n",
    "    # STATIC INITALIZATION BLOCK\n",
    "    # pre-compie all expressions to save execution time\n",
    "    for lang in language_regex_dict.keys():\n",
    "        language_regex_dict[lang] = re.compile(language_regex_dict[lang])\n",
    "\n",
    "\n",
    "    def __init__(self, include_greeklish=True):\n",
    "        \"\"\"\n",
    "        Initialize a new :class:`RegexClassifier` instance.\n",
    "\n",
    "        :param include_greeklish: Optional. If True, include Greeklish in the language identification process.\n",
    "        \"\"\"\n",
    "        self.include_greeklish = include_greeklish\n",
    "\n",
    "    def predict(self, x):\n",
    "        preds = []\n",
    "\n",
    "        for text in x:\n",
    "            pred = \"other\"\n",
    "\n",
    "            if self.include_greeklish and \\\n",
    "                re.search(RegexClassifier.language_regex_dict[\"el\"], text) and \\\n",
    "                re.search(RegexClassifier.language_regex_dict[\"en\"], text):\n",
    "                    pred = \"greeklish\"\n",
    "            else:\n",
    "                for lang_code, regex_pattern in RegexClassifier.language_regex_dict.items():\n",
    "                    match = re.search(regex_pattern, text)\n",
    "                    if match:\n",
    "                        pred = lang_code\n",
    "                        break\n",
    "            preds.append(pred)\n",
    "        return np.array(preds)\n",
    "    \n",
    "    def fit(self, x, y):\n",
    "        return self"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "96a77733-0bfa-45e0-a5a7-4432baab8455",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Detected Language Code: ['other' 'en' 'el']\n"
     ]
    }
   ],
   "source": [
    "# Example usage\n",
    "user_input_text = [\"你好，これはサンプルです。안녕하세요\", \"this is an english sentence\", \"δεν αντεχω αλλο\"]\n",
    "detected_language = RegexClassifier().predict(user_input_text)\n",
    "print(f\"Detected Language Code: {detected_language}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2e6a8254-d567-492c-9a98-c944e7ca5199",
   "metadata": {},
   "source": [
    "## Defining the Gold Standard\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0ca590d3-e5ac-4c6b-a81f-f45d034b059e",
   "metadata": {},
   "source": [
    "The first task we ought to complete is derive our \"gold\" dataset, aka the dataset on which our classifiers will be trained on."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a77d5818-afe2-4c09-9a29-6f615dd492ec",
   "metadata": {},
   "source": [
    "### Greek-English identification dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "35fbce76-e950-4efc-9b38-09da72af29c8",
   "metadata": {},
   "source": [
    "We will begin by creating a base dataset including three of the four categories of data we defined in our language identification task (those being \"Greek\", \"English\" and \"Other\").\n",
    "\n",
    "For this we will be using a subset of the [papluca language identification dataset](https://huggingface.co/datasets/papluca/language-identification), which is available to us through `huggingface`."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "54c319e7-d83b-4be3-9821-103acfacd4ba",
   "metadata": {},
   "source": [
    "We download the dataset and turn it into a pandas Dataframe, so we can easily combine it with other sources later on."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "7884d336-7550-4a37-9125-fd31d52561e3",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "\n",
    "def dataset_to_pd(dataset_dict: dict) -> pd.DataFrame:\n",
    "    \"\"\"\n",
    "    Convert a dictionary of datasets into a single pandas DataFrame.\n",
    "    \n",
    "    The datasets are assumed to be stored as values in the input dictionary,\n",
    "    with corresponding labels as keys.\n",
    "    \n",
    "    :param dataset_dict: A dictionary where keys are labels and values are datasets.\n",
    "    :type dataset_dict: dict\n",
    "    \n",
    "    :return: A concatenated pandas DataFrame with an additional 'set' column\n",
    "         indicating the original label of each row.\n",
    "    :rtype: pd.DataFrame\n",
    "    \n",
    "    :Example:\n",
    "    \n",
    "    .. code-block:: python\n",
    "    \n",
    "    >>> dataset_dict = {'A': dataset_A, 'B': dataset_B, 'C': dataset_C}\n",
    "    >>> result_df = dataset_to_pd(dataset_dict)\n",
    "    >>> print(result_df)\n",
    "         set  ...  (columns of the datasets)\n",
    "    0    A   ...  ...\n",
    "    1    A   ...  ...\n",
    "    ...      ...  ...\n",
    "    n    C   ...  ...\n",
    "    \"\"\"\n",
    "    df_ls = []\n",
    "    label_array = np.empty(shape=(sum([len(dataset) for dataset in dataset_dict.values()])), dtype=object)\n",
    "    last_idx = -1\n",
    "\n",
    "    for label, dataset in dataset_dict.items():\n",
    "        new_last_idx = len(dataset) + last_idx\n",
    "        label_array.put(np.arange(last_idx+1, new_last_idx+1, 1), label)\n",
    "        last_idx = new_last_idx\n",
    "\n",
    "        df_ls.append(pd.DataFrame(dataset))\n",
    "\n",
    "    full_df = pd.concat(df_ls, ignore_index=True)\n",
    "    full_df[\"set\"] = label_array\n",
    "    full_df.insert(0, \"set\", full_df.pop(\"set\"))\n",
    "\n",
    "    return full_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "321846d4-f30f-46f6-b67d-cae6566f7ea0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DatasetDict({\n",
       "    train: Dataset({\n",
       "        features: ['labels', 'text'],\n",
       "        num_rows: 70000\n",
       "    })\n",
       "    validation: Dataset({\n",
       "        features: ['labels', 'text'],\n",
       "        num_rows: 10000\n",
       "    })\n",
       "    test: Dataset({\n",
       "        features: ['labels', 'text'],\n",
       "        num_rows: 10000\n",
       "    })\n",
       "})"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from datasets import load_dataset\n",
    "\n",
    "\n",
    "dataset_dict = load_dataset(\"papluca/language-identification\")\n",
    "dataset_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "fd8bd535-bb09-413a-b30b-cc3c978f93fd",
   "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>set</th>\n",
       "      <th>labels</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>train</td>\n",
       "      <td>pt</td>\n",
       "      <td>os chefes de defesa da estónia, letónia, lituâ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>train</td>\n",
       "      <td>bg</td>\n",
       "      <td>размерът на хоризонталната мрежа може да бъде ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>train</td>\n",
       "      <td>zh</td>\n",
       "      <td>很好，以前从不去评价，不知道浪费了多少积分，现在知道积分可以换钱，就要好好评价了，后来我就把...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>train</td>\n",
       "      <td>th</td>\n",
       "      <td>สำหรับ ของเก่า ที่ จริงจัง ลอง   honeychurch  ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>train</td>\n",
       "      <td>ru</td>\n",
       "      <td>Он увеличил давление .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89995</th>\n",
       "      <td>test</td>\n",
       "      <td>zh</td>\n",
       "      <td>史料很充分，对岸的很多观点与大陆迥异啊。</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89996</th>\n",
       "      <td>test</td>\n",
       "      <td>tr</td>\n",
       "      <td>Örneğin, teşhis Yunanca bir kelimeden alındı (...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89997</th>\n",
       "      <td>test</td>\n",
       "      <td>vi</td>\n",
       "      <td>Nếu lite/light chỉ đơn giản là mô tả một đặc t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89998</th>\n",
       "      <td>test</td>\n",
       "      <td>bg</td>\n",
       "      <td>Например, една щатска столица, която посетихме...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89999</th>\n",
       "      <td>test</td>\n",
       "      <td>pl</td>\n",
       "      <td>Mam dla ciebie kilka propozycji:</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>90000 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         set labels                                               text\n",
       "0      train     pt  os chefes de defesa da estónia, letónia, lituâ...\n",
       "1      train     bg  размерът на хоризонталната мрежа може да бъде ...\n",
       "2      train     zh  很好，以前从不去评价，不知道浪费了多少积分，现在知道积分可以换钱，就要好好评价了，后来我就把...\n",
       "3      train     th  สำหรับ ของเก่า ที่ จริงจัง ลอง   honeychurch  ...\n",
       "4      train     ru                             Он увеличил давление .\n",
       "...      ...    ...                                                ...\n",
       "89995   test     zh                               史料很充分，对岸的很多观点与大陆迥异啊。\n",
       "89996   test     tr  Örneğin, teşhis Yunanca bir kelimeden alındı (...\n",
       "89997   test     vi  Nếu lite/light chỉ đơn giản là mô tả một đặc t...\n",
       "89998   test     bg  Например, една щатска столица, която посетихме...\n",
       "89999   test     pl                   Mam dla ciebie kilka propozycji:\n",
       "\n",
       "[90000 rows x 3 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lang_df = dataset_to_pd(dataset_dict)\n",
    "lang_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "b65d7720-bf89-4b94-8283-4544209ea5be",
   "metadata": {
    "tags": []
   },
   "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>labels</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>el</td>\n",
       "      <td>Παρά τον εαυτό μου , η γνώμη μου για τη σύνεση...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>en</td>\n",
       "      <td>Didnt really seem to work much.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>el</td>\n",
       "      <td>Ακόμα και το να ξέρεις ότι ήταν ένα άγαλμα που...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49</th>\n",
       "      <td>en</td>\n",
       "      <td>Highly recommend for those who don't like bein...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75</th>\n",
       "      <td>el</td>\n",
       "      <td>Έχω κάνει τέσσερις ή πέντε .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89961</th>\n",
       "      <td>en</td>\n",
       "      <td>It's super cute, really soft. Print is fine bu...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89965</th>\n",
       "      <td>en</td>\n",
       "      <td>One of them worked, the other one didn't. Ther...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89978</th>\n",
       "      <td>en</td>\n",
       "      <td>I only received one out of the three strikers :(</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89982</th>\n",
       "      <td>el</td>\n",
       "      <td>Οι οργανισμοί πρέπει να είναι σε θέση να μετρή...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89986</th>\n",
       "      <td>el</td>\n",
       "      <td>Το Abeam με την Arabella, απέναντι από το λιμά...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>9000 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      labels                                               text\n",
       "18        el  Παρά τον εαυτό μου , η γνώμη μου για τη σύνεση...\n",
       "39        en                    Didnt really seem to work much.\n",
       "40        el  Ακόμα και το να ξέρεις ότι ήταν ένα άγαλμα που...\n",
       "49        en  Highly recommend for those who don't like bein...\n",
       "75        el                       Έχω κάνει τέσσερις ή πέντε .\n",
       "...      ...                                                ...\n",
       "89961     en  It's super cute, really soft. Print is fine bu...\n",
       "89965     en  One of them worked, the other one didn't. Ther...\n",
       "89978     en   I only received one out of the three strikers :(\n",
       "89982     el  Οι οργανισμοί πρέπει να είναι σε θέση να μετρή...\n",
       "89986     el  Το Abeam με την Arabella, απέναντι από το λιμά...\n",
       "\n",
       "[9000 rows x 2 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "en_gr_cond = lang_df.labels.eq(\"el\") | lang_df.labels.eq(\"en\")\n",
    "en_gr_df = lang_df.loc[en_gr_cond, [\"labels\", \"text\"]]\n",
    "en_gr_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3ec0525f-7498-4319-a45b-30b4ca30b189",
   "metadata": {},
   "source": [
    "We sample a set of 2000 data records belonging to other languages and tag them as \"other\". The exact number is arbitrary, although in general we would like to have a size:\n",
    "- large enough to contain most common words from foreign languages\n",
    "- small enough to not unnecessarily burden training or bias our classifiers towards rare occurences in our operational set (the dataset which we claim little to no prior knowledge)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "68aa37a0-6802-471a-a55a-40636d718f12",
   "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>labels</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>18882</th>\n",
       "      <td>other</td>\n",
       "      <td>不错，很好，虽然送错地点了，但是还是解决了</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54679</th>\n",
       "      <td>other</td>\n",
       "      <td>给我寄的咖啡粉是2016年6月份的（18个月之前的）。而且外包装是裂开的。虐心啊！</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2154</th>\n",
       "      <td>other</td>\n",
       "      <td>Il Presidente del Consiglio dei Ministri Mario...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32815</th>\n",
       "      <td>other</td>\n",
       "      <td>1946 ' de ingiliz görev personeli tarafından i...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24227</th>\n",
       "      <td>other</td>\n",
       "      <td>Ugoda zawiera 4,1 miliona dolarów na honoraria...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12579</th>\n",
       "      <td>other</td>\n",
       "      <td>カンフーシーンをもうちょい減らしていいから人物背景をもっと増やしてくれたらよかった。 特に殺...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38271</th>\n",
       "      <td>other</td>\n",
       "      <td>値段なりの商品です。某100円ショップで売っていそうな商品です。</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24788</th>\n",
       "      <td>other</td>\n",
       "      <td>'   แต่ พ่อแม่ ผม ไม่ ชอบ แบบนี้</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37396</th>\n",
       "      <td>other</td>\n",
       "      <td>महान यहूदी linguist मैक ् स weinreich के अनुसा...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>65858</th>\n",
       "      <td>other</td>\n",
       "      <td>l'agenzia internazionale per l'energia atomica...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2000 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      labels                                               text\n",
       "18882  other                              不错，很好，虽然送错地点了，但是还是解决了\n",
       "54679  other          给我寄的咖啡粉是2016年6月份的（18个月之前的）。而且外包装是裂开的。虐心啊！\n",
       "2154   other  Il Presidente del Consiglio dei Ministri Mario...\n",
       "32815  other  1946 ' de ingiliz görev personeli tarafından i...\n",
       "24227  other  Ugoda zawiera 4,1 miliona dolarów na honoraria...\n",
       "...      ...                                                ...\n",
       "12579  other  カンフーシーンをもうちょい減らしていいから人物背景をもっと増やしてくれたらよかった。 特に殺...\n",
       "38271  other                   値段なりの商品です。某100円ショップで売っていそうな商品です。\n",
       "24788  other                   '   แต่ พ่อแม่ ผม ไม่ ชอบ แบบนี้\n",
       "37396  other  महान यहूदी linguist मैक ् स weinreich के अनुसा...\n",
       "65858  other  l'agenzia internazionale per l'energia atomica...\n",
       "\n",
       "[2000 rows x 2 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "others_df = lang_df.loc[~en_gr_cond, [\"labels\", \"text\"]]\n",
    "others_df = others_df.sample(2000)\n",
    "others_df.labels = \"other\"\n",
    "others_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "6b3b00e0-ad10-471d-bf7b-6ec73579d69a",
   "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>labels</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>el</td>\n",
       "      <td>Παρά τον εαυτό μου , η γνώμη μου για τη σύνεση...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>en</td>\n",
       "      <td>Didnt really seem to work much.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>el</td>\n",
       "      <td>Ακόμα και το να ξέρεις ότι ήταν ένα άγαλμα που...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>en</td>\n",
       "      <td>Highly recommend for those who don't like bein...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>el</td>\n",
       "      <td>Έχω κάνει τέσσερις ή πέντε .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10995</th>\n",
       "      <td>other</td>\n",
       "      <td>カンフーシーンをもうちょい減らしていいから人物背景をもっと増やしてくれたらよかった。 特に殺...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10996</th>\n",
       "      <td>other</td>\n",
       "      <td>値段なりの商品です。某100円ショップで売っていそうな商品です。</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10997</th>\n",
       "      <td>other</td>\n",
       "      <td>'   แต่ พ่อแม่ ผม ไม่ ชอบ แบบนี้</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10998</th>\n",
       "      <td>other</td>\n",
       "      <td>महान यहूदी linguist मैक ् स weinreich के अनुसा...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10999</th>\n",
       "      <td>other</td>\n",
       "      <td>l'agenzia internazionale per l'energia atomica...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>11000 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      labels                                               text\n",
       "0         el  Παρά τον εαυτό μου , η γνώμη μου για τη σύνεση...\n",
       "1         en                    Didnt really seem to work much.\n",
       "2         el  Ακόμα και το να ξέρεις ότι ήταν ένα άγαλμα που...\n",
       "3         en  Highly recommend for those who don't like bein...\n",
       "4         el                       Έχω κάνει τέσσερις ή πέντε .\n",
       "...      ...                                                ...\n",
       "10995  other  カンフーシーンをもうちょい減らしていいから人物背景をもっと増やしてくれたらよかった。 特に殺...\n",
       "10996  other                   値段なりの商品です。某100円ショップで売っていそうな商品です。\n",
       "10997  other                   '   แต่ พ่อแม่ ผม ไม่ ชอบ แบบนี้\n",
       "10998  other  महान यहूदी linguist मैक ् स weinreich के अनुसा...\n",
       "10999  other  l'agenzia internazionale per l'energia atomica...\n",
       "\n",
       "[11000 rows x 2 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gold1_df = pd.concat([en_gr_df, others_df], axis=0, ignore_index=True, copy=False)\n",
    "gold1_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c2697de5-d487-4603-8708-cc502034a3dd",
   "metadata": {},
   "source": [
    "### Greek-Greeklish identification dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d2453f47-2e4a-4453-9ea7-cbd205740cb9",
   "metadata": {},
   "source": [
    "A much harder task will be to incude Greeklish in our dataset. There are only few papers dedicated to Greeklish language identification or translation `[1,2]`. Out of these, only one provides [a comprehensive Greeklish dataset](http://metashare.ilsp.gr:8080/repository/browse/greeklish-database-from-actual-users/23e0174a63cb11e2b2f8842b2b6a04d7c72cf16d853e4c1db3995068af41a6c1/) `[2]`, which is not publically available.\n",
    "\n",
    "```\n",
    "[1] Aimilios Chalamandaris, Athanassios Protopapas, Pirros Tsiakoulis, and Spyros Raptis. 2006. All Greek to me! An automatic Greeklish to Greek transliteration system. In Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC’06), Genoa, Italy. European Language Resources Association (ELRA).\n",
    "[2] A. Chalamandaris, P. Tsiakoulis, S. Raptis, G. Giannopoulos, and G. Carayannis. 2004. Bypassing Greeklish!. In Proceedings of the Fourth International Conference on Language Resources and Evaluation (LREC’04), Lisbon, Portugal. European Language Resources Association (ELRA).\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "65986a2c-eb33-471a-ab66-07df4f6b6c1d",
   "metadata": {},
   "source": [
    "We thus need to create our own dataset. We begin with crawling a Greek gaming forum, which we have verified contains many posts in Greeklish and almost none in English."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "fbf1b45a-9a2f-419e-b900-4c6e873349cc",
   "metadata": {},
   "outputs": [],
   "source": [
    "head_url = \"https://forum.warmane.com\"\n",
    "warmane_url = \"https://forum.warmane.com/forumdisplay.php?f=20\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "03a32104-f828-44e2-b9e6-fb0a1d5bd71c",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing page 1 of 8...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:04<00:00,  4.15it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing page 2 of 8...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:03<00:00,  5.18it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing page 3 of 8...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:04<00:00,  4.97it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing page 4 of 8...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:03<00:00,  5.39it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing page 5 of 8...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:03<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing page 6 of 8...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:03<00:00,  5.71it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing page 7 of 8...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 70%|█████████████████████████████████████████████████████████▍                        | 14/20 [00:02<00:01,  4.97it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ERROR: Failed to get information on post  https://forum.warmane.com/showthread.php?t=272585\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:03<00:00,  5.35it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing page 8 of 8...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 78%|█████████████████████████████████████████████████████████████████▎                  | 7/9 [00:01<00:00,  5.45it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ERROR: Failed to get information on post  https://forum.warmane.com/showthread.php?t=278731\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 9/9 [00:01<00:00,  5.66it/s]\n"
     ]
    }
   ],
   "source": [
    "from src.crawling import fetch_soup\n",
    "from tasks.warmane import parse_warmane_thread\n",
    "from tqdm import tqdm\n",
    "\n",
    "\n",
    "threads = []\n",
    "\n",
    "for page in range(1, 9):\n",
    "    url = warmane_url + f\"&page={page}\"\n",
    "    soup = fetch_soup(url)\n",
    "\n",
    "    print(f\"Processing page {page} of 8...\")\n",
    "    thread_tags = soup.find_all(\"li\", {\"class\": \"threadbit\"})\n",
    "    for thread_tag in tqdm(thread_tags):\n",
    "        thread = parse_warmane_thread(head_url, thread_tag)\n",
    "        threads.append(thread)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "e0d9a523-1311-4960-ab66-1f9b98c03729",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "415"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import itertools\n",
    "\n",
    "# flatten nested lists \n",
    "posts = set(itertools.chain.from_iterable([thread.posts for thread in threads]))\n",
    "len(posts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "2425baf7-a694-4a49-98c0-11d565c2cb98",
   "metadata": {
    "tags": []
   },
   "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>thread_id</th>\n",
       "      <th>author</th>\n",
       "      <th>contents</th>\n",
       "      <th>date</th>\n",
       "      <th>reply_to</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</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>2926596</th>\n",
       "      <td>384475</td>\n",
       "      <td>Ripsin</td>\n",
       "      <td>Kalhspera paides,\\n\\r\\nEimai arketo kairo ston...</td>\n",
       "      <td>2018-05-22</td>\n",
       "      <td>-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2473988</th>\n",
       "      <td>300013</td>\n",
       "      <td>v4gflo</td>\n",
       "      <td>geia sas.psaxnw ellhniko guild ston Deathwing ...</td>\n",
       "      <td>2015-06-17</td>\n",
       "      <td>-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2420747</th>\n",
       "      <td>290921</td>\n",
       "      <td>AlexPan</td>\n",
       "      <td>Καλησπέρα παιδιά, το πρόβλημα είναι το εξής. \\...</td>\n",
       "      <td>2015-03-24</td>\n",
       "      <td>-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2981903</th>\n",
       "      <td>399822</td>\n",
       "      <td>xAchillesGate4x</td>\n",
       "      <td>Καλησπέρα παίδες. Ψάχνω Ελληνικό active raidin...</td>\n",
       "      <td>2019-03-03</td>\n",
       "      <td>-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2879517</th>\n",
       "      <td>371804</td>\n",
       "      <td>Csdas</td>\n",
       "      <td>Opoios gnwrizei kati as mou kanei /w Dremoria ...</td>\n",
       "      <td>2017-11-29</td>\n",
       "      <td>-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2877428</th>\n",
       "      <td>353812</td>\n",
       "      <td>Shiverbro</td>\n",
       "      <td>kalos private aksizei na ksekiniseis paidia?</td>\n",
       "      <td>2017-11-21</td>\n",
       "      <td>2875915</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3069941</th>\n",
       "      <td>423611</td>\n",
       "      <td>crystallenia898</td>\n",
       "      <td>Ε μεις αυξανόμαστε και θα αρχίσουμε να στήνουμ...</td>\n",
       "      <td>2020-07-26</td>\n",
       "      <td>3068345</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2801654</th>\n",
       "      <td>350071</td>\n",
       "      <td>Draculation</td>\n",
       "      <td>Bump! ICC25 6/12</td>\n",
       "      <td>2017-05-07</td>\n",
       "      <td>2795443</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2873339</th>\n",
       "      <td>370241</td>\n",
       "      <td>Ripsin</td>\n",
       "      <td>Kalhspera tha ithela na rwthsw an kapoios gnwr...</td>\n",
       "      <td>2017-11-07</td>\n",
       "      <td>-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2410495</th>\n",
       "      <td>289030</td>\n",
       "      <td>boolouk</td>\n",
       "      <td>Εμπειρια εχω, αλλα δεν εχω χρονο...αυτο ειναι ...</td>\n",
       "      <td>2015-03-13</td>\n",
       "      <td>2409274</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>415 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         thread_id           author  \\\n",
       "id                                    \n",
       "2926596     384475           Ripsin   \n",
       "2473988     300013           v4gflo   \n",
       "2420747     290921          AlexPan   \n",
       "2981903     399822  xAchillesGate4x   \n",
       "2879517     371804            Csdas   \n",
       "...            ...              ...   \n",
       "2877428     353812        Shiverbro   \n",
       "3069941     423611  crystallenia898   \n",
       "2801654     350071      Draculation   \n",
       "2873339     370241           Ripsin   \n",
       "2410495     289030          boolouk   \n",
       "\n",
       "                                                  contents       date  \\\n",
       "id                                                                      \n",
       "2926596  Kalhspera paides,\\n\\r\\nEimai arketo kairo ston... 2018-05-22   \n",
       "2473988  geia sas.psaxnw ellhniko guild ston Deathwing ... 2015-06-17   \n",
       "2420747  Καλησπέρα παιδιά, το πρόβλημα είναι το εξής. \\... 2015-03-24   \n",
       "2981903  Καλησπέρα παίδες. Ψάχνω Ελληνικό active raidin... 2019-03-03   \n",
       "2879517  Opoios gnwrizei kati as mou kanei /w Dremoria ... 2017-11-29   \n",
       "...                                                    ...        ...   \n",
       "2877428       kalos private aksizei na ksekiniseis paidia? 2017-11-21   \n",
       "3069941  Ε μεις αυξανόμαστε και θα αρχίσουμε να στήνουμ... 2020-07-26   \n",
       "2801654                                   Bump! ICC25 6/12 2017-05-07   \n",
       "2873339  Kalhspera tha ithela na rwthsw an kapoios gnwr... 2017-11-07   \n",
       "2410495  Εμπειρια εχω, αλλα δεν εχω χρονο...αυτο ειναι ... 2015-03-13   \n",
       "\n",
       "         reply_to  \n",
       "id                 \n",
       "2926596        -1  \n",
       "2473988        -1  \n",
       "2420747        -1  \n",
       "2981903        -1  \n",
       "2879517        -1  \n",
       "...           ...  \n",
       "2877428   2875915  \n",
       "3069941   3068345  \n",
       "2801654   2795443  \n",
       "2873339        -1  \n",
       "2410495   2409274  \n",
       "\n",
       "[415 rows x 5 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "warmane_df = pd.DataFrame.from_records([post.__dict__ for post in posts], index=\"id\")\n",
    "warmane_df.reply_to = warmane_df.reply_to.fillna(-1).astype(int)\n",
    "warmane_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6991044b-992a-4ca8-906a-0f322ed21e46",
   "metadata": {},
   "source": [
    "We will also clear out empty posts."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "dbffc615-c381-41d5-b39f-0d25f77a63e2",
   "metadata": {
    "tags": []
   },
   "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>thread_id</th>\n",
       "      <th>author</th>\n",
       "      <th>contents</th>\n",
       "      <th>date</th>\n",
       "      <th>reply_to</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>id</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>3082464</th>\n",
       "      <td>427259</td>\n",
       "      <td>malakas17</td>\n",
       "      <td></td>\n",
       "      <td>2020-10-20</td>\n",
       "      <td>3081822</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3113236</th>\n",
       "      <td>427259</td>\n",
       "      <td>malakas17</td>\n",
       "      <td></td>\n",
       "      <td>2021-05-12</td>\n",
       "      <td>3113009</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3099161</th>\n",
       "      <td>431660</td>\n",
       "      <td>malakas17</td>\n",
       "      <td></td>\n",
       "      <td>2021-02-10</td>\n",
       "      <td>3096432</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3113819</th>\n",
       "      <td>427259</td>\n",
       "      <td>malakas17</td>\n",
       "      <td></td>\n",
       "      <td>2021-05-16</td>\n",
       "      <td>3113236</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3099593</th>\n",
       "      <td>427259</td>\n",
       "      <td>boonick</td>\n",
       "      <td></td>\n",
       "      <td>2021-02-14</td>\n",
       "      <td>3093400</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3081820</th>\n",
       "      <td>427259</td>\n",
       "      <td>malakas17</td>\n",
       "      <td></td>\n",
       "      <td>2020-10-16</td>\n",
       "      <td>3080427</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3081822</th>\n",
       "      <td>427259</td>\n",
       "      <td>malakas17</td>\n",
       "      <td></td>\n",
       "      <td>2020-10-16</td>\n",
       "      <td>3081820</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         thread_id     author contents       date  reply_to\n",
       "id                                                         \n",
       "3082464     427259  malakas17          2020-10-20   3081822\n",
       "3113236     427259  malakas17          2021-05-12   3113009\n",
       "3099161     431660  malakas17          2021-02-10   3096432\n",
       "3113819     427259  malakas17          2021-05-16   3113236\n",
       "3099593     427259    boonick          2021-02-14   3093400\n",
       "3081820     427259  malakas17          2020-10-16   3080427\n",
       "3081822     427259  malakas17          2020-10-16   3081820"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "empty_contents = warmane_df.contents.apply(lambda x: x.isspace() | len(x)==0)\n",
    "warmane_df[empty_contents]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "29103c28-5579-41e8-b33e-a9fa7a49428d",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "warmane_df = warmane_df[~empty_contents]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c927eb64-514b-4e3d-87bc-0859f0d2beb8",
   "metadata": {},
   "source": [
    "While this dataset fits our needs, it is by no means large enough to accurately model Greeklish on traditional NLP models. \n",
    "\n",
    "Thus, we turn our attention early to YouTube scraping. YouTube is one of the few sites featuring lively comment sections in Greek informal enough for Greeklish to be present. Knowing that Greeklish are generally more prevalent towards younger generations, we can select certain videos where this demographic is present to scrape for comments. In our case, we select 5 Greek gaming videos."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "923224eb-9f12-44c7-b463-8fb262165bbb",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from src.crawling import ChromeDriverManager, jupyter_options\n",
    "\n",
    "\n",
    "ChromeDriverManager.set_options(jupyter_options())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "a77b074e-1271-421b-aabb-141681485e82",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                            | 0/5 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating new driver...\n",
      "New driver online.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 5/5 [01:35<00:00, 19.00s/it]\n"
     ]
    }
   ],
   "source": [
    "from tasks.youtube import extract_search_results, extract_comments, scrape_youtube\n",
    "\n",
    "\n",
    "greek_yt_urls = [\"https://www.youtube.com/watch?v=4Y2gxkqbsbA\", \n",
    "       \"https://www.youtube.com/watch?v=31LcJ9gqQvA\",\n",
    "       \"https://www.youtube.com/watch?v=1cZXAQ1JEJo\",\n",
    "       \"https://www.youtube.com/watch?v=x7lnS6jMS64\",\n",
    "       \"https://www.youtube.com/watch?v=ImilczGN-00\"]\n",
    "scrape_results = []\n",
    "\n",
    "for url in tqdm(greek_yt_urls):\n",
    "    scrape_results.append(scrape_youtube(ChromeDriverManager.get(), url, max_scrolls=10, verbose=False))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "7f61a6d4-3a7f-4f66-b990-fe070c7b1527",
   "metadata": {},
   "outputs": [],
   "source": [
    "from tasks.youtube import extract_comments\n",
    "\n",
    "\n",
    "all_comments = []\n",
    "for result in scrape_results:\n",
    "    comments, _ = extract_comments(result)\n",
    "    all_comments += comments\n",
    "\n",
    "all_comments = pd.Series(all_comments)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1d140310-d3f5-4be4-b3ab-51997d4e39f0",
   "metadata": {},
   "source": [
    "We now combine our two Greeklish datasets:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "7c965cdf-969f-4385-939f-21a7c20b4e7e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2926596    Kalhspera paides,\\n\\r\\nEimai arketo kairo ston...\n",
       "2473988    geia sas.psaxnw ellhniko guild ston Deathwing ...\n",
       "2420747    Καλησπέρα παιδιά, το πρόβλημα είναι το εξής. \\...\n",
       "2981903    Καλησπέρα παίδες. Ψάχνω Ελληνικό active raidin...\n",
       "2879517    Opoios gnwrizei kati as mou kanei /w Dremoria ...\n",
       "                                 ...                        \n",
       "744                                                    Πρώτη\n",
       "745                                             First of all\n",
       "746                                                    First\n",
       "747                                                        .\n",
       "748                                                   Πρώτος\n",
       "Length: 1157, dtype: object"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "greeklish_series = pd.concat([warmane_df.contents, all_comments])\n",
    "greeklish_series"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9d548f2b-c783-4c03-b9e8-76f4ac8e39aa",
   "metadata": {},
   "source": [
    "We filter out empty and \"junk\" comments."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "bcdd8c12-f28c-4ddb-83dc-0eb90a839d04",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2926596    Kalhspera paides,\\n\\r\\nEimai arketo kairo ston...\n",
       "2473988    geia sas.psaxnw ellhniko guild ston Deathwing ...\n",
       "2420747    Καλησπέρα παιδιά, το πρόβλημα είναι το εξής. \\...\n",
       "2981903    Καλησπέρα παίδες. Ψάχνω Ελληνικό active raidin...\n",
       "2879517    Opoios gnwrizei kati as mou kanei /w Dremoria ...\n",
       "                                 ...                        \n",
       "742                                                   Πρώτος\n",
       "743                                        Gianni παίξε Pubg\n",
       "744                                                    Πρώτη\n",
       "747                                                        .\n",
       "748                                                   Πρώτος\n",
       "Length: 1139, dtype: object"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "conditions = (greeklish_series.apply(lambda x: len(x) != 0)) & \\\n",
    "                (greeklish_series.apply(lambda x: \"RRR\" not in x)) & \\\n",
    "                (greeklish_series.apply(lambda x: \"PPP\" not in x)) & \\\n",
    "                (greeklish_series.apply(lambda x: \"First\" not in x))\n",
    "cleared_greeklish_series = greeklish_series[conditions]\n",
    "cleared_greeklish_series"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "110174d3-004c-4dbb-ad80-ed4b54206c0c",
   "metadata": {},
   "source": [
    "And annotate the entire Greeklish dataset using our rules-based (Regex) classifier. \n",
    "\n",
    "Since we selected videos exclusively in Greek, we can safely assume that the vast majority of comments not featuring Greek characters are in Greeklish. We can also safely assume from prior knowledge that Greeklish comments will conversly not feature any Greek characters. \n",
    "\n",
    "We thus classify all comments with English charactes are Greeklish."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "3c6bc888-adf2-4c94-bddf-4af870e9a0f6",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "regex_model = RegexClassifier(include_greeklish=False)\n",
    "preds = regex_model.predict(cleared_greeklish_series)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e71babb0-522b-4417-8d78-8df76e88c6ad",
   "metadata": {},
   "source": [
    "We can briefly verify that our assumptions are correct:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "46995910-7936-46c4-ac34-611a8fdaa307",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2926596    Kalhspera paides,\\n\\r\\nEimai arketo kairo ston...\n",
       "2473988    geia sas.psaxnw ellhniko guild ston Deathwing ...\n",
       "2879517    Opoios gnwrizei kati as mou kanei /w Dremoria ...\n",
       "2959390               Bubblethesap Icecrown wotlk horde belf\n",
       "2947119    den se vrisko kane add evvi  .\\nmou leei den u...\n",
       "                                 ...                        \n",
       "730                                                     Geia\n",
       "731                                                      Lol\n",
       "732        Ante Pali me ta atoma pou einai first.... Mhn ...\n",
       "733                                Protos Protos molis vgike\n",
       "740                                                     Hafa\n",
       "Length: 430, dtype: object"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cleared_greeklish_series[preds==\"en\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "17442e3b-93c7-4a53-82c2-bd26c63b0f16",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2420747    Καλησπέρα παιδιά, το πρόβλημα είναι το εξής. \\...\n",
       "2981903    Καλησπέρα παίδες. Ψάχνω Ελληνικό active raidin...\n",
       "2959391    Καλησπερα θα πας στις ιδιότητες (properties) σ...\n",
       "2719776    Originally Posted by celphecil\\n\\nΚαλησπέρα Σω...\n",
       "2971700    Ελληνικο guild , ψαχνουμε ατομα για runs ICC10...\n",
       "                                 ...                        \n",
       "741                                               Πρώτο like\n",
       "742                                                   Πρώτος\n",
       "743                                        Gianni παίξε Pubg\n",
       "744                                                    Πρώτη\n",
       "748                                                   Πρώτος\n",
       "Length: 703, dtype: object"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cleared_greeklish_series[preds==\"el\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "bba2d5fd-9fb7-4fd7-b8f5-a9944d337c07",
   "metadata": {
    "tags": []
   },
   "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>labels</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2926596</th>\n",
       "      <td>greeklish</td>\n",
       "      <td>Kalhspera paides,\\n\\r\\nEimai arketo kairo ston...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2473988</th>\n",
       "      <td>greeklish</td>\n",
       "      <td>geia sas.psaxnw ellhniko guild ston Deathwing ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2420747</th>\n",
       "      <td>el</td>\n",
       "      <td>Καλησπέρα παιδιά, το πρόβλημα είναι το εξής. \\...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2981903</th>\n",
       "      <td>el</td>\n",
       "      <td>Καλησπέρα παίδες. Ψάχνω Ελληνικό active raidin...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2879517</th>\n",
       "      <td>greeklish</td>\n",
       "      <td>Opoios gnwrizei kati as mou kanei /w Dremoria ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>742</th>\n",
       "      <td>el</td>\n",
       "      <td>Πρώτος</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>743</th>\n",
       "      <td>el</td>\n",
       "      <td>Gianni παίξε Pubg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>744</th>\n",
       "      <td>el</td>\n",
       "      <td>Πρώτη</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>747</th>\n",
       "      <td>el</td>\n",
       "      <td>.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>748</th>\n",
       "      <td>el</td>\n",
       "      <td>Πρώτος</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1139 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            labels                                               text\n",
       "2926596  greeklish  Kalhspera paides,\\n\\r\\nEimai arketo kairo ston...\n",
       "2473988  greeklish  geia sas.psaxnw ellhniko guild ston Deathwing ...\n",
       "2420747         el  Καλησπέρα παιδιά, το πρόβλημα είναι το εξής. \\...\n",
       "2981903         el  Καλησπέρα παίδες. Ψάχνω Ελληνικό active raidin...\n",
       "2879517  greeklish  Opoios gnwrizei kati as mou kanei /w Dremoria ...\n",
       "...            ...                                                ...\n",
       "742             el                                             Πρώτος\n",
       "743             el                                  Gianni παίξε Pubg\n",
       "744             el                                              Πρώτη\n",
       "747             el                                                  .\n",
       "748             el                                             Πρώτος\n",
       "\n",
       "[1139 rows x 2 columns]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels = np.where(preds==\"en\", \"greeklish\", \"el\")\n",
    "gold2_df = pd.DataFrame({\"labels\": labels, \"text\": cleared_greeklish_series})\n",
    "gold2_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ae40ce06-ee64-4289-be8b-4ed40eac04f5",
   "metadata": {},
   "source": [
    "Having our Greek-English-Other and our Greek-Greeklish datasets we can now combine them to form our gold dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "ff7306dd-67e1-4bf6-be4b-6c8974964b61",
   "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>labels</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>el</td>\n",
       "      <td>Παρά τον εαυτό μου , η γνώμη μου για τη σύνεση...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>en</td>\n",
       "      <td>Didnt really seem to work much.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>el</td>\n",
       "      <td>Ακόμα και το να ξέρεις ότι ήταν ένα άγαλμα που...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>en</td>\n",
       "      <td>Highly recommend for those who don't like bein...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>el</td>\n",
       "      <td>Έχω κάνει τέσσερις ή πέντε .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>742</th>\n",
       "      <td>el</td>\n",
       "      <td>Πρώτος</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>743</th>\n",
       "      <td>el</td>\n",
       "      <td>Gianni παίξε Pubg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>744</th>\n",
       "      <td>el</td>\n",
       "      <td>Πρώτη</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>747</th>\n",
       "      <td>el</td>\n",
       "      <td>.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>748</th>\n",
       "      <td>el</td>\n",
       "      <td>Πρώτος</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>12139 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    labels                                               text\n",
       "0       el  Παρά τον εαυτό μου , η γνώμη μου για τη σύνεση...\n",
       "1       en                    Didnt really seem to work much.\n",
       "2       el  Ακόμα και το να ξέρεις ότι ήταν ένα άγαλμα που...\n",
       "3       en  Highly recommend for those who don't like bein...\n",
       "4       el                       Έχω κάνει τέσσερις ή πέντε .\n",
       "..     ...                                                ...\n",
       "742     el                                             Πρώτος\n",
       "743     el                                  Gianni παίξε Pubg\n",
       "744     el                                              Πρώτη\n",
       "747     el                                                  .\n",
       "748     el                                             Πρώτος\n",
       "\n",
       "[12139 rows x 2 columns]"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gold_df = pd.concat([gold1_df, gold2_df])\n",
    "gold_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "be21cbf4-0e05-490c-b6a2-3c68b425c89b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "\n",
    "OUTPUT_DIR = \"output\"\n",
    "\n",
    "def csv_output(df: pd.DataFrame, filename: str) -> None:\n",
    "    \"\"\"\n",
    "    Save a pandas DataFrame to a CSV file.\n",
    "\n",
    "    :param df: The DataFrame to be saved.\n",
    "    :type df: pd.DataFrame\n",
    "\n",
    "    :param filename: The name of the CSV file.\n",
    "    :type filename: str\n",
    "\n",
    "    :return: This function does not return anything.\n",
    "    :rtype: None\n",
    "    \"\"\"\n",
    "    file = os.path.join(OUTPUT_DIR, filename)\n",
    "    df.to_csv(file, encoding = 'utf8')\n",
    "    print(f\"File saved successfully as {file}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "95c8b92b-4bde-4c06-b92f-86d31c24e1a0",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File saved successfully as output\\gold.csv\n"
     ]
    }
   ],
   "source": [
    "csv_output(gold_df, \"gold.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9345fd42-2426-4b4d-b7f6-e4e17fecb622",
   "metadata": {},
   "source": [
    "## Youtube Crawling"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f317a116-4373-4c53-82f1-64cc506d682f",
   "metadata": {},
   "source": [
    "Our search strategy consists of: \n",
    "\n",
    "- Searching YouTube for a specific Greek topic\n",
    "- Getting the links and video names from the search\n",
    "- For each link, crawl the comments for a set number of scrolling actions\n",
    "\n",
    "By using the YouTube search function we can guarantee that the crawled videos will be relevant (many comments in Greek-Greeklish) and popular (large number of comments).\n",
    "\n",
    "We will repeat this procedure twice for two distinct groups of videos: greek songs, since their comments are usually in more formal Greek, and Greek gaming videos where, as discussed above, because of the demographic Greeklish are more prevalent."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "fc364ba5-43ef-41dc-b47e-7b593ff4adcd",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Scrolling (0 out of max 5)...\n",
      "Scrolling (1 out of max 5)...\n",
      "Scrolling (2 out of max 5)...\n",
      "Scrolling (3 out of max 5)...\n",
      "Scrolling (4 out of max 5)...\n",
      "Scrolling (5 out of max 5)...\n"
     ]
    }
   ],
   "source": [
    "from tasks.youtube import extract_search_results, extract_comments\n",
    "\n",
    "\n",
    "# \"greek songs\" search in Greek\n",
    "song_search_url = \"https://www.youtube.com/results?search_query=%CE%B5%CE%BB%CE%BB%CE%B7%CE%BD%CE%B9%CE%BA%CE%B1+%CF%84%CF%81%CE%B1%CE%B3%CE%BF%CF%85%CE%B4%CE%B9%CE%B1\"\n",
    "search_soup = scrape_youtube(ChromeDriverManager.get(), song_search_url, max_scrolls=5, verbose=True)\n",
    "results_search_song = extract_search_results(search_soup)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "785dc0b6-77a8-4e32-b4f7-1cba15f70db5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Scrolling (0 out of max 5)...\n",
      "Scrolling (1 out of max 5)...\n",
      "Scrolling (2 out of max 5)...\n",
      "Scrolling (3 out of max 5)...\n",
      "Scrolling (4 out of max 5)...\n",
      "Scrolling (5 out of max 5)...\n"
     ]
    }
   ],
   "source": [
    "gaming_search_url = \"https://www.youtube.com/results?search_query=greek+fortnite\"\n",
    "gaming_soup = scrape_youtube(ChromeDriverManager.get(), gaming_search_url, max_scrolls=5, verbose=True)\n",
    "results_search_gaming = extract_search_results(gaming_soup)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "cccbdab7-02e4-4d4d-979d-7dcda9723c54",
   "metadata": {
    "tags": []
   },
   "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>title</th>\n",
       "      <th>link</th>\n",
       "      <th>source</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>\\n\\nGreek Hits 2023 | Non-Stop Mix by Elegant ...</td>\n",
       "      <td>/watch?v=RcSAggke-_U&amp;pp=ygUjzrXOu867zrfOvc65zr...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>\\n\\n00's GREEK MIX | KAPSOURA EDITION\\n</td>\n",
       "      <td>/watch?v=isCeE38TrXA&amp;pp=ygUjzrXOu867zrfOvc65zr...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>\\n\\nΜουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χ...</td>\n",
       "      <td>/watch?v=p5g82ta4sTk&amp;pp=ygUjzrXOu867zrfOvc65zr...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>\\n\\nThe Greek '90s Dance NonStopMix | OFFICIAL...</td>\n",
       "      <td>/watch?v=GEPvsn6JA_c&amp;pp=ygUjzrXOu867zrfOvc65zr...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>\\n\\nΤΑ ΛΑΪΚΑ ΤΗΣ ΤΑΒΕΡΝΑΣ | NON STOP MIX - Που...</td>\n",
       "      <td>/watch?v=C4f3xcZzr3s&amp;pp=ygUjzrXOu867zrfOvc65zr...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>262</th>\n",
       "      <td>\\n\\nΗ ΤΕΛΕΥΤΑΙΑ ΦΟΡΑ ΠΟΥ ΠΑΙΖΩ FORTNITE ONLY U...</td>\n",
       "      <td>/watch?v=KLBYBvxiTvQ&amp;pp=ygUOZ3JlZWsgZm9ydG5pdG...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>263</th>\n",
       "      <td>\\n\\n🔴 UNREAL STREAM Fortnite Greek Live Stream...</td>\n",
       "      <td>/watch?v=4lyI1hJlnKE&amp;pp=ygUOZ3JlZWsgZm9ydG5pdG...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>264</th>\n",
       "      <td>\\n\\nΟ ΧΑΜΕΝΟΣ ΤΡΩΕΙ ΜΙΑ ΑΠΑΙΣΙΑ PIZZA *ΜΠΛΙΑΧ*...</td>\n",
       "      <td>/watch?v=qFgEIRgj9pE&amp;pp=ygUOZ3JlZWsgZm9ydG5pdG...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>265</th>\n",
       "      <td>\\n\\nΔΟΚΙΜΑΖΩ ΤΟ CHAPTER 2 ΣΤΟ FORTNITE ONLY UP...</td>\n",
       "      <td>/watch?v=e7sawZlbL6c&amp;pp=ygUOZ3JlZWsgZm9ydG5pdG...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>266</th>\n",
       "      <td>\\n\\n😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με ...</td>\n",
       "      <td>/watch?v=6ay2HZwz2sA&amp;pp=ygUOZ3JlZWsgZm9ydG5pdG...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>267 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 title  \\\n",
       "0    \\n\\nGreek Hits 2023 | Non-Stop Mix by Elegant ...   \n",
       "1              \\n\\n00's GREEK MIX | KAPSOURA EDITION\\n   \n",
       "2    \\n\\nΜουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χ...   \n",
       "3    \\n\\nThe Greek '90s Dance NonStopMix | OFFICIAL...   \n",
       "4    \\n\\nΤΑ ΛΑΪΚΑ ΤΗΣ ΤΑΒΕΡΝΑΣ | NON STOP MIX - Που...   \n",
       "..                                                 ...   \n",
       "262  \\n\\nΗ ΤΕΛΕΥΤΑΙΑ ΦΟΡΑ ΠΟΥ ΠΑΙΖΩ FORTNITE ONLY U...   \n",
       "263  \\n\\n🔴 UNREAL STREAM Fortnite Greek Live Stream...   \n",
       "264  \\n\\nΟ ΧΑΜΕΝΟΣ ΤΡΩΕΙ ΜΙΑ ΑΠΑΙΣΙΑ PIZZA *ΜΠΛΙΑΧ*...   \n",
       "265  \\n\\nΔΟΚΙΜΑΖΩ ΤΟ CHAPTER 2 ΣΤΟ FORTNITE ONLY UP...   \n",
       "266  \\n\\n😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με ...   \n",
       "\n",
       "                                                  link  source  \n",
       "0    /watch?v=RcSAggke-_U&pp=ygUjzrXOu867zrfOvc65zr...    song  \n",
       "1    /watch?v=isCeE38TrXA&pp=ygUjzrXOu867zrfOvc65zr...    song  \n",
       "2    /watch?v=p5g82ta4sTk&pp=ygUjzrXOu867zrfOvc65zr...    song  \n",
       "3    /watch?v=GEPvsn6JA_c&pp=ygUjzrXOu867zrfOvc65zr...    song  \n",
       "4    /watch?v=C4f3xcZzr3s&pp=ygUjzrXOu867zrfOvc65zr...    song  \n",
       "..                                                 ...     ...  \n",
       "262  /watch?v=KLBYBvxiTvQ&pp=ygUOZ3JlZWsgZm9ydG5pdG...  gaming  \n",
       "263  /watch?v=4lyI1hJlnKE&pp=ygUOZ3JlZWsgZm9ydG5pdG...  gaming  \n",
       "264  /watch?v=qFgEIRgj9pE&pp=ygUOZ3JlZWsgZm9ydG5pdG...  gaming  \n",
       "265  /watch?v=e7sawZlbL6c&pp=ygUOZ3JlZWsgZm9ydG5pdG...  gaming  \n",
       "266  /watch?v=6ay2HZwz2sA&pp=ygUOZ3JlZWsgZm9ydG5pdG...  gaming  \n",
       "\n",
       "[267 rows x 3 columns]"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_df = pd.DataFrame({\"title\": results_search_song[0] + results_search_gaming[0],\n",
    "                           \"link\":  results_search_song[1] + results_search_gaming[1],\n",
    "                          \"source\": np.array(len(results_search_song[0])*[\"song\"] + len(results_search_gaming[0])*[\"gaming\"]) })\n",
    "results_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "33ad8596-1462-41d4-a81f-2c00bf2d7e39",
   "metadata": {
    "tags": []
   },
   "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>title</th>\n",
       "      <th>link</th>\n",
       "      <th>source</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Greek Hits 2023 | Non-Stop Mix by Elegant Gree...</td>\n",
       "      <td>https://www.youtube.com/watch?v=RcSAggke-_U&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>00's GREEK MIX | KAPSOURA EDITION</td>\n",
       "      <td>https://www.youtube.com/watch?v=isCeE38TrXA&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>The Greek '90s Dance NonStopMix | OFFICIAL Part 1</td>\n",
       "      <td>https://www.youtube.com/watch?v=GEPvsn6JA_c&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ΤΑ ΛΑΪΚΑ ΤΗΣ ΤΑΒΕΡΝΑΣ | NON STOP MIX - Πουλόπο...</td>\n",
       "      <td>https://www.youtube.com/watch?v=C4f3xcZzr3s&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>262</th>\n",
       "      <td>Η ΤΕΛΕΥΤΑΙΑ ΦΟΡΑ ΠΟΥ ΠΑΙΖΩ FORTNITE ONLY UP CH...</td>\n",
       "      <td>https://www.youtube.com/watch?v=KLBYBvxiTvQ&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>263</th>\n",
       "      <td>🔴 UNREAL STREAM Fortnite Greek Live Stream Now...</td>\n",
       "      <td>https://www.youtube.com/watch?v=4lyI1hJlnKE&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>264</th>\n",
       "      <td>Ο ΧΑΜΕΝΟΣ ΤΡΩΕΙ ΜΙΑ ΑΠΑΙΣΙΑ PIZZA *ΜΠΛΙΑΧ*  (F...</td>\n",
       "      <td>https://www.youtube.com/watch?v=qFgEIRgj9pE&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>265</th>\n",
       "      <td>ΔΟΚΙΜΑΖΩ ΤΟ CHAPTER 2 ΣΤΟ FORTNITE ONLY UP * R...</td>\n",
       "      <td>https://www.youtube.com/watch?v=e7sawZlbL6c&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>266</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>267 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 title  \\\n",
       "0    Greek Hits 2023 | Non-Stop Mix by Elegant Gree...   \n",
       "1                    00's GREEK MIX | KAPSOURA EDITION   \n",
       "2    Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "3    The Greek '90s Dance NonStopMix | OFFICIAL Part 1   \n",
       "4    ΤΑ ΛΑΪΚΑ ΤΗΣ ΤΑΒΕΡΝΑΣ | NON STOP MIX - Πουλόπο...   \n",
       "..                                                 ...   \n",
       "262  Η ΤΕΛΕΥΤΑΙΑ ΦΟΡΑ ΠΟΥ ΠΑΙΖΩ FORTNITE ONLY UP CH...   \n",
       "263  🔴 UNREAL STREAM Fortnite Greek Live Stream Now...   \n",
       "264  Ο ΧΑΜΕΝΟΣ ΤΡΩΕΙ ΜΙΑ ΑΠΑΙΣΙΑ PIZZA *ΜΠΛΙΑΧ*  (F...   \n",
       "265  ΔΟΚΙΜΑΖΩ ΤΟ CHAPTER 2 ΣΤΟ FORTNITE ONLY UP * R...   \n",
       "266  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "\n",
       "                                                  link  source  \n",
       "0    https://www.youtube.com/watch?v=RcSAggke-_U&pp...    song  \n",
       "1    https://www.youtube.com/watch?v=isCeE38TrXA&pp...    song  \n",
       "2    https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song  \n",
       "3    https://www.youtube.com/watch?v=GEPvsn6JA_c&pp...    song  \n",
       "4    https://www.youtube.com/watch?v=C4f3xcZzr3s&pp...    song  \n",
       "..                                                 ...     ...  \n",
       "262  https://www.youtube.com/watch?v=KLBYBvxiTvQ&pp...  gaming  \n",
       "263  https://www.youtube.com/watch?v=4lyI1hJlnKE&pp...  gaming  \n",
       "264  https://www.youtube.com/watch?v=qFgEIRgj9pE&pp...  gaming  \n",
       "265  https://www.youtube.com/watch?v=e7sawZlbL6c&pp...  gaming  \n",
       "266  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming  \n",
       "\n",
       "[267 rows x 3 columns]"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_df.title = results_df.title.apply(lambda x: x.strip())\n",
    "results_df.link = results_df.link.apply(lambda x: \"https://www.youtube.com\" + x) \n",
    "results_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7aacd6fc-66e9-445c-b125-9941bbce48d5",
   "metadata": {},
   "source": [
    "We will use our rules-based (Regex) classifier to once again filter the videos, so they only include titles in Greek. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "1ef13a6a-2c51-419d-ba62-a58a7fd7e36a",
   "metadata": {
    "tags": []
   },
   "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>title</th>\n",
       "      <th>link</th>\n",
       "      <th>source</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ΤΑ ΛΑΪΚΑ ΤΗΣ ΤΑΒΕΡΝΑΣ | NON STOP MIX - Πουλόπο...</td>\n",
       "      <td>https://www.youtube.com/watch?v=C4f3xcZzr3s&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Νατάσα Θεοδωρίδου - Αν πας με άλλη (AI Cover)</td>\n",
       "      <td>https://www.youtube.com/watch?v=HILSvOQV_bc&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Ελληνικά disco, 80s &amp; 90s (Non-stop Party Mix)...</td>\n",
       "      <td>https://www.youtube.com/watch?v=Y9rbyT0ZNq4&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>Βασίλης Καρράς - Τραγούδια Επιτυχίες | Vasilis...</td>\n",
       "      <td>https://www.youtube.com/watch?v=mQkIg8Rg3m4&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>261</th>\n",
       "      <td>🔴 Η ΣΕΖΟΝ 4 ΕΙΝΑΙ ΕΔΩ!!! 🔥 ΝΕΟ MAP, BATTLE PAS...</td>\n",
       "      <td>https://www.youtube.com/watch?v=SMVnHVPRm_Q&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>262</th>\n",
       "      <td>Η ΤΕΛΕΥΤΑΙΑ ΦΟΡΑ ΠΟΥ ΠΑΙΖΩ FORTNITE ONLY UP CH...</td>\n",
       "      <td>https://www.youtube.com/watch?v=KLBYBvxiTvQ&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>264</th>\n",
       "      <td>Ο ΧΑΜΕΝΟΣ ΤΡΩΕΙ ΜΙΑ ΑΠΑΙΣΙΑ PIZZA *ΜΠΛΙΑΧ*  (F...</td>\n",
       "      <td>https://www.youtube.com/watch?v=qFgEIRgj9pE&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>265</th>\n",
       "      <td>ΔΟΚΙΜΑΖΩ ΤΟ CHAPTER 2 ΣΤΟ FORTNITE ONLY UP * R...</td>\n",
       "      <td>https://www.youtube.com/watch?v=e7sawZlbL6c&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>266</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>238 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 title  \\\n",
       "2    Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "4    ΤΑ ΛΑΪΚΑ ΤΗΣ ΤΑΒΕΡΝΑΣ | NON STOP MIX - Πουλόπο...   \n",
       "6        Νατάσα Θεοδωρίδου - Αν πας με άλλη (AI Cover)   \n",
       "7    Ελληνικά disco, 80s & 90s (Non-stop Party Mix)...   \n",
       "8    Βασίλης Καρράς - Τραγούδια Επιτυχίες | Vasilis...   \n",
       "..                                                 ...   \n",
       "261  🔴 Η ΣΕΖΟΝ 4 ΕΙΝΑΙ ΕΔΩ!!! 🔥 ΝΕΟ MAP, BATTLE PAS...   \n",
       "262  Η ΤΕΛΕΥΤΑΙΑ ΦΟΡΑ ΠΟΥ ΠΑΙΖΩ FORTNITE ONLY UP CH...   \n",
       "264  Ο ΧΑΜΕΝΟΣ ΤΡΩΕΙ ΜΙΑ ΑΠΑΙΣΙΑ PIZZA *ΜΠΛΙΑΧ*  (F...   \n",
       "265  ΔΟΚΙΜΑΖΩ ΤΟ CHAPTER 2 ΣΤΟ FORTNITE ONLY UP * R...   \n",
       "266  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "\n",
       "                                                  link  source  \n",
       "2    https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song  \n",
       "4    https://www.youtube.com/watch?v=C4f3xcZzr3s&pp...    song  \n",
       "6    https://www.youtube.com/watch?v=HILSvOQV_bc&pp...    song  \n",
       "7    https://www.youtube.com/watch?v=Y9rbyT0ZNq4&pp...    song  \n",
       "8    https://www.youtube.com/watch?v=mQkIg8Rg3m4&pp...    song  \n",
       "..                                                 ...     ...  \n",
       "261  https://www.youtube.com/watch?v=SMVnHVPRm_Q&pp...  gaming  \n",
       "262  https://www.youtube.com/watch?v=KLBYBvxiTvQ&pp...  gaming  \n",
       "264  https://www.youtube.com/watch?v=qFgEIRgj9pE&pp...  gaming  \n",
       "265  https://www.youtube.com/watch?v=e7sawZlbL6c&pp...  gaming  \n",
       "266  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming  \n",
       "\n",
       "[238 rows x 3 columns]"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "greeklish_model = RegexClassifier(include_greeklish=True)\n",
    "preds = greeklish_model.predict(results_df.title)\n",
    "gr_res_df = results_df[(preds != \"en\") & (preds != None)]\n",
    "gr_res_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "63b99638-1668-48a1-9824-c302f78f968c",
   "metadata": {},
   "source": [
    "Below we begin the process of crawling, processing and packaging the crawled data into a unified dataframe:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "a24623c0-ce6d-4e3d-bf26-87b79cd2c390",
   "metadata": {},
   "outputs": [],
   "source": [
    "from selenium.common.exceptions import JavascriptException\n",
    "import bs4\n",
    "\n",
    "\n",
    "def scrape(urls: list[str]) -> list[tuple[str, bs4.BeautifulSoup]]:\n",
    "    \"\"\"\n",
    "    Scrape YouTube videos using the provided URLs.\n",
    "\n",
    "    :param urls: A list of YouTube video URLs to scrape.\n",
    "    :type urls: list[str]\n",
    "\n",
    "    :return: A list of tuples where each tuple contains the original URL and\n",
    "             the corresponding BeautifulSoup object containing the scraped data.\n",
    "    :rtype: list[tuple[str, bs4.BeautifulSoup]]\n",
    "\n",
    "    :Example:\n",
    "\n",
    "    .. code-block:: python\n",
    "\n",
    "        >>> scrape_results = scrape(['https://www.youtube.com/watch?v=example1', 'https://www.youtube.com/watch?v=example2'])\n",
    "        >>> print(scrape_results)\n",
    "        [('https://www.youtube.com/watch?v=example1', <BeautifulSoup object>), ('https://www.youtube.com/watch?v=example2', <BeautifulSoup object>)]\n",
    "    \"\"\"\n",
    "    scrape_results = []\n",
    "    \n",
    "    print(\"Scraping videos...\")\n",
    "    for url in tqdm(urls):\n",
    "        try:\n",
    "            scrape_results.append((url, scrape_youtube(ChromeDriverManager.get(), url, max_scrolls=10, scroll_wait_secs=1.3, verbose=False)))\n",
    "        except JavascriptException:\n",
    "            continue\n",
    "        except Exception as e:\n",
    "            print(e)\n",
    "            continue  \n",
    "    return scrape_results\n",
    "\n",
    "\n",
    "def process_scraped(scrape_results: list[tuple[str, bs4.BeautifulSoup]]) -> pd.DataFrame:\n",
    "    \"\"\"\n",
    "    Process the scraped YouTube video data to extract comments and dates.\n",
    "\n",
    "    :param scrape_results: A list of tuples where each tuple contains the original URL\n",
    "                          and the corresponding BeautifulSoup object containing the scraped data.\n",
    "    :type scrape_results: list[tuple[str, bs4.BeautifulSoup]]\n",
    "\n",
    "    :return: A pandas DataFrame containing the processed data with columns 'link', 'text', and 'date'.\n",
    "    :rtype: pd.DataFrame\n",
    "\n",
    "    :Example:\n",
    "\n",
    "    .. code-block:: python\n",
    "\n",
    "        >>> processed_data = process_scraped([('https://www.youtube.com/watch?v=example1', <BeautifulSoup object>), ('https://www.youtube.com/watch?v=example2', <BeautifulSoup object>)])\n",
    "        >>> print(processed_data)\n",
    "                    link                 text        date\n",
    "        0  https://www.youtube.com/watch?v=example1  Comment 1  2023-01-01\n",
    "        1  https://www.youtube.com/watch?v=example1  Comment 2  2023-01-02\n",
    "        2  https://www.youtube.com/watch?v=example2  Comment 3  2023-01-03\n",
    "    \"\"\"\n",
    "    scraped_urls = []\n",
    "    comments = []\n",
    "    dates = []\n",
    "    \n",
    "    print(\"Processing comments...\")\n",
    "    print(type(scrape_results[0]))\n",
    "    for url, result in tqdm(scrape_results):\n",
    "        if result is not None:\n",
    "            new_comments, new_dates = extract_comments(result)\n",
    "            comments += new_comments\n",
    "            dates += new_dates\n",
    "            scraped_urls += ([url] * len(new_comments))\n",
    "    return pd.DataFrame({\"link\": scraped_urls, \"text\": comments, \"date\": dates})\n",
    "\n",
    "\n",
    "def filter_comments(df: pd.DataFrame) -> pd.DataFrame:\n",
    "    preds = greeklish_model.predict(df.text)\n",
    "    mask = ((preds != \"el\") & (preds != \"greeklish\"))\n",
    "    return comments_df[mask]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "061846d6-f0dd-4eda-9b78-425a4f3acbbb",
   "metadata": {},
   "source": [
    "We will crawl a maximum of 150 videos, using a random uniform mix of both video groups. \n",
    "\n",
    "Initially multi-threading was considered but the processes proved too demanding on main memory and CPU resources. Thus, we run the crawl in a single thread."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "a24c1a4b-1b63-437f-8480-065b30d872a1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Scraping videos...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████| 150/150 [34:02<00:00, 13.62s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing comments...\n",
      "<class 'tuple'>\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████| 135/135 [01:06<00:00,  2.02it/s]\n"
     ]
    }
   ],
   "source": [
    "max_videos = 150\n",
    "urls = gr_res_df.link.sample(max_videos) if len(gr_res_df.link) > max_videos else gr_res_df.link\n",
    "\n",
    "scraped = scrape(urls)\n",
    "comments_df = process_scraped(scraped)\n",
    "\n",
    "comments_df.date = comments_df.date.apply(lambda x: x.date() if x is not None else None)\n",
    "comments_df = filter_comments(comments_df)\n",
    "crawl_df = pd.merge(gr_res_df, comments_df, how=\"inner\", on=\"link\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "493f60af-eac6-4099-914c-0a81658750a0",
   "metadata": {},
   "source": [
    "We again clear the dataset of empty comments and other anomalies:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "04eaef1b-837b-423c-a261-ceec238d04d0",
   "metadata": {
    "scrolled": true
   },
   "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>title</th>\n",
       "      <th>link</th>\n",
       "      <th>source</th>\n",
       "      <th>text</th>\n",
       "      <th>date</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>ANTIGUAS CANCIONES DE GRECIA PAIS NATAL DE MIS...</td>\n",
       "      <td>2022-11-29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>Încă  o zi petrecută cu muzica voastră  fantas...</td>\n",
       "      <td>2022-11-29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>FelicităriSuperb \\nSă fiți mereu bine \\nMomen...</td>\n",
       "      <td>2022-11-29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>Lovely collection! Thank you &lt;3</td>\n",
       "      <td>2022-11-29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>Beautiful thank you  so many memories awesome ...</td>\n",
       "      <td>2022-11-29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2692</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Copy by fantaros</td>\n",
       "      <td>2021-11-29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2693</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Fist</td>\n",
       "      <td>2021-11-29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2694</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Yui</td>\n",
       "      <td>2022-11-29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2695</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Uihhii</td>\n",
       "      <td>2022-11-29</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2696</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Zzfitdt</td>\n",
       "      <td>2021-11-29</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2359 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  title  \\\n",
       "0     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "1     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "2     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "3     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "4     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "...                                                 ...   \n",
       "2692  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2693  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2694  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2695  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2696  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "\n",
       "                                                   link  source  \\\n",
       "0     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "1     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "2     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "3     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "4     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "...                                                 ...     ...   \n",
       "2692  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2693  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2694  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2695  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2696  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "\n",
       "                                                   text        date  \n",
       "0     ANTIGUAS CANCIONES DE GRECIA PAIS NATAL DE MIS...  2022-11-29  \n",
       "1     Încă  o zi petrecută cu muzica voastră  fantas...  2022-11-29  \n",
       "2      FelicităriSuperb \\nSă fiți mereu bine \\nMomen...  2022-11-29  \n",
       "3                       Lovely collection! Thank you <3  2022-11-29  \n",
       "4     Beautiful thank you  so many memories awesome ...  2022-11-29  \n",
       "...                                                 ...         ...  \n",
       "2692                                   Copy by fantaros  2021-11-29  \n",
       "2693                                               Fist  2021-11-29  \n",
       "2694                                                Yui  2022-11-29  \n",
       "2695                                             Uihhii  2022-11-29  \n",
       "2696                                            Zzfitdt  2021-11-29  \n",
       "\n",
       "[2359 rows x 5 columns]"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "crawl_df = crawl_df.dropna()\n",
    "crawl_df = crawl_df[~crawl_df.text.apply(lambda x: len(x.strip())==0)]\n",
    "crawl_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bffa511d-05ac-4682-b08d-c7ad10ba0406",
   "metadata": {},
   "source": [
    "And close the driver we used for our crawling to relinquish the resources we obtained from the OS."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "a10c703c-74d6-4d27-a22f-58a7d3d0a04b",
   "metadata": {},
   "outputs": [],
   "source": [
    "ChromeDriverManager.quit()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cf0324f7-ccbc-4157-983b-441cbb1c83c6",
   "metadata": {},
   "source": [
    "## Language Identification"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "44dea405-d131-49a8-9894-22cea73e6986",
   "metadata": {},
   "source": [
    "In this part, we create our optimal ML classifier for the language identification task we outlined in Part 1."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "da930721-08a8-46ab-8137-a4064578d493",
   "metadata": {},
   "source": [
    "We will use the `gold.csv` dataset we built in Part 1 to create a train, test and validation split for our models.\n",
    "\n",
    "We use stratified sampling, since our datasets are wildly unbalanced, especially in respect to Greeklish. This will ensure an adequate number of Greeklish posts will be included in all splits.\n",
    "\n",
    "Given this imbalance in fact, we are tempted to use undersampling or oversampling in order to prevent our classifiers from simply ignoring the class. This however will likely not aid us significantly, given that most models (such as Logistic Regression `[3]`) are robust to class imbalance `[4]`.\n",
    "\n",
    "\n",
    "```\n",
    "[3] Gary King and Langche Zeng. 2001. “Logistic Regression in Rare Events Data.” Political Analysis, 9, Pp. 137–163\n",
    "\n",
    "[4] Stephan Kolassa (https://stats.stackexchange.com/users/1352/stephan-kolassa), Are unbalanced datasets problematic, and (how) does oversampling (purport to) help?, URL (version: 2023-11-14): https://stats.stackexchange.com/q/357466\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0b65412b-d28e-46c3-ae34-204a0bd592ad",
   "metadata": {},
   "source": [
    "We will use a train-test-validation split of 70%-10%-20%. We don't expect a great need for validation since we will be using relatively simple ML models. Additionally, the difficulty presented by the small sample size of Greeklish comments means we must focus as much data as possible in making sure our classifiers can understand Greeklish in the first place."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "852afc7e-a026-4457-9cfe-a451da710b32",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from src.ml import train_test_val_split\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "data_train, data_val, data_test = train_test_val_split(gold_df, train_ratio=0.7, val_ratio=0.1, test_ratio=0.2,\n",
    "                                                        random_state=42, stratify_col=\"labels\")\n",
    "# code block from Ioannis Pavlopoulos\n",
    "axes = pd.DataFrame({\"train\": data_train.labels.value_counts(),\n",
    "              \"validation\": data_val.labels.value_counts(),\n",
    "              \"test\": data_test.labels.value_counts()}\n",
    "            ).plot.pie(subplots=True, \n",
    "                       textprops={'fontsize': 8},\n",
    "                       autopct=f'%1.1f%%', # print percent% results \n",
    "                       legend=False)\n",
    "\n",
    "axes[0].set_xlabel(data_train.shape[0])\n",
    "axes[1].set_xlabel(data_val.shape[0])\n",
    "axes[2].set_xlabel(data_test.shape[0])\n",
    "\n",
    "axes[1].text(0, -2, 'Number of records', ha='center')\n",
    "\n",
    "plt.title(\"Train-Test-Val split for the Language Identification Dataset\")\n",
    "plt.tight_layout(pad=2.0)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e7954000-3144-4632-831c-055a561e6838",
   "metadata": {},
   "source": [
    "We will encode our data as TF-IDF vectors. Count vectors could also work for this specific problem, but the computational cost of TF-IDF is minimal compared to acquiring the data fitting our classifiers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "0a182208-c3db-4892-8617-2209629396fd",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "\n",
    "\n",
    "vectorizer = TfidfVectorizer().fit(data_train.text)\n",
    "x_train = vectorizer.transform(data_train.text)\n",
    "y_train = data_train.labels\n",
    "x_val = vectorizer.transform(data_val.text)\n",
    "y_val = data_val.labels\n",
    "x_test = vectorizer.transform(data_test.text)\n",
    "y_test = data_test.labels"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "91d5fa31-d0eb-402c-b1b8-8836067179f3",
   "metadata": {},
   "source": [
    "The metric we will be using is Macro-F1 average.\n",
    "\n",
    "- **F1** is a metric used to balance the need for making sure our classifications for a category are both correct (precision) and represent as many of the actual cases of the category as possible (recall).\n",
    "- **Macro-F1** is the unweighted average of all F1 metrics for each class. We choose Macro F1 instead of a weighted average because\n",
    "    * We have an unbalanced dataset (Greeklish data are a small fraction of overall data)\n",
    "    * We are much more interested in the small classes (here Greeklish)\n",
    "\n",
    "Thus, we want to use a metric which favors both thorough and precise classifiers, and which also assigns equal importance to our smaller classes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "586c604d-1467-498c-9b04-57b00bf5db71",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import cross_val_score\n",
    "\n",
    "\n",
    "def cross_val_res(model, x, y, scoring=None, cv=10):\n",
    "    \"\"\"\n",
    "    Minor utility method, wraping cross_val_score.\n",
    "    \"\"\"\n",
    "    if scoring is None:\n",
    "        scoring = \"f1_macro\"\n",
    "    res = cross_val_score(model, x, y, cv=cv, scoring=scoring)\n",
    "    return res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "a52c1a13-bf6a-4de9-a281-c43996874a5b",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import f1_score\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.metrics import ConfusionMatrixDisplay\n",
    "import matplotlib.pyplot as plt\n",
    "import warnings\n",
    "\n",
    "\n",
    "def get_statistics(y_test, y_pred):\n",
    "    \"\"\"\n",
    "    Minor utility method printing average Macro F1 score and classification report\n",
    "    as well as displaying the classifier's Confusion Matrix.\n",
    "    \"\"\"\n",
    "    with warnings.catch_warnings():\n",
    "        warnings.simplefilter(\"ignore\")\n",
    "\n",
    "        print(f\"Macro F1: {f1_score(y_test, y_pred, average='macro', zero_division=0)}\")\n",
    "        print(classification_report(y_test, y_pred, zero_division=0))\n",
    "        ConfusionMatrixDisplay.from_predictions(y_test, \n",
    "                                                y_pred, \n",
    "                                                colorbar=True)\n",
    "        plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c98930e1-54ab-4543-880f-1ee8b43c0f67",
   "metadata": {},
   "source": [
    "For the Language Identification task the following models are considered:\n",
    "\n",
    "- The previously implemented rules-based (Regex) model\n",
    "- Naive Bayes\n",
    "- Logistic Regression\n",
    "- Random Forest\n",
    "- Adaboost Model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "10e77140-ffb9-41d2-9984-5764179617b3",
   "metadata": {},
   "source": [
    "### Dummy Classifier"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9a5112eb-7cf0-4e03-8206-643a444e7a16",
   "metadata": {},
   "source": [
    "We will first run a \"fake\" classifier which only guesses the majority category.\n",
    "\n",
    "This dummy model thus completely disregards the input features and serves as a useful baseline with which to compare the subsequent classifiers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "fc35cfee-456b-4ebb-a6bb-f91d7b8aa0c2",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.14709302325581394\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "          el       0.42      1.00      0.59      1012\n",
      "          en       0.00      0.00      0.00       942\n",
      "   greeklish       0.00      0.00      0.00        79\n",
      "       other       0.00      0.00      0.00       395\n",
      "\n",
      "    accuracy                           0.42      2428\n",
      "   macro avg       0.10      0.25      0.15      2428\n",
      "weighted avg       0.17      0.42      0.25      2428\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.dummy import DummyClassifier\n",
    "\n",
    "\n",
    "majority = DummyClassifier(strategy=\"most_frequent\")\n",
    "majority.fit(x_train, y_train)\n",
    "majority_res = majority.predict(x_test)\n",
    "\n",
    "get_statistics(y_test, majority_res)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eb0b9c1a-4593-4890-9090-46b31dcc7866",
   "metadata": {},
   "source": [
    "### Regex Classification"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "66b8748b-d850-473f-80be-3b095c07ca8e",
   "metadata": {},
   "source": [
    "Let's now compare our rules-based classifier with the baseline. This classifier does use the input features to make decisions, but in a very simple and naive way. It also does not benefit from any information that could be gained through training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "0eb47656-a440-41c2-ae24-175969c76e06",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Regex Classifier mean macro F1: 0.5690, std: 0.5614\n"
     ]
    }
   ],
   "source": [
    "with warnings.catch_warnings():\n",
    "    warnings.simplefilter(\"ignore\")\n",
    "\n",
    "    regex_model = RegexClassifier()\n",
    "    res = cross_val_res(regex_model, data_train.text, y_train, scoring=\"f1_macro\")\n",
    "    print(f\"Regex Classifier mean macro F1: {res[0]:.4f}, std: {res[1]:.4f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "cdfa7f60-198c-4860-90a0-8f52e5b6151c",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.5427840052990762\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "          el       1.00      0.69      0.82      1012\n",
      "          en       0.74      1.00      0.85       942\n",
      "   greeklish       0.00      0.00      0.00        79\n",
      "       other       0.98      0.34      0.51       395\n",
      "\n",
      "    accuracy                           0.73      2428\n",
      "   macro avg       0.68      0.51      0.54      2428\n",
      "weighted avg       0.86      0.73      0.75      2428\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "regex_model = RegexClassifier(include_greeklish=True).fit(data_train.text, y_train)\n",
    "regex_res = regex_model.predict(data_test.text)\n",
    "get_statistics(y_test, regex_res)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ff92ea38-afd4-4799-9d73-125ad46d0b39",
   "metadata": {},
   "source": [
    "We notice that this classifiers performs relatively well at classifiying Greek and English, but is generally easily confused and can not catch Greeklish at all."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d85fd2ac-c5de-4721-baf7-c52a5c286364",
   "metadata": {},
   "source": [
    "### Naive Bayes"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a855d45e-e89c-4a8a-acf4-caa6c72b6438",
   "metadata": {},
   "source": [
    "Naive Bayes is a very cheap and easy-to-interpret classifier, which checks for the probability that each individual word in the text will belong in any language. We generally want to use the simplest model for the job, and so we start with this reliable model which has proven itself in many fields in the past.\n",
    "\n",
    "The `sklearn` library gives us access to many variations of Naive Bayes, each specialized in its own field. For this NLP task, we will be using `MultinomialNB`, which was suggested by [this blogpost](https://towardsdatascience.com/naive-bayes-classifiers-for-text-classification-be0d133d35ba)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "833bff3c-0a75-4f58-bafb-3cd28cdb4e1c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Naive Bayes mean macro F1-score 0.7848, std: 0.8102\n"
     ]
    }
   ],
   "source": [
    "from sklearn.naive_bayes import MultinomialNB\n",
    "\n",
    "# naive bayes needs dense arrays to work \n",
    "naive_x_train = x_train.toarray()\n",
    "naive_x_test = x_test.toarray()\n",
    "\n",
    "naive_model = MultinomialNB()\n",
    "res = cross_val_res(naive_model, naive_x_train, y_train, cv=5)\n",
    "print(f\"Naive Bayes mean macro F1-score {res[0]:.4f}, std: {res[1]:.4f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "f1f56e9a-1573-4229-8746-5eb7aed053b3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.8651470835475744\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "          el       0.92      1.00      0.96      1012\n",
      "          en       0.96      1.00      0.98       942\n",
      "   greeklish       1.00      0.47      0.64        79\n",
      "       other       0.99      0.80      0.88       395\n",
      "\n",
      "    accuracy                           0.95      2428\n",
      "   macro avg       0.97      0.82      0.87      2428\n",
      "weighted avg       0.95      0.95      0.94      2428\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "naive_model = MultinomialNB().fit(naive_x_train, y_train)\n",
    "naive_res = naive_model.predict(naive_x_test)\n",
    "\n",
    "get_statistics(y_test, naive_res)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2264438c-cc3d-4e33-a825-bd883286b824",
   "metadata": {},
   "source": [
    "Compared to our rules-based classifier, this is a great step-up. Classification of Greek and English is very good, and the \"other\" languages are reliably identified. \n",
    "\n",
    "Greeklish however are not reliably caught. The classification report states that when the classifier guesses Greeklish, it is always correct, but most Greeklish comments are confused with either English or Greek. Thus, the problem of distinguishing these three categories probably requires a more complex model."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "059ecd9d-3525-4f86-a102-5117125ddf5f",
   "metadata": {},
   "source": [
    "### Logistic Regression"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f14ebf97-2488-464a-b872-c5f0379325f6",
   "metadata": {},
   "source": [
    "LogisticRegression despite its name is a linear classifier, meaning that it attempts to linearly separate the data into distinct categories. This interpretation does not apply well to a NLP task, but means that the classifier retains some very useful properties:\n",
    "\n",
    "- The solution we get is a global optimum, meaning that it's the best we can get with the provided data. This means no hyper-parameter tuning is necessary and we can use the classifier as-is.\n",
    "- It's a simple and very easy to compute classifier, since it solves a (mathematically simple) linear problem, albeit with some restrictions (technically those restrictions force it to use gradient descent, but the calculations are much easier than say, a neural network)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "51d879dd-7534-4cb6-b887-17bc89d3cb4c",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Logistic Regression mean macro F1-score 0.9042, std: 0.8825\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "\n",
    "with warnings.catch_warnings():\n",
    "    # ignore warnings about deprecated methods in libraries\n",
    "    warnings.simplefilter(\"ignore\")\n",
    "\n",
    "    lr = LogisticRegression(max_iter=1000)\n",
    "    res = cross_val_res(lr, x_train, y_train)\n",
    "    print(f\"Logistic Regression mean macro F1-score {res[0]:.4f}, std: {res[1]:.4f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "0fc257a8-a52f-4f5b-832d-2e6e8d57f112",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.9243052241829711\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "          el       0.99      0.98      0.99      1012\n",
      "          en       0.99      0.99      0.99       942\n",
      "   greeklish       0.98      0.63      0.77        79\n",
      "       other       0.90      1.00      0.95       395\n",
      "\n",
      "    accuracy                           0.98      2428\n",
      "   macro avg       0.97      0.90      0.92      2428\n",
      "weighted avg       0.98      0.98      0.98      2428\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjEAAAGwCAYAAABYazQUAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy81sbWrAAAACXBIWXMAAA9hAAAPYQGoP6dpAABR2UlEQVR4nO3deVhU9f4H8PcAw7AOq6wCgoK7yaLmvmGWWnotNzQzySU1wg01NdFU0hItKzXvTbjmUjfbNc3ccskF3DfcEEFFUNlkG5g5vz/4OTUiOjDDjId5v57nPI9zzvd75nMOI/Phux2JIAgCiIiIiETGzNgBEBEREdUEkxgiIiISJSYxREREJEpMYoiIiEiUmMQQERGRKDGJISIiIlFiEkNERESiZGHsAKgylUqFW7duwd7eHhKJxNjhEBFRNQmCgIKCAnh5ecHMrPbaC0pKSqBQKHQ+j6WlJaysrPQQkWExiXkG3bp1Cz4+PsYOg4iIdJSeno769evXyrlLSkrg72eHzCylzufy8PBAamqq6BIZJjHPIHt7ewBA2vEGkNuxx88Q/hXU0tghENUqiUxm7BBMSrlQhv2KH9S/z2uDQqFAZpYSackNILev+XdFfoEKfqHXoVAomMSQ7h52IcntzHT6YJL2LCRSY4dAVKsk/IwbhSGGBNjZS2BnX/P3UUG8wxaYxBAREYmYUlBBqcNTEJWCSn/BGBiTGCIiIhFTQYAKNc9idKlrbOyrICIiIlFiSwwREZGIqaCCLh1CutU2LiYxREREIqYUBCiFmncJ6VLX2NidRERERKLElhgiIiIRM+WBvUxiiIiIREwFAUoTTWLYnURERESixJYYIiIiEWN3EhEREYkSZycRERERiQxbYoiIiERM9f+bLvXFikkMERGRiCl1nJ2kS11jYxJDREQkYkoBOj7FWn+xGBrHxBAREZEosSWGiIhIxDgmhoiIiERJBQmUkOhUX6zYnURERESixJYYIiIiEVMJFZsu9cWKSQwREZGIKXXsTtKlrrGxO4mIiIhEiS0xREREImbKLTFMYoiIiERMJUigEnSYnaRDXWNjdxIRERGJEltiiIiIRIzdSURERCRKSphBqUPHilKPsRgakxgiIiIRE3QcEyNwTAwRERGRYbElhoiISMQ4JoaIiIhESSmYQSnoMCZGxI8dYHcSERERiRJbYoiIiERMBQlUOrRJqCDephgmMURERCJmymNi2J1EREREosSWGCIiIhHTfWAvu5OIiIjICCrGxOjwAEh2JxEREREZFlti6KmKHpghcaknDv3mgNx7FmjYvBhvf5CBxq2LAQA52Rb4zyIvJO+zR2GeOVo8/wATF2bAO0ChPoeiVIK1C7yw90cnlJZIENzpASbFZaCeV5mxLkv0+r1xF4PezoazWxnSLllh9fteOHvUzthh1Vkt2j3AoAnZCGxZBBePcsSOboC/tjsYO6w6Ycjbt9Cxdw7qNyyGosQM54/b4aslPsi4Zq0u07H3ffSJyEKjFkVwcC7HhD7Nce2CrRGjfnaodHx2kphnJ7ElxgD27t0LiUSC3NxcY4dSI8un+uD4n3aIWZmG1bsuIrRrAWYOaYS7t6UQBGD+aH/cTrNE7Lpr+Pz3FLjXV2DmkEYoKfr747V6njcObXfArFXXEf/jFRQXmeH9kQFQivnJY0bU9ZUcjJ9/C5s+dcOEF4Jw9ogtFm5IRT1vxdMrU41Y2ahw7ZwVPp/tbexQ6pyW7Qrwy3o3TB7YDLNGNoG5uYBF/02BzPrvXxBWNiqcS7LHuqX1jRjps+nhmBhdNrESb+RkEKXFEhzY5oi35txGy+cL4e2vwOvTMuHho8Cv/3XBzWsyXEi2xTsfVrTM+DQqxaS4DBQXmWHPD44AgMJ8M+zY5Iwx799CSJcHaNSyGDNWpuH6RSuc2G9v3AsUqYFj72LHJmds3+iC9CtWWD3PG9m3pOg38p6xQ6uzkvbIkbjUEwd/czR2KHXOnFGNsXNLPaRdtkHqBRvExwTA3VuBwJaF6jK7fnDFxpXeOHGArV+PUsFM502sxBs5GYRSKYFKKYGlTKWxX2atwrmjdihTVAwI++dxc3NAKhVw7lhF18bl0zYoLzNDaNcCdRkXj3L4NSnB+WNsDq4uC6kKga2KkLxPMwFM3mePZmGFVdQiEg8b+4oWmIJcjnigJ2MSoyeCIGDp0qUICAiAtbU1nnvuOXz33Xda1S0tLUV+fr7G9qywsVOhaWghNq7wwL1MCyiVwK4tTrh43Ab371jAp1EJ3Osr8FWcJwpyzVGmkOCblW64nyXF/TsVv4DuZ1lAaqmCvaNm35GTaxlysvlLqrrkzkqYWwC5dzXvXW62BZzcyo0UFZG+CBg35wbOHrND2iUbYwcjCkpBovMmVvwG0ZM5c+bg+++/x6pVqxAYGIg///wTI0aMQL169Z5aNy4uDvPnzzdAlDUTszIN8VN8ERHSAmbmAhq1LEL3f+XgyhkbWEiBuf9ORfwUX7zWrCXMzAUEdy5Amx5PT8QEQQIRz+wzukeXdpBIABGPzyMCAExckAb/JkWYOqiZsUMRDaWOA3uVIv7FwSRGDwoLCxEfH4/du3ejffv2AICAgAAcOHAAa9aswdixY59Yf9asWZgyZYr6dX5+Pnx8fGo15urwaqDAx99fQUmRGQoLzODiXo5F4/zg4VsKAAhsVYxVf6SgMN8MZWUSOLooEdU3EEGtigAAzm7lKFOYoSDXXKM1JveeBbs/aiD/vjmU5YBTPc1WFwfXcrZskai9HXsdz/fMxbQhTXE309LY4ZAIsDtJD86fP4+SkhL06tULdnZ26u2///0vrl69+tT6MpkMcrlcY3sWWdmo4OJejoJccyTvk6N9b83WFlu5Co4uSty8ZonLp2zUxwNbFcFCqsLxP/8ew3HvjgXSLlqhWRsmMdVVXmaGy6dtENKlQGN/SJcCnE/iGCMSIwET5l9Hx945mDG8Ce5kyIwdkKioBDOdN7Hin216oFJVDGrdunUrvL01p1/KZDKtEplnWdJeewgC4NOwFDdTLfHvD7xRv2EJXhhSMRPmz18c4OCihJu3AqkXrLD6/fpo/2IeQrtVfMnaylXoPew+vpzvBblTOewdlVj7gRcaNClBcOeCJ701VeH7L10x/dN0XDptjQtJtugz4h7cvMuw9b8uxg6tzrKyUcLL/+8p7B4+CgQ0L0ZBrjmyb7LVQBcTF6She/97mD82EMUPzODkWnGfCwssoCit+IK1cyiHm1cpXNwr1paqH1ACAMjJliLnrmnff3YnkU6aNWsGmUyGGzduoGvXrpWOiz2JKcw3x7o4T9y9LYW9oxId++TizZm3YSGtOH7/jhRrYr2Re9cCzm7lCB90HxHRdzTOMT72ZsXaD+MbQFFshtadCjA/8RrMzY1wQXXAvp+dYO+kxPDJd+DsVo60FCvMGeGPLH6Z1pqg54rx0Za//y+Pn38LAPD7N05YNtnXWGHVCS+/ngUA+GjzRY39y6b5Y+eWinGF7cNzMPXjVPWx9z6r+Fl8vcILX3/CtWNMlUQQRPzkp2fInDlzsHr1aixbtgydOnVCfn4+Dh06BDs7O/j5+aF79+7IycmBo6PjU8+Vn58PBwcH5FwKgNxevM18YtLbq7WxQyCqVRIZu2gMqVwow57Sb5GXl1drQwQeflesOR4Ka7uat0kUPyjHuJDkWo21trAlRk8++OADuLm5IS4uDteuXYOjoyNCQkLw3nvvqbubiIiI9E3XBevEvNgdkxg9kUgkiIqKQlRU1GOPs8GLiIhIv5jEEBERiZiuzz8S87OTmMQQERGJmAoSqHRYOVSXusbGJIaIiEjETLklRryRExERkUljSwwREZGI6b7YnXjbM5jEEBERiZhKkEClw5OodalrbOJNv4iIiMiksSWGiIhIxFQ6didxsTsiIiIyCl2fRC3mp1iLN3IiIiIyaWyJISIiEjElJFDqsGCdLnWNjUkMERGRiLE7iYiIiEhkmMQQERGJmBJ/dynVbKue8vJyzJkzB/7+/rC2tkZAQAAWLFgAlUqlLiMIAmJjY+Hl5QVra2t069YN586d0zhPaWkp3nnnHbi6usLW1havvPIKMjIyqhULkxgiIiIRe9idpMtWHUuWLMHq1avx2Wef4cKFC1i6dCk++ugjrFy5Ul1m6dKliI+Px2effYZjx47Bw8MDvXr1QkFBgbpMdHQ0fvjhB2zevBkHDhzAgwcP0K9fPyiV2qdVHBNDREQkYoZ+AORff/2F/v37o2/fvgCABg0aYNOmTUhKSgJQ0QqzYsUKzJ49GwMHDgQAJCYmwt3dHRs3bsS4ceOQl5eH//znP1i/fj3Cw8MBAF9//TV8fHzwxx9/oHfv3lrFwpYYIiIiQn5+vsZWWlr62HKdOnXCrl27cOnSJQDAqVOncODAAfTp0wcAkJqaiszMTLzwwgvqOjKZDF27dsWhQ4cAAMnJySgrK9Mo4+XlhRYtWqjLaIMtMURERCImQAKVDtOkhf+v6+Pjo7F/3rx5iI2NrVR+xowZyMvLQ5MmTWBubg6lUolFixZh2LBhAIDMzEwAgLu7u0Y9d3d3pKWlqctYWlrCycmpUpmH9bXBJIaIiEjE9NWdlJ6eDrlcrt4vk8keW/6bb77B119/jY0bN6J58+Y4efIkoqOj4eXlhTfeeENdTiLRTKwEQai071HalPknJjFEREQEuVyukcRUZfr06Zg5cyaGDh0KAGjZsiXS0tIQFxeHN954Ax4eHgAqWls8PT3V9bKystStMx4eHlAoFMjJydFojcnKykKHDh20jpljYoiIiERMJUh03qqjqKgIZmaa6YO5ubl6irW/vz88PDywc+dO9XGFQoF9+/apE5TQ0FBIpVKNMrdv38bZs2erlcSwJYaIiEjElDo+xbq6dV9++WUsWrQIvr6+aN68OU6cOIH4+HiMHj0aQEU3UnR0NBYvXozAwEAEBgZi8eLFsLGxQUREBADAwcEBkZGRmDp1KlxcXODs7Ixp06ahZcuW6tlK2mASQ0RERFpbuXIl5s6diwkTJiArKwteXl4YN24c3n//fXWZmJgYFBcXY8KECcjJyUG7du3w+++/w97eXl1m+fLlsLCwwODBg1FcXIyePXsiISEB5ubmWsciEQRB0OvVkc7y8/Ph4OCAnEsBkNuzx88Qenu1NnYIRLVKUsUgTaod5UIZ9pR+i7y8PK3GmdTEw++KqAP9IbOT1vg8pQ/K8Gmnn2o11trClhgiIiIRU8EMKh26k3Spa2zijZyIiIhMGltiiIiIREwpSKCs5gyjR+uLFZMYIiIiEavJNOlH64sVkxgiIiIRE2rwJOpH64uVeCMnIiIik8aWGCIiIhFTQgKlDg+A1KWusTGJISIiEjGVoNu4FpWIV4tjdxIRERGJEltiiIiIREyl48BeXeoaG5MYIiIiEVNBApUO41p0qWts4k2/iIiIyKSxJYaIiEjEuGIvERERiRLHxNAz6V9BLWEhqfnj1Ul7q9MOGDsEkzPer5OxQzApgkJh7BBMiiCUGTsEk8AkhoiISMRU0PHZSSIe2MskhoiISMQEHWcnCUxiiIiIyBhM+SnW4h3NQ0RERCaNLTFEREQixtlJREREJErsTiIiIiISGbbEEBERiZgpPzuJSQwREZGIsTuJiIiISGTYEkNERCRiptwSwySGiIhIxEw5iWF3EhEREYkSW2KIiIhEzJRbYpjEEBERiZgA3aZJC/oLxeCYxBAREYmYKbfEcEwMERERiRJbYoiIiETMlFtimMQQERGJmCknMexOIiIiIlFiSwwREZGImXJLDJMYIiIiERMECQQdEhFd6hobu5OIiIhIlNgSQ0REJGIqSHRa7E6XusbGJIaIiEjETHlMDLuTiIiISJTYEkNERCRipjywl0kMERGRiJlydxKTGCIiIhEz5ZYYjokhIiIiUWJLDBERkYgJOnYnibklhkkMERGRiAkABEG3+mLF7iQiIiISJbbEEBERiZgKEki4Yi8RERGJDWcnEREREYkMW2KIiIhETCVIIOFid0RERCQ2gqDj7CQRT09idxIRERGJEltiiIiIRMyUB/YyiSEiIhIxJjFEetDvjbsY9HY2nN3KkHbJCqvf98LZo3bGDkuUSh6Y4+dlvji5wwUFd6XwaV6IwbHX0OC5BwCAX5b7IukXV+TcksFCKsC35QP0n34d/sEVx++myzCnU5vHnnvMFxcQ2veewa6lLuFn3DBGTLmN16fe0dh3P8sCw4JbGCmiZxsH9hLpqOsrORg//xY+e88b547aou/r97BwQyrGdGuM7JuWxg5PdNbPaIRbKTZ4c/klOLgrcOQHN6wY3gLz/jgOJw8F3P2LMXTBVbj6lqCsxBy7/u2FT15vgQ/2JcHepRzOXqVYcuyIxjkPbPLA76vro3m3HCNdlbjxM25Y1y9aYebQhurXKqV4v2ip9nBgL+nFwLF3sWOTM7ZvdEH6FSusnueN7FtS9BvJv/irS1FihhO/uWLgrOsIbJcPtwYleHnyDbj6lODP9R4AgLYDstG0Ux7q+ZbCK6gIr81NRUmBBW5esAUAmJkDDm5lGtvJ7S4I7ZcNK1uVMS9PtPgZNyylEsjJlqq3vPv8m7sqD2cn6bKJFZOYGhAEAUuXLkVAQACsra3x3HPP4bvvvgMA7N27FxKJBLt27UJYWBhsbGzQoUMHpKSkGDnq2mMhVSGwVRGS99lr7E/eZ49mYYVGikq8VOUSqJQSSGWayYZUpsKVJIdK5csVEuzf6AFreTnqN3v8/U47Y4v083boOOTOY4/Tk/Ezbnje/gpsTD6LxL/OY9YX1+HhW2rskJ5ZFYmIRIfN2FdQc0xta2DOnDn4/vvvsWrVKgQGBuLPP//EiBEjUK9ePXWZ2bNnY9myZahXrx7Gjx+P0aNH4+DBg489X2lpKUpL//4Pmp+fX+vXoE9yZyXMLYDcu5ofp9xsCzi5lRspKvGyslMiICQfW1f6wiMwBXJXBY79VA/XT9rDzb9YXe70Lif8Z1ITKIrNIHdT4N2vz8LO+fH3++BmD3g0KkLDsAJDXUadws+4YV08YYuP3rVGxjUZnOqVY1hUJpb/dBljezRBQQ6/tuhv/DRUU2FhIeLj47F79260b98eABAQEIADBw5gzZo1GDt2LABg0aJF6Nq1KwBg5syZ6Nu3L0pKSmBlZVXpnHFxcZg/f77hLqKWPJrNSyQQ9zPejejNFZfw3+mBmNm2LczMBfi0eIA2/bNx4+zfg0gbt8/D7N9O4MF9KQ5scsfaCU0w46dTkLuWaZxLUWKGYz/XQ5930g19GXUOP+OGkbRHrv739YvA+SQbJBy6gF6D7uP7L92MGNmzibOTSGvnz59HSUkJevXqpbFfoVAgODhY/bpVq1bqf3t6egIAsrKy4OvrW+mcs2bNwpQpU9Sv8/Pz4ePjo+/Qa03+fXMoywGnepp/kTq4liMnmx+xmqjnV4Kp355BaZEZSgrM4eBehrUTG8PVp0RdRmajgluDErg1KEFASAHmdg3FoW/c8eLEDI1zHd/mAkWxGZ5/lV1JNcXPuHGVFpvj+kUrePuzS+lxBOiWS4s5D+f/vmpSqSrGKWzduhXe3t4ax2QyGa5evQoAkEql6v0SiUSj7qNkMhlkMllthGsQ5WVmuHzaBiFdCnBo+99jNkK6FOCvHZXHcJD2ZDYqyGxUKMwzx/k/nTBwVmrVhQWgTFF5mNvBbzzQKvw+7F3Y7VFT/Iwbl9RSBZ/AUpw9wunspIkDe6upWbNmkMlkuHHjBho1aqSxian1RN++/9IVL0bcxwtD78GnUQnGxd6Em3cZtv7XxdihidK5fY44t9cRd2/IcH6/I5YPbQn3gGJ0GJSF0iIz/LjUD9eO2+Nehgw3zthifUwj5GTKENr3rsZ5sq5b4coROToNzTTSldQd/Iwbzpi5N9Hy+Qdw9ylF4+BCzPnyOmzslNj5P2djh/ZM0m1Qb826om7evIkRI0bAxcUFNjY2aN26NZKTk/8Rk4DY2Fh4eXnB2toa3bp1w7lz5zTOUVpainfeeQeurq6wtbXFK6+8goyMjEff6onYElNN9vb2mDZtGiZPngyVSoVOnTohPz8fhw4dgp2dHfz8/IwdolHs+9kJ9k5KDJ98B85u5UhLscKcEf7I4voZNVJcYIEfl/ghN1MGG4dyBL90FwOmp8FcKkClBDKvWOOv75qgMEcKW8cy+D33ANP+dxpeQUUa5zn0rTscPRRo2iXXOBdSh/AzbjiunmWY9fl1yJ2VyLtngYvHbRD9chDvdVUM3J+Uk5ODjh07onv37vjtt9/g5uaGq1evwtHRUV1m6dKliI+PR0JCAoKCgrBw4UL06tULKSkpsLevmOUXHR2NX375BZs3b4aLiwumTp2Kfv36ITk5Gebm5lrFIhEEMU+uMg5BELBy5Up88cUXuHbtGhwdHRESEoL33nsPKpUK3bt3R05OjvoHevLkSQQHByM1NRUNGjR46vnz8/Ph4OCAbugPC4n0qeVJd6vTDhg7BJMz3q+TsUMwLRLxDt4Uo3KhDHuFH5GXlwe5XP70CjXw8LsiIGE2zGwqTxrRlqqoBNdGLdI61pkzZ+LgwYPYv3//Y48LggAvLy9ER0djxowZACpaXdzd3bFkyRKMGzcOeXl5qFevHtavX48hQ4YAAG7dugUfHx9s27YNvXv31ip2difVgEQiQVRUFC5evAiFQoGsrCxs374dXbp0Qbdu3SAIgkZG2rp1awiCoFUCQ0REZAz5+fka2z+X/vinn3/+GWFhYRg0aBDc3NwQHByMtWvXqo+npqYiMzMTL7zwgnqfTCZD165dcejQIQBAcnIyysrKNMp4eXmhRYsW6jLaYBJDREQkYvpasdfHxwcODg7qLS4u7rHvd+3aNfU6aTt27MD48eMRFRWF//73vwCAzMyKMXju7u4a9dzd3dXHMjMzYWlpCScnpyrLaINjYoiIiERMX+vEpKena3QnVTVrVqVSISwsDIsXLwYABAcH49y5c1i1ahVGjhypLid5pAtTEIRK+yrH8vQy/8SWGCIiIoJcLtfYqkpiPD090axZM419TZs2xY0bNwAAHh4Vz3h7tEUlKytL3Trj4eEBhUKBnJycKstog0kMERGRmAkS3bdq6NixY6XnAV66dEk9O9ff3x8eHh7YuXOn+rhCocC+ffvQoUMHAEBoaCikUqlGmdu3b+Ps2bPqMtpgdxIREZGI6fok6urWnTx5Mjp06IDFixdj8ODBOHr0KL788kt8+eWXACq6kaKjo7F48WIEBgYiMDAQixcvho2NDSIiIgAADg4OiIyMxNSpU+Hi4gJnZ2dMmzYNLVu2RHh4uNaxMIkhIiIirbVp0wY//PADZs2ahQULFsDf3x8rVqzA8OHD1WViYmJQXFyMCRMmICcnB+3atcPvv/+uXiMGAJYvXw4LCwsMHjwYxcXF6NmzJxISErReIwbgOjHPJK4TY3hcJ8bwuE6MgXGdGIMy5Doxfmvn6rxOTNqYD2o11trClhgiIiIR41Osn+LTTz/V+oRRUVE1DoaIiIhIW1olMcuXL9fqZA9XsiUiIiIDMtGBIVolMampqbUdBxEREdWAKXcn1XidGIVCgZSUFJSXl+szHiIiIqoOQQ+bSFU7iSkqKkJkZCRsbGzQvHlz9Qp9UVFR+PDDD/UeIBEREdHjVDuJmTVrFk6dOoW9e/fCyurvKV3h4eH45ptv9BocERERPY1ED5s4VXuK9Y8//ohvvvkGzz//vMZDmpo1a4arV6/qNTgiIiJ6Cl27hEypOyk7Oxtubm6V9hcWFlbryZNEREREuqh2EtOmTRts3bpV/fph4rJ27Vq0b99ef5ERERHR05nwwN5qdyfFxcXhxRdfxPnz51FeXo5PPvkE586dw19//YV9+/bVRoxERERUlRo8ibpSfZGqdktMhw4dcPDgQRQVFaFhw4b4/fff4e7ujr/++guhoaG1ESMRERFRJTV6dlLLli2RmJio71iIiIiomgShYtOlvljVKIlRKpX44YcfcOHCBUgkEjRt2hT9+/eHhQWfJ0lERGRQJjw7qdpZx9mzZ9G/f39kZmaicePGAIBLly6hXr16+Pnnn9GyZUu9B0lERET0qGqPiXnrrbfQvHlzZGRk4Pjx4zh+/DjS09PRqlUrjB07tjZiJCIioqo8HNiryyZS1W6JOXXqFJKSkuDk5KTe5+TkhEWLFqFNmzZ6DY6IiIieTCJUbLrUF6tqt8Q0btwYd+7cqbQ/KysLjRo10ktQREREpCUTXidGqyQmPz9fvS1evBhRUVH47rvvkJGRgYyMDHz33XeIjo7GkiVLajteIiIiIgBadic5OjpqPFJAEAQMHjxYvU/4//lZL7/8MpRKZS2ESURERI9lwovdaZXE7Nmzp7bjICIioprgFOsn69q1a23HQURERFQtNV6drqioCDdu3IBCodDY36pVK52DIiIiIi2xJUZ72dnZePPNN/Hbb7899jjHxBARERmQCScx1Z5iHR0djZycHBw+fBjW1tbYvn07EhMTERgYiJ9//rk2YiQiIiKqpNotMbt378ZPP/2ENm3awMzMDH5+fujVqxfkcjni4uLQt2/f2oiTiIiIHseEZydVuyWmsLAQbm5uAABnZ2dkZ2cDqHiy9fHjx/UbHRERET3RwxV7ddnEqkYr9qakpAAAWrdujTVr1uDmzZtYvXo1PD099R4gERER0eNUuzspOjoat2/fBgDMmzcPvXv3xoYNG2BpaYmEhAR9x0dERERPYsIDe6udxAwfPlz97+DgYFy/fh0XL16Er68vXF1d9RocERERUVVqvE7MQzY2NggJCdFHLERERFRNEuj4FGu9RWJ4WiUxU6ZM0fqE8fHxNQ6GiIiISFtaJTEnTpzQ6mT/fEgk6YFEUrFRrRvfoLOxQzA55k6Oxg7BpKgeFBo7BJMiEQCUGejNTHiKNR8ASUREJGYmPLC32lOsiYiIiJ4FOg/sJSIiIiMy4ZYYJjFEREQipuuquya1Yi8RERHRs4AtMURERGJmwt1JNWqJWb9+PTp27AgvLy+kpaUBAFasWIGffvpJr8ERERHRUwh62ESq2knMqlWrMGXKFPTp0we5ublQKpUAAEdHR6xYsULf8RERERE9VrWTmJUrV2Lt2rWYPXs2zM3N1fvDwsJw5swZvQZHRERET/ZwYK8um1hVe0xMamoqgoODK+2XyWQoLOSKkERERAZlwiv2Vrslxt/fHydPnqy0/7fffkOzZs30ERMRERFpy4THxFS7JWb69OmYOHEiSkpKIAgCjh49ik2bNiEuLg7//ve/ayNGIiIiokqqncS8+eabKC8vR0xMDIqKihAREQFvb2988sknGDp0aG3ESERERFUw5cXuarROzJgxYzBmzBjcvXsXKpUKbm5u+o6LiIiItGHC68TotNidq6urvuIgIiIiqpZqJzH+/v6QSKoeyXzt2jWdAiIiIqJq0HWatCm1xERHR2u8Lisrw4kTJ7B9+3ZMnz5dX3ERERGRNtidpL133333sfs///xzJCUl6RwQERERkTb09hTrl156CVu2bNHX6YiIiEgbXCdGd9999x2cnZ31dToiIiLSAqdYV0NwcLDGwF5BEJCZmYns7Gx88cUXeg2OiIiIqCrVTmIGDBig8drMzAz16tVDt27d0KRJE33FRURERPRE1UpiysvL0aBBA/Tu3RseHh61FRMRERFpy4RnJ1VrYK+FhQXefvttlJaW1lY8REREVA0Px8TosolVtWcntWvXDidOnKiNWIiIiIi0Vu0xMRMmTMDUqVORkZGB0NBQ2Nraahxv1aqV3oIjIiIiLYi4NUUXWicxo0ePxooVKzBkyBAAQFRUlPqYRCKBIAiQSCRQKpX6j5KIiIgez4THxGidxCQmJuLDDz9EampqbcZDREREpBWtkxhBqEjV/Pz8ai0YIiIiqh4udqelJz29moiIiIyA3UnaCQoKemoic//+fZ0CIiIiItJGtZKY+fPnw8HBobZiISIiomoydndSXFwc3nvvPbz77rtYsWIFgIohKPPnz8eXX36JnJwctGvXDp9//jmaN2+urldaWopp06Zh06ZNKC4uRs+ePfHFF1+gfv36Wr93tZKYoUOHws3NrTpViIiIqDYZsTvp2LFj+PLLLystr7J06VLEx8cjISEBQUFBWLhwIXr16oWUlBTY29sDAKKjo/HLL79g8+bNcHFxwdSpU9GvXz8kJyfD3Nxcq/fXerE7jochIiKqu/Lz8zW2p63O/+DBAwwfPhxr166Fk5OTer8gCFixYgVmz56NgQMHokWLFkhMTERRURE2btwIAMjLy8N//vMfLFu2DOHh4QgODsbXX3+NM2fO4I8//tA6Zq2TmIezk4iIiOgZIuhhA+Dj4wMHBwf1FhcX98S3nThxIvr27Yvw8HCN/ampqcjMzMQLL7yg3ieTydC1a1ccOnQIAJCcnIyysjKNMl5eXmjRooW6jDa07k5SqVRan5SIiIgMQ19jYtLT0yGXy9X7ZTJZlXU2b96M48eP49ixY5WOZWZmAgDc3d019ru7uyMtLU1dxtLSUqMF52GZh/W1Ue3HDhAREdEzRE9jYuRyuUYSU5X09HS8++67+P3332FlZVVluUeHoTxc2f+JoWhR5p+q/QBIIiIiMl3JycnIyspCaGgoLCwsYGFhgX379uHTTz+FhYWFugXm0RaVrKws9TEPDw8oFArk5ORUWUYbTGKIiIjETE9jYrTVs2dPnDlzBidPnlRvYWFhGD58OE6ePImAgAB4eHhg586d6joKhQL79u1Dhw4dAAChoaGQSqUaZW7fvo2zZ8+qy2iD3UlEREQiZuh1Yuzt7dGiRQuNfba2tnBxcVHvj46OxuLFixEYGIjAwEAsXrwYNjY2iIiIAAA4ODggMjISU6dOhYuLC5ydnTFt2jS0bNmy0kDhJ2ESQ3rh4qFA5Hu30aZHPiytVLh5TYb4qb64csbG2KHVSSOm3MbrU+9o7LufZYFhwS2qqEHVMXxCKoZPTNPYd/+uFCO6dvz/VwKGT7iOFwfdhp28HCmn7fHFwiDcuGpr+GDrgBZtC/DauNsIbFkEF/cyzB/TCH/9/veATysbJUbPzED7F3IgdyrHnQwZflrnjq1fc92yZ1VMTAyKi4sxYcIE9WJ3v//+u3qNGABYvnw5LCwsMHjwYPVidwkJCVqvEQOYaBKTkJCA6Oho5ObmPvb43r170b17d+Tk5MDR0fGp5atz7rrIzqEc8T9exulD9pgzIgC5dy3g2UCBwnztP4hUfdcvWmHm0Ibq1yol13LSp+uXbTD7refUr5X/uL+vRabjX29kIH52E9y8bo2h49Kw6N+nMLZvWxQXmeSvVZ1Y2SiResEGO//nirlrrlY6Pu79G3iufQE+ig7AnQwZQjrnYdLCNNy7I8XhnU6POaOJeQaenbR3716N1xKJBLGxsYiNja2yjpWVFVauXImVK1fW+H35v00LQ4YMQZ8+fYwdxjNr8IQs3L1liWVTfNX77mRUPTWP9EOpBHKypcYOo85SKiXIufu4z7GAAa9nYPOXfjj0Rz0AwLL3mmLjnwfRrW8Wfvufl2EDrQOS9joiaa/j/7+qnMQ0DSnEH1tccfpwxcyZ3za5oc/wbAS1KmQSA+M/dsCYRDWwV6FQGOV9ra2t+biFJ3j+hTxcOm2D2WtS8c2ps/h8Rwpeirhn7LDqPG9/BTYmn0XiX+cx64vr8PB98uqaVD3evsVYv+cQvtpxGDM+OgeP+sUAAI/6JXCup8Dxg39/eZaXmeFMkiOaBucZK9w67dwxOzwfngMXdwUAAa3a58PbvwTJ+/gsP1Nn1CSmoKAAw4cPh62tLTw9PbF8+XJ069YN0dHRAIAGDRpg4cKFGDVqFBwcHDBmzBgAwKFDh9ClSxdYW1vDx8cHUVFRKCwsVJ9XoVAgJiYG3t7esLW1Rbt27So1df3TvXv30LZtW7zyyisoKSmpdDwhIQGOjo7q16dOnUL37t1hb28PuVyO0NBQJCUladTZsWMHmjZtCjs7O7z44ou4fft2le9fWlpaablnMfH0VaDf63dxK1WG9yICsHW9C95ekIHw1/hE89py8YQtPnrXF+8Nb4gVMT5wqleG5T9dhr1TubFDqxNSTsux7L2mmDu2FT6dFwQnVwU+3nAc9g5lcHKt+GMq956lRp3ce5bqY6Rfq2J9kXbZGhuOnsKvV5KxMPESPp/jh3NJ9k+vbAoMPDvpWWLUJGbKlCk4ePAgfv75Z+zcuRP79+/H8ePHNcp89NFHaNGiBZKTkzF37lycOXMGvXv3xsCBA3H69Gl88803OHDgACZNmqSu8+abb+LgwYPYvHkzTp8+jUGDBuHFF1/E5cuXK8WQkZGBzp07o0mTJvj++++fuHDPQ8OHD0f9+vVx7NgxJCcnY+bMmZBK/27WLyoqwscff4z169fjzz//xI0bNzBt2rQqzxcXF6ex1LOPj482t++ZITEDrpy1xroPvXD1nA22fe2K3za6oO/Iu8YOrc5K2iPHgW2OuH7RGif222PuyAAAQK9BTBz1IemACw7urIfrl+1w8rAz5k2oeLhd+IC/17149EksEknlfaQf/d+8g6bBhZg3OhDv9GuGtYt8MHFhGoI7suULgEknMUYbE1NQUIDExERs3LgRPXv2BACsW7cOXl6a/ck9evTQSABGjhyJiIgIdWtNYGAgPv30U3Tt2hWrVq3CzZs3sWnTJmRkZKjPNW3aNGzfvh3r1q3D4sWL1ee6dOkSevXqhf79++OTTz7RepXAGzduYPr06WjSpIk6hn8qKyvD6tWr0bBhxaDLSZMmYcGCBVWeb9asWZgyZYr6dX5+vqgSmftZFki7pJn8pV+xQqc+/AVjKKXF5rh+0Qre/uxSqg2lxeZIu2QHL99i/LXLFQDg5KrQGDPj4Kyo1DpDurOUqTBq+k18MK4Rju52BACkXrRBw2ZFeHVsJk4cZJeSKTNaEnPt2jWUlZWhbdu26n0ODg5o3LixRrmwsDCN18nJybhy5Qo2bNig3icIAlQqFVJTU3H27FkIgoCgoCCNeqWlpXBxcVG/Li4uRqdOnTBs2DB88skn1Yp9ypQpeOutt7B+/XqEh4dj0KBB6oQFAGxsbDRee3p6Iisrq8rzyWSyJz6j4ll3/pgtfBpqfnl6B5Qi6yYHnRqK1FIFn8BSnD1iZ+xQ6iQLqQo+AYU4e9wBmRlWuJ9tiZAOObh20V59vGVYLtbFN3zKmai6LKQCpJYCHn18n0opgURUozprj+T/N13qi5XRkpiHT8V+3LMV/snWVnPdBZVKhXHjxiEqKqrSOX19fXH69GmYm5sjOTm50lxzO7u/f8HLZDKEh4dj69atmD59OurXr6917LGxsYiIiMDWrVvx22+/Yd68edi8eTP+9a9/AYBG19LDa6zLTwH/fq0blv90CUPfuYM/f3FE49ZF6DP8HlbEaH9PqXrGzL2JwzsdkHVTCkfXckS8ewc2dkrs/J+zsUOrEyKnXcGRva7Ivi2Do3MZho5Pg42dErt+9AAgwY/r62PwmDTcTLPGrTRrDBl7A6Ul5ti7lRMAasLKRgmvBn//IeThU4qAZkUoyDVH9i0ZTv9lj7fey4CixAx3bsrQql0Ber56F19+4PuEs5qQZ2CKtbEYLYlp2LAhpFIpjh49qu46yc/Px+XLl9G1a9cq64WEhODcuXNo1KjRY48HBwdDqVQiKysLnTt3rvI8ZmZmWL9+PSIiItCjRw/s3bu3UlfWkwQFBSEoKAiTJ0/GsGHDsG7dOnUSY2ounbLBgrf88ebM2xgenYnMdEusnueNPT/wC7W2uHqWYdbn1yF3ViLvngUuHrdB9MtByLrJ7gx9cHUvxYyPzkPuVIa8+1KknJZjckQIsm5XdJt+9x8fyGRKTJx7GXbyMqSclmPOmFZcI6aGgloVYuk3KerX495PBwDs/J8Llk0LQNw7DfFmTAZiPrkGe8dyZGXIkPhRfWz9up6xQn6mmPIUa6P9j7O3t8cbb7yB6dOnw9nZGW5ubpg3bx7MzMyeODZlxowZeP755zFx4kSMGTMGtra2uHDhAnbu3ImVK1ciKCgIw4cPx8iRI7Fs2TIEBwfj7t272L17N1q2bKmx3ou5uTk2bNiAYcOGqRMZDw+PJ8ZdXFyM6dOn47XXXoO/vz8yMjJw7NgxvPrqq3q7N2J05A8HHPmDfdOGEjehgbFDqNOWTG/+lBISbPjCHxu+8DdIPHXd6cNyvOjXpsrjOdlSxE/nvabKjNqjGB8fj/bt26Nfv34IDw9Hx44d0bRp0yfOEGrVqhX27duHy5cvo3PnzggODsbcuXPh6empLrNu3TqMHDkSU6dORePGjfHKK6/gyJEjjx0sa2FhgU2bNqF58+bo0aPHE8euABWJz7179zBy5EgEBQVh8ODBeOmllzB//vya3wgiIqKaMuHZSRLhGRqsUVhYCG9vbyxbtgyRkZHGDsdo8vPz4eDggG6SAbCQcHAs1U3m/1h7iWqf6kHh0wuR3pQLZdhT9j/k5eVBLpfXyns8/K5oPm4xzC2fvjxIVZSKEpxb816txlpbjNqBe+LECVy8eBFt27ZFXl6eehpy//79jRkWERERiYDRR6F9/PHHSElJgaWlJUJDQ7F//364uroaOywiIiJR4MBeIwkODkZycrIxQyAiIhI3E55izaWCiIiISJSM3p1ERERENcfuJCIiIhIndicRERERiQtbYoiIiESM3UlEREQkTibcncQkhoiISMxMOInhmBgiIiISJbbEEBERiRjHxBAREZE4sTuJiIiISFzYEkNERCRiEkGARKh5c4oudY2NSQwREZGYsTuJiIiISFzYEkNERCRinJ1ERERE4sTuJCIiIiJxYUsMERGRiLE7iYiIiMTJhLuTmMQQERGJmCm3xHBMDBEREYkSW2KIiIjEjN1JREREJFZi7hLSBbuTiIiISJTYEkNERCRmglCx6VJfpJjEEBERiRhnJxERERGJDFtiiIiIxIyzk4iIiEiMJKqKTZf6YsXuJCIiIhIltsQQERGJGbuTiIiISIxMeXYSkxgiIiIxM+F1YjgmhoiIiESJLTFEREQixu4kejYJuo7WInp2KXNyjB2CSbk5o4OxQzApytISYPn/DPNmJjywl91JREREJEpsiSEiIhIxdicRERGROHF2EhEREZG4sCWGiIhIxNidREREROLE2UlERERE4sKWGCIiIhFjdxIRERGJk0qo2HSpL1JMYoiIiMSMY2KIiIiIxIUtMURERCImgY5jYvQWieExiSEiIhIzrthLRERE9HRxcXFo06YN7O3t4ebmhgEDBiAlJUWjjCAIiI2NhZeXF6ytrdGtWzecO3dOo0xpaSneeecduLq6wtbWFq+88goyMjKqFQuTGCIiIhF7OMVal6069u3bh4kTJ+Lw4cPYuXMnysvL8cILL6CwsFBdZunSpYiPj8dnn32GY8eOwcPDA7169UJBQYG6THR0NH744Qds3rwZBw4cwIMHD9CvXz8olUqtY2F3EhERkZgZeHbS9u3bNV6vW7cObm5uSE5ORpcuXSAIAlasWIHZs2dj4MCBAIDExES4u7tj48aNGDduHPLy8vCf//wH69evR3h4OADg66+/ho+PD/744w/07t1bq1jYEkNERETIz8/X2EpLS7Wql5eXBwBwdnYGAKSmpiIzMxMvvPCCuoxMJkPXrl1x6NAhAEBycjLKyso0ynh5eaFFixbqMtpgEkNERCRiEkHQeQMAHx8fODg4qLe4uLinvrcgCJgyZQo6deqEFi1aAAAyMzMBAO7u7hpl3d3d1ccyMzNhaWkJJyenKstog91JREREYqb6/02X+gDS09Mhl8vVu2Uy2VOrTpo0CadPn8aBAwcqHZNINCdvC4JQad+jtCnzT2yJISIiIsjlco3taUnMO++8g59//hl79uxB/fr11fs9PDwAoFKLSlZWlrp1xsPDAwqFAjk5OVWW0QaTGCIiIhHTV3eStgRBwKRJk/D9999j9+7d8Pf31zju7+8PDw8P7Ny5U71PoVBg37596NChAwAgNDQUUqlUo8zt27dx9uxZdRltsDuJiIhIzAw8O2nixInYuHEjfvrpJ9jb26tbXBwcHGBtbQ2JRILo6GgsXrwYgYGBCAwMxOLFi2FjY4OIiAh12cjISEydOhUuLi5wdnbGtGnT0LJlS/VsJW0wiSEiIhIzA6/Yu2rVKgBAt27dNPavW7cOo0aNAgDExMSguLgYEyZMQE5ODtq1a4fff/8d9vb26vLLly+HhYUFBg8ejOLiYvTs2RMJCQkwNzfXOhYmMURERKQ1QYukRyKRIDY2FrGxsVWWsbKywsqVK7Fy5coax8IkhoiISMRqsuruo/XFikkMERGRmPEBkERERETiwpYYIiIiEZOoKjZd6osVkxgiIiIxY3cSERERkbiwJYaIiEjMDLzY3bOESQwREZGI1eTRAY/WFyt2JxEREZEosSWGiIhIzEx4YC+TGCIiIjETAOgyTVq8OQyTGCIiIjHjmBgiIiIikWFLDBERkZgJ0HFMjN4iMTgmMURERGJmwgN72Z1EREREosSWGNKbfm/cxaC3s+HsVoa0S1ZY/b4Xzh61M3ZYdRbvt+HxnuvHkJZnMaTVOXjZFwAArtx3xuojoTiQ5gcAcLEpwuSOh9HBNx32MgWSb3pi8b5OuJHrqD7Huld/Qpv6tzTO+1tKI0zf3stg1/HMUAGQ6FhfpJjEVKFbt25o3bo1VqxYYexQRKHrKzkYP/8WPnvPG+eO2qLv6/ewcEMqxnRrjOyblsYOr87h/TY83nP9yXxgh+UHn8eNXAcAQP+mKVj58na8tnEQrt53wif9tqNcZYaoX1/Cg1IpRoacxr//9Qv6rx+K4nKp+jz/O9MUnx1uq35dWm5u8Gt5FnB2kgnbu3cvJBIJcnNzjR2KqA0cexc7Njlj+0YXpF+xwup53si+JUW/kfeMHVqdxPtteLzn+rMvtQH2X/dDWq4j0nId8elf7VBUJsVznnfg55iH1p538MHuLjh7xw3Xc52wcE9n2EjL0KfxZY3zlJRb4F6RjXp7oJAZ6YrIWEw+iTGksrIyY4dQKyykKgS2KkLyPnuN/cn77NEsrNBIUdVdvN+Gx3tee8wkKrwUdBnWFmU4edsdluZKAIBC+XerikowQ5nKHMFemRp1+za+jP1j1+HHEZsxrdMh2EgVBo39mfFwYK8um0iZRBJTWlqKqKgouLm5wcrKCp06dcKxY8dw/fp1dO/eHQDg5OQEiUSCUaNGqeupVCrExMTA2dkZHh4eiI2N1ThvXl4exo4dCzc3N8jlcvTo0QOnTp1SH4+NjUXr1q3x1VdfISAgADKZDIKIPyxVkTsrYW4B5N7V7J3MzbaAk1u5kaKqu3i/DY/3XP8CXe7h6NtrcXzSl5jb40+8u/VFXLvvjNQcR9zMt8e7HY5ALiuFhZkSkWHHUc+2CPVsi9T1f70YiJjtvfDmllew5mgowhtdwyf9dhjxiozIhJMYkxgTExMTgy1btiAxMRF+fn5YunQpevfujcuXL2PLli149dVXkZKSArlcDmtra3W9xMRETJkyBUeOHMFff/2FUaNGoWPHjujVqxcEQUDfvn3h7OyMbdu2wcHBAWvWrEHPnj1x6dIlODs7AwCuXLmCb7/9Flu2bIG5+eP7a0tLS1FaWqp+nZ+fX7s3pJY8+v9AIoGo1x941vF+Gx7vuf6k5jji1Y2DIZeVoleja1jUazdGbemPa/edMXlrbywI34ND479CuUqCwzfq48/rvhr1t5xrpv73lXsuSMt1xLfDvkPTetm4kF3P0JdDRlLnk5jCwkKsWrUKCQkJeOmllwAAa9euxc6dO/HVV1+hTZs2AAA3Nzc4Ojpq1G3VqhXmzZsHAAgMDMRnn32GXbt2oVevXtizZw/OnDmDrKwsyGQV/bAff/wxfvzxR3z33XcYO3YsAEChUGD9+vWoV6/q/1RxcXGYP3++vi/dYPLvm0NZDjjV0/yL1MG1HDnZdf4jZnC834bHe65/5SpzpOdVDOw9l+WG5u5ZGNH6DBbs7orzWfXw2sbBsLMshdRchZxia2wcsgXn7lT9e/R8livKlGbwc8wzvSSG68TUXVevXkVZWRk6duyo3ieVStG2bVtcuHDhiXVbtWql8drT0xNZWVkAgOTkZDx48AAuLi6ws7NTb6mpqbh69aq6jp+f3xMTGACYNWsW8vLy1Ft6enp1L9OoysvMcPm0DUK6FGjsD+lSgPNJtkaKqu7i/TY83vPaJ5FAPR7moQcKGXKKreHrmIvmbtnYc82/yvqNXO5Daq5CdqFNbYf67FHpYROpOv8nxMMxKBKJpNL+R/c9SiqVaryWSCRQqSp+2iqVCp6enti7d2+lev9s0bG1ffovOJlMpm7NEavvv3TF9E/Tcem0NS4k2aLPiHtw8y7D1v+6GDu0Oon32/B4z/Xn3Q6Hsf+6LzIL7GBrWYaXgq6gjfctjP+pLwDghUZXkVNshdsF9gh0vYeZXQ9i97UGOHTDBwDg45BXMaj3ui9yiq3Q0CUH0zsfwvksV5y47WHMSzMKU55iXeeTmEaNGsHS0hIHDhxAREQEgIpZQklJSYiOjoalZcX6Dkql8kmnqSQkJASZmZmwsLBAgwYN9B226Oz72Qn2TkoMn3wHzm7lSEuxwpwR/sji+hm1gvfb8HjP9cfFphhxvXejnk0hChSWuHTXBeN/6ou//j9JqWdbiJguB+FiU4zsQhv8fKExVh8NVdcvU5qjnU8GRrQ+DRtpGTIf2OHPVD98cSQMKqHOdzDQP9T5JMbW1hZvv/02pk+fDmdnZ/j6+mLp0qUoKipCZGQkioqKIJFI8Ouvv6JPnz6wtraGnd3TV+AMDw9H+/btMWDAACxZsgSNGzfGrVu3sG3bNgwYMABhYWEGuLpny6+Jrvg10dXYYZgM3m/D4z3Xj/f/6P7E4xtOtcKGU62qPJ75wA5vbhmg56hEjGNi6rYPP/wQr776Kl5//XWEhITgypUr2LFjB5ycnODt7Y358+dj5syZcHd3x6RJk7Q6p0QiwbZt29ClSxeMHj0aQUFBGDp0KK5fvw53d/daviIiIqL/pxJ030RKItTFhUtELj8/Hw4ODuiG/rCQSJ9egYjoKW7O6GDsEEyKsrQEl5a/h7y8PMjl8lp5j4ffFeENo2FhXvNxleXKUvxxdUWtxlpb6nx3EhERUZ1mwt1JTGKIiIhETddVd8WbxJjEmBgiIiKqe9gSQ0REJGbsTiIiIiJRUgnQqUtIxLOT2J1EREREosSWGCIiIjETVBWbLvVFikkMERGRmHFMDBEREYkSx8QQERERiQtbYoiIiMSM3UlEREQkSgJ0TGL0FonBsTuJiIiIRIktMURERGLG7iQiIiISJZUKgA5rvajEu04Mu5OIiIhIlNgSQ0REJGbsTiIiIiJRMuEkht1JREREJEpsiSEiIhIzE37sAJMYIiIiERMEFQQdnkStS11jYxJDREQkZoKgW2sKx8QQERERGRZbYoiIiMRM0HFMjIhbYpjEEBERiZlKBUh0GNci4jEx7E4iIiIiUWJLDBERkZixO4mIiIjESFCpIOjQnSTmKdbsTiIiIiJRYksMERGRmLE7iYiIiERJJQAS00xi2J1EREREosSWGCIiIjETBAC6rBMj3pYYJjFEREQiJqgECDp0JwlMYoiIiMgoBBV0a4nhFGsiIiIyIV988QX8/f1hZWWF0NBQ7N+/3+AxMIkhIiISMUEl6LxV1zfffIPo6GjMnj0bJ06cQOfOnfHSSy/hxo0btXCFVWMSQ0REJGaCSvetmuLj4xEZGYm33noLTZs2xYoVK+Dj44NVq1bVwgVWjWNinkEPB1mVo0yn9YuIiB5SlpYYOwST8vB+G2LQrK7fFeUoAwDk5+dr7JfJZJDJZJXKKxQKJCcnY+bMmRr7X3jhBRw6dKjmgdQAk5hnUEFBAQDgALYZORIiqjOW/2TsCExSQUEBHBwcauXclpaW8PDwwIFM3b8r7Ozs4OPjo7Fv3rx5iI2NrVT27t27UCqVcHd319jv7u6OzMxMnWOpDiYxzyAvLy+kp6fD3t4eEonE2OFUS35+Pnx8fJCeng65XG7scOo83m/D4v02PLHec0EQUFBQAC8vr1p7DysrK6SmpkKhUOh8LkEQKn3fPK4V5p8eLf+4c9Q2JjHPIDMzM9SvX9/YYehELpeL6heO2PF+Gxbvt+GJ8Z7XVgvMP1lZWcHKyqrW3+efXF1dYW5uXqnVJSsrq1LrTG3jwF4iIiLSmqWlJUJDQ7Fz506N/Tt37kSHDh0MGgtbYoiIiKhapkyZgtdffx1hYWFo3749vvzyS9y4cQPjx483aBxMYkivZDIZ5s2b99S+VNIP3m/D4v02PN7zZ9OQIUNw7949LFiwALdv30aLFi2wbds2+Pn5GTQOiSDmhyYQERGRyeKYGCIiIhIlJjFEREQkSkxiiIiISJSYxFCt2bt3LyQSCXJzc40dCpFeJSQkwNHRscrjj372n1a+Ouc2Bd26dUN0dLSxwyARYBJDRFTLhgwZgkuXLhk7jGcO/9AhXXGKNRGZDIVCAUtLS4O/r7W1NaytrQ3+vvS3srIySKVSY4dBesaWGNKJIAhYunQpAgICYG1tjeeeew7fffedscOqc550nx/+Nbtr1y6EhYXBxsYGHTp0QEpKipGjrn0FBQUYPnw4bG1t4enpieXLl2t0RTRo0AALFy7EqFGj4ODggDFjxgAADh06hC5dusDa2ho+Pj6IiopCYWGh+rwKhQIxMTHw9vaGra0t2rVrh71791YZx71799C2bVu88sorKCmp/LToR7uITp06he7du8Pe3h5yuRyhoaFISkrSqLNjxw40bdoUdnZ2ePHFF3H79u2a3ygjKi0tRVRUFNzc3GBlZYVOnTrh2LFjuH79Orp37w4AcHJygkQiwahRo9T1VCoVYmJi4OzsDA8Pj0oPIszLy8PYsWPh5uYGuVyOHj164NSpU+rjsbGxaN26Nb766isEBARAJpMZ5InSZFhMYkgnc+bMwbp167Bq1SqcO3cOkydPxogRI7Bv3z5jh1anaHOfZ8+ejWXLliEpKQkWFhYYPXq0ESM2jClTpuDgwYP4+eefsXPnTuzfvx/Hjx/XKPPRRx+hRYsWSE5Oxty5c3HmzBn07t0bAwcOxOnTp/HNN9/gwIEDmDRpkrrOm2++iYMHD2Lz5s04ffo0Bg0ahBdffBGXL1+uFENGRgY6d+6MJk2a4Pvvv9fqOTbDhw9H/fr1cezYMSQnJ2PmzJkarQRFRUX4+OOPsX79evz555+4ceMGpk2bpsOdMp6YmBhs2bIFiYmJOH78OBo1aoTevXvD3t4eW7ZsAQCkpKTg9u3b+OSTT9T1EhMTYWtriyNHjmDp0qVYsGCBepl7QRDQt29fZGZmYtu2bUhOTkZISAh69uyJ+/fvq89x5coVfPvtt9iyZQtOnjxp0OsmAxGIaujBgweClZWVcOjQIY39kZGRwrBhw4Q9e/YIAIScnBzjBFhHaHuf//jjD/WxrVu3CgCE4uJiQ4drMPn5+YJUKhX+97//qffl5uYKNjY2wrvvvisIgiD4+fkJAwYM0Kj3+uuvC2PHjtXYt3//fsHMzEwoLi4Wrly5IkgkEuHmzZsaZXr27CnMmjVLEARBWLduneDg4CCkpKQIvr6+wjvvvCOoVCp12Uc/+w/LP2Rvby8kJCQ89rrWrVsnABCuXLmi3vf5558L7u7u2t2YZ8iDBw8EqVQqbNiwQb1PoVAIXl5ewtKlS6v8HdG1a1ehU6dOGvvatGkjzJgxQxAEQdi1a5cgl8uFkpISjTINGzYU1qxZIwiCIMybN0+QSqVCVlZWLVwZPSs4JoZq7Pz58ygpKUGvXr009isUCgQHBxspqrpH2/vcqlUr9b89PT0BVDxV1tfX1zCBGti1a9dQVlaGtm3bqvc5ODigcePGGuXCwsI0XicnJ+PKlSvYsGGDep8gCFCpVEhNTcXZs2chCAKCgoI06pWWlsLFxUX9uri4GJ06dcKwYcM0WhC0MWXKFLz11ltYv349wsPDMWjQIDRs2FB93MbGRuO1p6cnsrKyqvUez4KrV6+irKwMHTt2VO+TSqVo27YtLly4gDZt2lRZ95+fZ0DzHiQnJ+PBgwcaPw+g4mdy9epV9Ws/Pz/Uq1dPH5dCzygmMVRjKpUKALB161Z4e3trHJPJZBq/TKjmtL3P/+yOkEgkGnXrIuH/xzc8vNZH9z9ka2ur8VqlUmHcuHGIioqqdE5fX1+cPn0a5ubmSE5Ohrm5ucZxOzs79b9lMhnCw8OxdetWTJ8+HfXr19c69tjYWERERGDr1q347bffMG/ePGzevBn/+te/AKDSAFSJRCLK8RxP+hk9uu9Rj7sHDz/PKpUKnp6ejx2n9M+xR4/+7KnuYRJDNdasWTPIZDLcuHEDXbt2rXScSYx+8D4/XsOGDSGVSnH06FH4+PgAAPLz83H58uXH3qeHQkJCcO7cOTRq1Oixx4ODg6FUKpGVlYXOnTtXeR4zMzOsX78eERER6NGjB/bu3QsvLy+t4w8KCkJQUBAmT56MYcOGYd26deokpq5o1KgRLC0tceDAAURERAComCWUlJSE6Oho9UwxpVJZrfOGhIQgMzMTFhYWaNCggb7DJhFhEkM1Zm9vj2nTpmHy5MlQqVTo1KkT8vPzcejQIdjZ2Rn8aaZ1Fe/z49nb2+ONN97A9OnT4ezsDDc3N8ybNw9mZmZP/Ct/xowZeP755zFx4kSMGTMGtra2uHDhAnbu3ImVK1ciKCgIw4cPx8iRI7Fs2TIEBwfj7t272L17N1q2bIk+ffqoz2Vubo4NGzZg2LBh6kTGw8PjiXEXFxdj+vTpeO211+Dv74+MjAwcO3YMr776qt7uzbPC1tYWb7/9tvpn5Ovri6VLl6KoqAiRkZEoKiqCRCLBr7/+ij59+sDa2lqjtasq4eHhaN++PQYMGIAlS5agcePGuHXrFrZt24YBAwZU6kKkuotJDOnkgw8+gJubG+Li4nDt2jU4OjoiJCQE7733Xp3uyjA03ufHi4+Px/jx49GvXz/I5XLExMQgPT39iTOEWrVqhX379mH27Nno3LkzBEFAw4YNMWTIEHWZdevWYeHChZg6dSpu3rwJFxcXtG/fXiOBecjCwgKbNm3CkCFD1InMk5ibm+PevXsYOXIk7ty5A1dXVwwcOBDz58+v8X14ln344YdQqVR4/fXXUVBQgLCwMOzYsQNOTk5wcnLC/PnzMXPmTLz55psYOXIkEhISnnpOiUSCbdu2Yfbs2Rg9ejSys7Ph4eGBLl26wN3dvfYvip4ZEkGMHa1ERI9RWFgIb29vLFu2DJGRkcYOh4hqGVtiiEi0Tpw4gYsXL6Jt27bIy8vDggULAAD9+/c3cmREZAhMYohI1D7++GOkpKTA0tISoaGh2L9/P1xdXY0dFhEZALuTiIiISJT42AEiIiISJSYxREREJEpMYoiIiEiUmMQQERGRKDGJISIiIlFiEkNEjxUbG4vWrVurX48aNQoDBgwweBzXr1+HRCLByZMnqyzToEEDrFixQutzJiQkaDwosKYkEgl+/PFHnc9DRDXDJIZIREaNGgWJRAKJRAKpVIqAgABMmzYNhYWFtf7en3zyiVZLwgPaJR5ERLriYndEIvPiiy9i3bp1KCsrw/79+/HWW2+hsLAQq1atqlS2rKwMUqlUL+/r4OCgl/MQEekLW2KIREYmk8HDwwM+Pj6IiIjA8OHD1V0aD7uAvvrqKwQEBEAmk0EQBOTl5WHs2LFwc3ODXC5Hjx49cOrUKY3zfvjhh3B3d4e9vT0iIyNRUlKicfzR7iSVSoUlS5agUaNGkMlk8PX1xaJFiwAA/v7+AIDg4GBIJBJ069ZNXW/dunVo2rQprKys0KRJE3zxxRca73P06FEEBwfDysoKYWFhOHHiRLXvUXx8PFq2bAlbW1v4+PhgwoQJePDgQaVyP/74I4KCgmBlZYVevXohPT1d4/gvv/yC0NBQWFlZISAgAPPnz0d5eXm14yGi2sEkhkjkrK2tUVZWpn595coVfPvtt9iyZYu6O6dv377IzMzEtm3bkJycjJCQEPTs2RP3798HAHz77beYN28eFi1ahKSkJHh6elZKLh41a9YsLFmyBHPnzsX58+exceNG9ROEjx49CgD4448/cPv2bXz//fcAgLVr12L27NlYtGgRLly4gMWLF2Pu3LlITEwEUPEAx379+qFx48ZITk5GbGwspk2bVu17YmZmhk8//RRnz55FYmIidu/ejZiYGI0yRUVFWLRoERITE3Hw4EHk5+dj6NCh6uM7duzAiBEjEBUVhfPnz2PNmjVISEhQJ2pE9AwQiEg03njjDaF///7q10eOHBFcXFyEwYMHC4IgCPPmzROkUqmQlZWlLrNr1y5BLpcLJSUlGudq2LChsGbNGkEQBKF9+/bC+PHjNY63a9dOeO655x773vn5+YJMJhPWrl372DhTU1MFAMKJEyc09vv4+AgbN27U2PfBBx8I7du3FwRBENasWSM4OzsLhYWF6uOrVq167Ln+yc/PT1i+fHmVx7/99lvBxcVF/XrdunUCAOHw4cPqfRcuXBAACEeOHBEEQRA6d+4sLF68WOM869evFzw9PdWvAQg//PBDle9LRLWLY2KIRObXX3+FnZ0dysvLUVZWhv79+2PlypXq435+fqhXr576dXJyMh48eAAXFxeN8xQXF+Pq1asAgAsXLmD8+PEax9u3b489e/Y8NoYLFy6gtLQUPXv21Dru7OxspKenIzIyEmPGjFHvLy8vV4+3uXDhAp577jnY2NhoxFFde/bsweLFi3H+/Hnk5+ejvLwcJSUlKCwshK2tLQDAwsICYWFh6jpNmjSBo6MjLly4gLZt2yI5ORnHjh3TaHlRKpUoKSlBUVGRRoxEZBxMYohEpnv37li1ahWkUim8vLwqDdx9+CX9kEqlgqenJ/bu3VvpXDWdZmxtbV3tOiqVCkBFl1K7du00jpmbmwMABD08jzYtLQ19+vTB+PHj8cEHH8DZ2RkHDhxAZGSkRrcbUDFF+lEP96lUKsyfPx8DBw6sVMbKykrnOIlId0xiiETG1tYWjRo10rp8SEgIMjMzYWFhgQYNGjy2TNOmTXH48GGMHDlSve/w4cNVnjMwMBDW1tbYtWsX3nrrrUrHLS0tAVS0XDzk7u4Ob29vXLt2DcOHD3/seZs1a4b169ejuLhYnSg9KY7HSUpKQnl5OZYtWwYzs4phf99++22lcuXl5UhKSkLbtm0BACkpKcjNzUWTJk0AVNy3lJSUat1rIjIsJjFEdVx4eDjat2+PAQMGYMmSJWjcuDFu3bqFbdu2YcCAAQgLC8O7776LN954A2FhYejUqRM2bNiAc+fOISAg4LHntLKywowZMxATEwNLS0t07NgR2dnZOHfuHCIjI+Hm5gZra2ts374d9evXh5WVFRwcHBAbG4uoqCjI5XK89NJLKC0tRVJSEnJycjBlyhRERERg9uzZiIyMxJw5c3D9+nV8/PHH1brehg0bory8HCtXrsTLL7+MgwcPYvXq1ZXKSaVSvPPOO/j0008hlUoxadIkPP/88+qk5v3330e/fv3g4+ODQYMGwczMDKdPn8aZM2ewcOHC6v8giEjvODuJqI6TSCTYtm0bunTpgtGjRyMoKAhDhw7F9evX1bOJhgwZgvfffx8zZsxAaGgo0tLS8Pbbbz/xvHPnzsXUqVPx/vvvo2nTphgyZAiysrIAVIw3+fTTT7FmzRp4eXmhf//+AIC33noL//73v5GQkICWLVuia9euSEhIUE/JtrOzwy+//ILz588jODgYs2fPxpIlS6p1va1bt0Z8fDyWLFmCFi1aYMOGDYiLi6tUzsbGBjNmzEBERATat28Pa2trbN68WX28d+/e+PXXX7Fz5060adMGzz//POLj4+Hn51eteIio9kgEfXRCExERERkYW2KIiIhIlJjEEBERkSgxiSEiIiJRYhJDREREosQkhoiIiESJSQwRERGJEpMYIiIiEiUmMURERCRKTGKIiIhIlJjEEBERkSgxiSEiIiJR+j/CFEiZyQW4+AAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "with warnings.catch_warnings():\n",
    "    warnings.simplefilter(\"ignore\")\n",
    "    lr = LogisticRegression(max_iter=1000).fit(x_train, y_train)\n",
    "    lr_res = lr.predict(x_test)\n",
    "\n",
    "get_statistics(y_test, lr_res)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "079f3e7e-a15b-49b8-83e3-9389b5519e14",
   "metadata": {},
   "source": [
    "These results are very encouraging, showing an almost excellent distinction between Greek and English and \"Others\" and a reliable classification of Greeklish."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b0ce1d27-89b9-4ca1-9063-79e677fe64ec",
   "metadata": {},
   "source": [
    "### Random Forest"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4fdde0b0-e8ac-4805-875e-fd567d6fee84",
   "metadata": {},
   "source": [
    "Random Forest is an ensemble algorithm, which means it uses many simpler algorithms which then \"vote\" on a final decision. It has proven to be a good classifier on complex tasks, it combats overfitting by design (essentially by utilizing random chance in its training phase) and is still fairly easy to interpret.\n",
    "\n",
    "The drawback is first and foremost computational, since we need to train many smaller classifiers, which may by themselves be computationally expensive (this is somewhat offset by the fact that the classifiers are indepednent and can be computed in parallel). Additionally, Random Forest is a non-parametric method which means that it is generally memory-intensive and may be slow to run on operational data. Finally, we also need to tune hyperparameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "c20afacc-038c-4ba4-94c5-532f3048670c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Random Forest mean macro F1: 0.9292, std: 0.9111\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "\n",
    "forest_model = RandomForestClassifier(n_estimators=500, \n",
    "                                      n_jobs=-1, \n",
    "                                      criterion=\"entropy\")\n",
    "res = cross_val_res(forest_model, x_train, y_train, cv=3)\n",
    "print(f\"Random Forest mean macro F1: {res[0]:.4f}, std: {res[1]:.4f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "11adac17-016c-4c15-8f84-71818e662763",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:    2.1s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:   10.9s\n",
      "[Parallel(n_jobs=-1)]: Done 434 tasks      | elapsed:   25.7s\n",
      "[Parallel(n_jobs=-1)]: Done 500 out of 500 | elapsed:   30.3s finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.0s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    0.1s\n",
      "[Parallel(n_jobs=8)]: Done 434 tasks      | elapsed:    0.4s\n",
      "[Parallel(n_jobs=8)]: Done 500 out of 500 | elapsed:    0.4s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.9120798978330256\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "          el       0.91      1.00      0.95      1012\n",
      "          en       0.99      1.00      0.99       942\n",
      "   greeklish       0.90      0.80      0.85        79\n",
      "       other       0.99      0.76      0.86       395\n",
      "\n",
      "    accuracy                           0.95      2428\n",
      "   macro avg       0.95      0.89      0.91      2428\n",
      "weighted avg       0.95      0.95      0.95      2428\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "forest_model = RandomForestClassifier(n_estimators=500, \n",
    "                                      n_jobs=-1, \n",
    "                                      criterion=\"entropy\", \n",
    "                                      verbose=1).fit(x_train, y_train)\n",
    "forest_pred = forest_model.predict(x_test)\n",
    "get_statistics(y_test, forest_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e068b646-cdba-4b71-8b4c-0cf88a99a5f1",
   "metadata": {},
   "source": [
    "The results are somewhat similar to our Logistic Regression classifier. The Cross-Validation score reports an equal on-average performance. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bfc319b9-5825-43be-88d5-707ed906d325",
   "metadata": {},
   "source": [
    "### Adaboost"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0365ddfa-7804-4f86-a8a6-1c5de37b9e0c",
   "metadata": {},
   "source": [
    "Adaboost is the logical conclusion of Random Forests, where each voter considers a very specific \"rule\" that needs to be followed. The next voter then considers the most important rule to distinguish between the categories for all the clases that the first could not reliably classify, and so on.\n",
    "\n",
    "This classifier is generally more compact and competent than a simple Random Forest, but is more computationally expensive during training because we cannot train it in parallel."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "372a74b3-5354-463a-9957-d153cfbae201",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AdaBoost mean macro F1: 0.8230, std: 0.7829\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import AdaBoostClassifier\n",
    "\n",
    "\n",
    "ada_model = AdaBoostClassifier(n_estimators=100)\n",
    "res = cross_val_res(ada_model, x_train, y_train, cv=3)\n",
    "print(f\"AdaBoost mean macro F1: {res[0]:.4f}, std: {res[1]:.4f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "285ec6f1-717a-48a8-8a8b-3d9e93caa531",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.8173169647781657\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "          el       1.00      0.87      0.93      1012\n",
      "          en       0.99      0.98      0.99       942\n",
      "   greeklish       0.80      0.42      0.55        79\n",
      "       other       0.68      0.98      0.80       395\n",
      "\n",
      "    accuracy                           0.92      2428\n",
      "   macro avg       0.87      0.81      0.82      2428\n",
      "weighted avg       0.94      0.92      0.92      2428\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.ensemble import AdaBoostClassifier\n",
    "\n",
    "\n",
    "ada_model = AdaBoostClassifier(n_estimators=100).fit(x_train, y_train)\n",
    "ada_pred = ada_model.predict(x_test)\n",
    "get_statistics(y_test, ada_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "59c69381-553e-4495-acc5-8b02549f0612",
   "metadata": {},
   "source": [
    "This classifier despite its complexity is even worse, being generally unable to distinguish Greeklish from \"Other\" languages and even placing many Greek comments to the latter category.\n",
    "\n",
    "This can be clearly seen by looking at the misclassification of adaboost compared to the misclassification of Random Forest."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "e185b12c-828e-4d51-a4dd-e47183ddb0eb",
   "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>predicted</th>\n",
       "      <th>actual</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1373</th>\n",
       "      <td>greeklish</td>\n",
       "      <td>en</td>\n",
       "      <td>The lights look beautiful!</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4411</th>\n",
       "      <td>other</td>\n",
       "      <td>el</td>\n",
       "      <td>Στην εταιρεία χρηματοπιστωτικών υπηρεσιών , η ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4060</th>\n",
       "      <td>other</td>\n",
       "      <td>el</td>\n",
       "      <td>Αυτός ήταν ο μπαμπάς μου .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2679752</th>\n",
       "      <td>other</td>\n",
       "      <td>greeklish</td>\n",
       "      <td>old expancion lordaeron server litch king</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>410</th>\n",
       "      <td>other</td>\n",
       "      <td>el</td>\n",
       "      <td>Περαστικά</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4193</th>\n",
       "      <td>other</td>\n",
       "      <td>el</td>\n",
       "      <td>6 % ως αποτέλεσμα απροσδιόριστο μόνιμων δράσεω...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1568</th>\n",
       "      <td>other</td>\n",
       "      <td>el</td>\n",
       "      <td>Αυτό κάνει ο καπιταλισμός της αγοράς .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2577</th>\n",
       "      <td>other</td>\n",
       "      <td>el</td>\n",
       "      <td>Εντάξει λοιπόν καλή σας νύχτα</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4773</th>\n",
       "      <td>other</td>\n",
       "      <td>el</td>\n",
       "      <td>Ο θείος της νταν ήταν δέκα χρόνια μεγαλύτερος ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7108</th>\n",
       "      <td>other</td>\n",
       "      <td>el</td>\n",
       "      <td>αλλά έκαναν κάτι λίγο διαφορετικό μαζί τους</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>202 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         predicted     actual  \\\n",
       "1373     greeklish         en   \n",
       "4411         other         el   \n",
       "4060         other         el   \n",
       "2679752      other  greeklish   \n",
       "410          other         el   \n",
       "...            ...        ...   \n",
       "4193         other         el   \n",
       "1568         other         el   \n",
       "2577         other         el   \n",
       "4773         other         el   \n",
       "7108         other         el   \n",
       "\n",
       "                                                      text  \n",
       "1373                            The lights look beautiful!  \n",
       "4411     Στην εταιρεία χρηματοπιστωτικών υπηρεσιών , η ...  \n",
       "4060                            Αυτός ήταν ο μπαμπάς μου .  \n",
       "2679752          old expancion lordaeron server litch king  \n",
       "410                                              Περαστικά  \n",
       "...                                                    ...  \n",
       "4193     6 % ως αποτέλεσμα απροσδιόριστο μόνιμων δράσεω...  \n",
       "1568                Αυτό κάνει ο καπιταλισμός της αγοράς .  \n",
       "2577                         Εντάξει λοιπόν καλή σας νύχτα  \n",
       "4773     Ο θείος της νταν ήταν δέκα χρόνια μεγαλύτερος ...  \n",
       "7108           αλλά έκαναν κάτι λίγο διαφορετικό μαζί τους  \n",
       "\n",
       "[202 rows x 3 columns]"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "missed = y_test != ada_pred\n",
    "pd.DataFrame({\"predicted\": ada_pred[missed], \"actual\": data_test[missed].labels, \"text\": data_test[missed].text})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "36893bd6-9b9f-4aa1-b2d3-07ecdbb70f3a",
   "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>predicted</th>\n",
       "      <th>actual</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9602</th>\n",
       "      <td>el</td>\n",
       "      <td>other</td>\n",
       "      <td>前にもセットで購入したのですがグリップ無くてもって思って弟にあげちゃいましたが、やっぱりあっ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9632</th>\n",
       "      <td>el</td>\n",
       "      <td>other</td>\n",
       "      <td>Syria \"zgłosić broń chemiczną i podpisać konwe...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10505</th>\n",
       "      <td>el</td>\n",
       "      <td>other</td>\n",
       "      <td>Pakistańscy talibowie wyznaczają nowego przywó...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9644</th>\n",
       "      <td>el</td>\n",
       "      <td>other</td>\n",
       "      <td>Оба окончательных правила были изданы кцб в со...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9118</th>\n",
       "      <td>el</td>\n",
       "      <td>other</td>\n",
       "      <td>Dali-schilderij gestolen uit galerie</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>533</th>\n",
       "      <td>en</td>\n",
       "      <td>greeklish</td>\n",
       "      <td>Warwick greek voice is actor is da best</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10503</th>\n",
       "      <td>el</td>\n",
       "      <td>other</td>\n",
       "      <td>2019.07.03にこちらで購入して、数回使用して故障しました！商品自体はよかったですが、...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10865</th>\n",
       "      <td>el</td>\n",
       "      <td>other</td>\n",
       "      <td>เอ่อ ทำไมคุณไม่เริ่มต้นล่ะ เพราะว่าคุณมีเวลามา...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10561</th>\n",
       "      <td>el</td>\n",
       "      <td>other</td>\n",
       "      <td>Hapana shaka rudisha .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10438</th>\n",
       "      <td>el</td>\n",
       "      <td>other</td>\n",
       "      <td>イヤーパッドが思ったより小さくて残念。音はこんなものか。おすすめではない。</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>118 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      predicted     actual                                               text\n",
       "9602         el      other  前にもセットで購入したのですがグリップ無くてもって思って弟にあげちゃいましたが、やっぱりあっ...\n",
       "9632         el      other  Syria \"zgłosić broń chemiczną i podpisać konwe...\n",
       "10505        el      other  Pakistańscy talibowie wyznaczają nowego przywó...\n",
       "9644         el      other  Оба окончательных правила были изданы кцб в со...\n",
       "9118         el      other               Dali-schilderij gestolen uit galerie\n",
       "...         ...        ...                                                ...\n",
       "533          en  greeklish            Warwick greek voice is actor is da best\n",
       "10503        el      other  2019.07.03にこちらで購入して、数回使用して故障しました！商品自体はよかったですが、...\n",
       "10865        el      other  เอ่อ ทำไมคุณไม่เริ่มต้นล่ะ เพราะว่าคุณมีเวลามา...\n",
       "10561        el      other                             Hapana shaka rudisha .\n",
       "10438        el      other              イヤーパッドが思ったより小さくて残念。音はこんなものか。おすすめではない。\n",
       "\n",
       "[118 rows x 3 columns]"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "missed = y_test != forest_pred\n",
    "pd.DataFrame({\"predicted\": forest_pred[missed], \"actual\": data_test[missed].labels, \"text\": data_test[missed].text})"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "389cc971-8bb2-4ecb-ab23-99239ac71308",
   "metadata": {},
   "source": [
    "### Hyperparameter tuning"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "10d50cd4-7e6d-4433-9a86-bc0506bbbf2c",
   "metadata": {},
   "source": [
    "Despite the allure of Logistic Regression's properties, we will stick to the more complex model of Random Forest, since we anticipate that the sample we procured for training and testing may not necessarily be very close to the actual operational data. We thus value the stability and robustness of a non-parametric method than the computational complexity and theoretical benefits of Logistic Regression.\n",
    "\n",
    "Since training Random Forest models is computationally intensive, we will only execute hyperparameter tuning on the most significant hyper-parameter; the number of trees which will vote during testing. This is also where our validation set comes into play."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "ce091a09-186f-4906-90fc-d1a5e0ead56a",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 10/10 [07:11<00:00, 43.19s/it]\n"
     ]
    }
   ],
   "source": [
    "estimators = []\n",
    "scores = []\n",
    "\n",
    "for n_estimators in tqdm([int(x) for x in np.linspace(start = 200, stop = 1200, num = 10)]):\n",
    "    estim = RandomForestClassifier(n_estimators=n_estimators, \n",
    "                                  n_jobs=-1, \n",
    "                                  criterion=\"entropy\").fit(x_train, y_train)\n",
    "    score = f1_score(y_val, estim.predict(x_val), average='macro', zero_division=0)\n",
    "\n",
    "    estimators.append(estim)\n",
    "    scores.append(score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "3ae6d6a4-2a6b-4bc8-8374-f20fc49ebd69",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best model RandomForestClassifier(criterion='entropy', n_estimators=977, n_jobs=-1) with macro F1 score of 0.9372640339591569\n"
     ]
    }
   ],
   "source": [
    "best_model = estimators[np.argmax(scores)]\n",
    "print(f\"Best model {best_model} with macro F1 score of {max(scores)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "de1fcfcc-e684-4e53-9253-0069aa838e77",
   "metadata": {},
   "source": [
    "### Annotating the operational dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2039b5bb-7839-4723-8bed-384763550524",
   "metadata": {},
   "source": [
    "We use our optimal classifier to identify the language of our operational dataset, which in this case are the crawled YouTube comments:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "36f59042-4429-4486-9a70-ea6af98c8c76",
   "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>title</th>\n",
       "      <th>link</th>\n",
       "      <th>source</th>\n",
       "      <th>text</th>\n",
       "      <th>date</th>\n",
       "      <th>language</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>ANTIGUAS CANCIONES DE GRECIA PAIS NATAL DE MIS...</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>other</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>Încă  o zi petrecută cu muzica voastră  fantas...</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>el</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>FelicităriSuperb \\nSă fiți mereu bine \\nMomen...</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>greeklish</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>Lovely collection! Thank you &lt;3</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>greeklish</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>Beautiful thank you  so many memories awesome ...</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>en</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2692</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Copy by fantaros</td>\n",
       "      <td>2021-11-29</td>\n",
       "      <td>el</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2693</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Fist</td>\n",
       "      <td>2021-11-29</td>\n",
       "      <td>el</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2694</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Yui</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>el</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2695</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Uihhii</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>el</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2696</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Zzfitdt</td>\n",
       "      <td>2021-11-29</td>\n",
       "      <td>el</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2359 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  title  \\\n",
       "0     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "1     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "2     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "3     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "4     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "...                                                 ...   \n",
       "2692  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2693  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2694  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2695  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2696  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "\n",
       "                                                   link  source  \\\n",
       "0     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "1     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "2     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "3     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "4     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "...                                                 ...     ...   \n",
       "2692  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2693  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2694  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2695  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2696  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "\n",
       "                                                   text        date   language  \n",
       "0     ANTIGUAS CANCIONES DE GRECIA PAIS NATAL DE MIS...  2022-11-29      other  \n",
       "1     Încă  o zi petrecută cu muzica voastră  fantas...  2022-11-29         el  \n",
       "2      FelicităriSuperb \\nSă fiți mereu bine \\nMomen...  2022-11-29  greeklish  \n",
       "3                       Lovely collection! Thank you <3  2022-11-29  greeklish  \n",
       "4     Beautiful thank you  so many memories awesome ...  2022-11-29         en  \n",
       "...                                                 ...         ...        ...  \n",
       "2692                                   Copy by fantaros  2021-11-29         el  \n",
       "2693                                               Fist  2021-11-29         el  \n",
       "2694                                                Yui  2022-11-29         el  \n",
       "2695                                             Uihhii  2022-11-29         el  \n",
       "2696                                            Zzfitdt  2021-11-29         el  \n",
       "\n",
       "[2359 rows x 6 columns]"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_oper = vectorizer.transform(crawl_df.text)\n",
    "crawl_df[\"language\"] = best_model.predict(x_oper)\n",
    "crawl_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ba7474bf-b737-44a1-8342-d95cf5132683",
   "metadata": {},
   "source": [
    "### Exploring the operational dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e5f918a-4afc-4a58-8ce9-79e9bd30d6ea",
   "metadata": {},
   "source": [
    "Now that we have annotated to the best of our abilities the operational dataset, it's time to run some quick analysis on our findings. \n",
    "\n",
    "The analysis and descriptions can be found in the `report.pdf` file. This notebook will only generate the graphs and Figures used there."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "b80cf5e2-fefe-4fa8-93f1-692b1bd45ced",
   "metadata": {},
   "outputs": [],
   "source": [
    "RESOURCE_OUTPUT = \"results\"\n",
    "\n",
    "def save_plot(filename):\n",
    "    path = os.path.join(RESOURCE_OUTPUT, filename)\n",
    "    plt.savefig(path, bbox_inches=\"tight\")\n",
    "    print(f\"Figured saved to \" + path)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6cbb7d94-3a2b-4369-8613-544d3b07adbc",
   "metadata": {},
   "source": [
    "#### Total language frequency"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "6ab7fb17-6863-4730-84c3-2852eada4bad",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define a common color palette for all graphs\n",
    "palette = {\"el\": \"blue\", \"en\": \"red\", \"greeklish\": \"green\", \"other\": \"black\"}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "7e0777af-dc21-4f23-a6e7-c1078af74c78",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\user\\AppData\\Local\\Temp\\ipykernel_32116\\3670115791.py:5: FutureWarning: \n",
      "\n",
      "Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `y` variable to `hue` and set `legend=False` for the same effect.\n",
      "\n",
      "  sns.barplot(crawl_df.language, palette=palette, legend=False)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Figured saved to results\\lang_dis.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "\n",
    "\n",
    "# I really don't have time to fix this warning, sorry :(\n",
    "sns.barplot(crawl_df.language, palette=palette, legend=False)\n",
    "plt.title(\"Post languages\")\n",
    "plt.xlabel(\"Number of observed comments\")\n",
    "\n",
    "save_plot(\"lang_dis.png\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0ad51153-8014-4fba-b936-7c7c98b67839",
   "metadata": {},
   "source": [
    "#### Average length of comments by language"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "d7036f2c-bfdb-4d71-b049-8fbbf8ef375b",
   "metadata": {},
   "outputs": [],
   "source": [
    "languages = np.unique(crawl_df.language)\n",
    "\n",
    "lang_col = []\n",
    "len_col = []\n",
    "\n",
    "for language in languages:\n",
    "    comments_text = crawl_df.loc[crawl_df.language == language, \"text\"]\n",
    "    comments_length = comments_text.apply(lambda x: len(x))\n",
    "    \n",
    "    lang_col += [language] * len(comments_length)\n",
    "    len_col += list(comments_length)\n",
    "\n",
    "len_df = pd.DataFrame({\"language\": lang_col, \"comment_length\": len_col})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "df6b95b4-0e0f-44ac-bd39-add2a9067390",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 613.125x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.displot(len_df, x=\"comment_length\", hue=\"language\", multiple=\"stack\", palette=palette)\n",
    "plt.title(\"Stacked plot of comment length (characters) by language\")\n",
    "plt.xlabel(\"Length of comment (characters)\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "79794708-f426-4754-9f2d-914eea3adacc",
   "metadata": {},
   "source": [
    "The wild variations between the main distribution and its long tail make parsing the graph very difficult. We will thus split it in two graphs, one containing the main body of the distribution and the other the long tail."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "f541276f-6d2d-4400-a8bd-518bae837223",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Figured saved to results\\length_dis.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax1, ax2) = plt.subplots(1,2)\n",
    "sns.histplot(len_df[len_df.comment_length<=250],\n",
    "             x=\"comment_length\", \n",
    "             hue=\"language\", \n",
    "             multiple=\"stack\", \n",
    "             palette=palette,\n",
    "             ax=ax1)\n",
    "ax1.set_xlabel(\"Length of comment (characters < 250)\")\n",
    "ax1.set_ylabel(\"Number of comments\")\n",
    "\n",
    "sns.histplot(len_df[len_df.comment_length>250], \n",
    "             x=\"comment_length\", \n",
    "             hue=\"language\", \n",
    "             multiple=\"stack\",\n",
    "             palette=palette,\n",
    "             ax=ax2)\n",
    "ax2.set_xlabel(\"Length of comment (characters)\")\n",
    "ax2.set_ylabel(\"Number of comments\")\n",
    "\n",
    "fig.suptitle(\"Stacked plot of long and short comments length by language\")\n",
    "save_plot(\"length_dis.png\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4ecd8abc-b093-47b6-904c-0a8b88b503aa",
   "metadata": {},
   "source": [
    "#### Emoji usage by language"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "8d0046d1-6546-4678-9a37-7fd5f9539658",
   "metadata": {},
   "outputs": [],
   "source": [
    "emoji_pattern = re.compile(\n",
    "    r'[\\U0001F600-\\U0001F64F\\U0001F300-\\U0001F5FF\\U0001F680-\\U0001F6FF'\n",
    "    r'\\U0001F700-\\U0001F77F\\U0001F780-\\U0001F7FF\\U0001F800-\\U0001F8FF'\n",
    "    r'\\U0001F900-\\U0001F9FF\\U0001FA00-\\U0001FA6F\\u2600-\\u26FF\\u2700-\\u27BF'\n",
    "    r'\\u2B50\\u2B06\\u2934\\u2935\\u2B05\\u2194-\\u2199\\u21A9\\u21AA\\u2139\\u2328'\n",
    "    r'\\u23CF\\u23E9-\\u23F3\\u231A\\u23F8-\\u23FA\\u231B\\u23F0\\u231A\\u1F004'\n",
    "    r'\\u1F0CF\\u1F18E\\u3030\\u303D]', flags=re.UNICODE\n",
    ")\n",
    "\n",
    "emojis_col = []\n",
    "\n",
    "for language in languages:\n",
    "    comments_text = crawl_df.loc[crawl_df.language == language, \"text\"]\n",
    "    comments_length = comments_text.apply(lambda x: len(emoji_pattern.findall(x)))\n",
    "\n",
    "    emojis_col += list(comments_length)\n",
    "\n",
    "emoji_df = pd.DataFrame({\"language\": lang_col, \"emojis\": emojis_col})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "cda75427-7fc2-4580-8044-1de1c5bdf218",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 613.125x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.displot(emoji_df[emoji_df.emojis > 0],\n",
    "            x=\"emojis\", \n",
    "            hue=\"language\",\n",
    "            multiple=\"stack\",\n",
    "            palette=palette)\n",
    "plt.title(\"Stacked plot of emojis used in comments by language\")\n",
    "plt.xlabel(\"Number of emojis\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ad7538b4-2c6d-467b-8598-59bd53aec809",
   "metadata": {},
   "source": [
    "We will perform the same operation as above for the same reasons."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "aa5b1104-b84b-4cf5-8ec2-06a7532618c3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Figured saved to results\\emojis_dis.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax1, ax2) = plt.subplots(1, 2)\n",
    "sns.histplot(emoji_df[(emoji_df.emojis > 0) & (emoji_df.emojis < 10)],\n",
    "             x=\"emojis\",\n",
    "             hue=\"language\",\n",
    "             multiple=\"stack\",\n",
    "             palette=palette,\n",
    "             ax=ax1)\n",
    "ax1.set_xlabel(\"Number of emojis (<10)\")\n",
    "ax1.set_ylabel(\"Number of comments\")\n",
    "\n",
    "sns.histplot(emoji_df[emoji_df.emojis > 10],\n",
    "             x=\"emojis\", \n",
    "             hue=\"language\", \n",
    "             multiple=\"stack\",\n",
    "             palette=palette,\n",
    "             ax=ax2)\n",
    "ax2.set_xlabel(\"Number of emojis (>10)\")\n",
    "ax2.set_ylabel(\"Number of comments\")\n",
    "\n",
    "fig.suptitle(\"Stacked plot of emojis used in comments by language\")\n",
    "save_plot(\"emojis_dis.png\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "140968c1-3c38-41ee-8ba3-b329318d04ae",
   "metadata": {},
   "source": [
    "#### Observed language usage through time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "09c4fd84-0c84-4152-949b-6f1f2bfda69c",
   "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></th>\n",
       "      <th>title</th>\n",
       "      <th>link</th>\n",
       "      <th>source</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>date</th>\n",
       "      <th>language</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">2018-11-29</th>\n",
       "      <th>el</th>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>en</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"3\" valign=\"top\">2019-11-29</th>\n",
       "      <th>el</th>\n",
       "      <td>49</td>\n",
       "      <td>49</td>\n",
       "      <td>49</td>\n",
       "      <td>49</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>en</th>\n",
       "      <td>38</td>\n",
       "      <td>38</td>\n",
       "      <td>38</td>\n",
       "      <td>38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>greeklish</th>\n",
       "      <td>231</td>\n",
       "      <td>231</td>\n",
       "      <td>231</td>\n",
       "      <td>231</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"4\" valign=\"top\">2023-11-28</th>\n",
       "      <th>el</th>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>en</th>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>greeklish</th>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>other</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2023-11-29</th>\n",
       "      <th>greeklish</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>107 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                      title  link  source  text\n",
       "date       language                            \n",
       "2018-11-29 el             3     3       3     3\n",
       "           en             1     1       1     1\n",
       "2019-11-29 el            49    49      49    49\n",
       "           en            38    38      38    38\n",
       "           greeklish    231   231     231   231\n",
       "...                     ...   ...     ...   ...\n",
       "2023-11-28 el             5     5       5     5\n",
       "           en             3     3       3     3\n",
       "           greeklish      2     2       2     2\n",
       "           other          1     1       1     1\n",
       "2023-11-29 greeklish      1     1       1     1\n",
       "\n",
       "[107 rows x 4 columns]"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "date_df = crawl_df.groupby([\"date\", \"language\"]).count()\n",
    "date_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "9ddf42fe-128e-419c-a764-90a0dd7d9985",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0     2018-11-29\n",
       "1     2018-11-29\n",
       "2     2019-11-29\n",
       "3     2019-11-29\n",
       "4     2019-11-29\n",
       "         ...    \n",
       "102   2023-11-28\n",
       "103   2023-11-28\n",
       "104   2023-11-28\n",
       "105   2023-11-28\n",
       "106   2023-11-29\n",
       "Name: date, Length: 107, dtype: datetime64[ns]"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "date_df2 = date_df.reset_index()\n",
    "date_df2.date = pd.to_datetime(date_df2.date)\n",
    "date_df2.date"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "f1fb70c8-acd3-47af-aa91-23dcc4b5f874",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Figured saved to results\\time_plot.png\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjsAAAGpCAYAAABrkPeOAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy81sbWrAAAACXBIWXMAAA9hAAAPYQGoP6dpAACdT0lEQVR4nOzdd1jTVxcH8G+AEPbegoCKE3DvgXtvrQMr2lrH697WWhXqqlpX3a2tG9xbq+LCXUcd4MCFW0QQ2YRA7vvHJYEwlGBCIJzP8+SR/PJLckiQHO499x4BY4yBEEIIIURL6Wg6AEIIIYQQdaJkhxBCCCFajZIdQgghhGg1SnYIIYQQotUo2SGEEEKIVqNkhxBCCCFajZIdQgghhGg1SnYIIYQQotUo2SGEEEKIVqNkh+Rp06ZNEAgE8ouBgQEcHBzQokULLFiwAFFRUYV+7Pv378Pf3x/Pnz9XXcBqcPr0adSpUwfGxsYQCAQ4cOBAkTyvQCDA6NGji+y5/P395dc/994EBgZi+fLlaonDzc0NgwcPVstjlybHjh1TeD9VoSh+Hp8/fw6BQIBNmzap9XlI6UXJDvmsjRs34sqVKwgODsbq1atRo0YNLFy4EFWqVMGpU6cK9Zj3799HQEBAsU52GGPo06cPhEIhDh06hCtXrsDHx0fTYanclStX8MMPP8ivf+69UWeyQ1Tj2LFjCAgI0HQYhBQ7epoOgBRvnp6eqFOnjvx6r169MGHCBDRp0gQ9e/bE48ePYW9vr8EI1ePt27f4+PEjevTogVatWmk6HLVp0KCBpkMghBC1o5EdorSyZctiyZIlSEhIwPr16+XHb9y4gX79+sHNzQ2GhoZwc3ND//798eLFC/k5mzZtwjfffAMAaNGihXyaLPvw9alTp9CqVSuYmZnByMgIjRs3xunTpxVi+PDhA4YNGwYXFxeIRCLY2tqicePGBRptunjxIlq1agVTU1MYGRmhUaNGOHr0qPx2f39/ODs7AwCmTZsGgUAANze3fB9PKpVi7ty5qFSpEgwNDWFhYQFvb2+sWLFCfs7gwYPzfAx/f38IBII8H3f9+vWoWLEiRCIRqlatih07dijcLptqPHPmDIYOHQpra2uYmZnBz88PSUlJiIyMRJ8+fWBhYQFHR0dMnjwZEolE4TGyT2N97r1p3rw5jh49ihcvXihMb8qkpaVh7ty5qFy5svz9+O677/DhwweF55NIJJg6dSocHBxgZGSEJk2a4Nq1a/m+tjmJxWL88ssvqFKlCgwMDGBtbY0WLVrg8uXL8nNSU1Mxffp0uLu7Q19fH2XKlMGoUaPw6dMnhcdyc3ND586dceTIEdSsWROGhoaoUqUKjhw5In89qlSpAmNjY9SrVw83btxQuP/gwYNhYmKChw8fol27djA2NoajoyN+/fVXAMDVq1fRpEkTGBsbo2LFiti8eXOu7ycyMhLDhw+Hs7Mz9PX14e7ujoCAAKSnp8vPkU3x/Pbbb1i6dCnc3d1hYmKChg0b4urVqwrxrF69GgAU3iPZKN3u3btRv359mJubw8jICOXKlcP3339f4Nf+cz+Pz58/h56eHhYsWJDrfufPn4dAIMDu3bsL/FwA8OTJE3z33Xfw8PCAkZERypQpgy5duiA0NFThvHPnzkEgECAoKAgzZsyAk5MTzMzM0Lp1a4SHhyucyxjD/Pnz4erqCgMDA9SpUwfBwcFo3rw5mjdvLj9P9n8r5win7LnOnTsnPxYcHIxu3brB2dkZBgYGqFChAoYPH47o6Ohc39PBgwfh7e0NkUiEcuXKYcWKFXn+DmCMYc2aNahRowYMDQ1haWmJ3r1749mzZ0q9hiQbRkgeNm7cyACw69ev53l7YmIi09XVZa1atZIf2717N5s1axbbv38/CwkJYTt27GA+Pj7M1taWffjwgTHGWFRUFJs/fz4DwFavXs2uXLnCrly5wqKiohhjjG3dupUJBALWvXt3tm/fPnb48GHWuXNnpqury06dOiV/rnbt2jFbW1v2xx9/sHPnzrEDBw6wWbNmsR07dnz2+zp37hwTCoWsdu3abOfOnezAgQOsbdu2TCAQyO/76tUrtm/fPgaAjRkzhl25coX9999/+T7mggULmK6uLps9ezY7ffo0O378OFu+fDnz9/eXnzNo0CDm6uqa676zZ89mOf8bAmAuLi6satWqLCgoiB06dIi1b9+eAWC7d++Wnyd7j9zd3dmkSZPYyZMn2cKFC5muri7r378/q1WrFps7dy4LDg5m06ZNYwDYkiVLcj3X7Nmzv/je3Lt3jzVu3Jg5ODjIj1+5coUxxlhGRgZr3749MzY2ZgEBASw4OJht2LCBlSlThlWtWpUlJycrvA4CgYBNmTKFnTx5ki1dupSVKVOGmZmZsUGDBn32vZNIJKxFixZMT0+PTZ48mR07dowdOnSI/fTTTywoKIgxxphUKmXt2rVjenp6bObMmezkyZPst99+Y8bGxqxmzZosNTVV/niurq7M2dmZeXp6sqCgIHbs2DFWv359JhQK2axZs1jjxo3Zvn372P79+1nFihWZvb19ru9FX1+fValSha1YsYIFBwez7777jgFg06dPZxUrVmR//fUXO3HiBOvcuTMDwG7cuCG//7t375iLiwtzdXVl69evZ6dOnWJz5sxhIpGIDR48WH5eREQEA8Dc3NxY+/bt2YEDB9iBAweYl5cXs7S0ZJ8+fWKMMfbkyRPWu3dvBkDhPUpNTWWXL19mAoGA9evXjx07doydOXOGbdy4kQ0cOPCzr7kyP489evRgZcuWZenp6Qr3/+abb5iTkxOTSCT5Pofse9y4caP8WEhICJs0aRLbs2cPCwkJYfv372fdu3dnhoaG7OHDh/Lzzp49K399BgwYwI4ePcqCgoJY2bJlmYeHh0I806dPZwDYsGHD2PHjx9mff/7JypYtyxwdHZmPj4/8PNn/rYiICIU4Zc919uxZ+bG1a9eyBQsWsEOHDrGQkBC2efNmVr16dVapUiWWlpYmP++ff/5hOjo6rHnz5mz//v1s9+7drH79+szNzS3X74ChQ4cyoVDIJk2axI4fP84CAwNZ5cqVmb29PYuMjMz3dST5o2SH5OlLyQ5jjNnb27MqVarke3t6ejpLTExkxsbGbMWKFfLju3fvzvULgzHGkpKSmJWVFevSpYvC8YyMDFa9enVWr149+TETExM2fvx4Jb8rxho0aMDs7OxYQkKCQpyenp7M2dmZSaVSxljWL9/Fixd/8TE7d+7MatSo8dlzlE12DA0NFX6ppaens8qVK7MKFSrIj8neozFjxijcv3v37gwAW7p0qcLxGjVqsFq1auV6Llmyw1j+7w1jjHXq1CnP7yEoKIgBYHv37lU4fv36dQaArVmzhjHG2IMHDxgANmHCBIXztm/fzgB8MdnZsmULA8D+/PPPfM85fvw4A8AWLVqkcHznzp0MAPvjjz/kx1xdXZmhoSF7/fq1/Njt27cZAObo6MiSkpLkxw8cOMAAsEOHDsmPDRo0KNf3LZFImK2tLQOgkCDHxMQwXV1dNnHiRPmx4cOHMxMTE/bixQuFWH/77TcGgN27d48xlvWz6OXlpfDBfe3aNQZAnugxxtioUaNy/Txlf0xZYqSMgv48yhKB/fv3y4+9efOG6enpsYCAgM8+R17JTk7p6eksLS2NeXh4KPwMyZ63Y8eOCufv2rVLnvgxxtjHjx+ZSCRiffv2VTjvypUrDEChk53spFIpk0gk7MWLFwwAO3jwoPy2unXrMhcXFyYWi+XHEhISmLW1tcJ7Josn5x8mr169YoaGhmzq1Kn5vkYkfzSNRQqNMaZwPTExEdOmTUOFChWgp6cHPT09mJiYICkpCQ8ePPji412+fBkfP37EoEGDkJ6eLr9IpVK0b98e169fR1JSEgCgXr162LRpE+bOnYurV6/mmp7JS1JSEv7991/07t0bJiYm8uO6uroYOHAgXr9+nWvYuyDq1auHO3fuYOTIkThx4gTi4+OVfoycWrVqpVALpauri759++LJkyd4/fq1wrmdO3dWuF6lShUAQKdOnXIdzz6lqCpHjhyBhYUFunTpovC+1ahRAw4ODvIh/7NnzwIABgwYoHD/Pn36QE/vy+WD//zzDwwMDD479XLmzBkAyLWy65tvvoGxsXGu6dAaNWqgTJky8uuy16558+YwMjLKdTzn6ycQCNCxY0f5dT09PVSoUAGOjo6oWbOm/LiVlRXs7OwU7n/kyBG0aNECTk5OCq9bhw4dAAAhISEKz9WpUyfo6urKr3t7e+cZU17q1q0LgL/Wu3btwps3b754n+wK8vPYvHlzVK9eXT6VBgDr1q2DQCDAsGHDlHo+AEhPT8f8+fNRtWpV6OvrQ09PD/r6+nj8+HGev0+6du2qcD3n63P16lWIxWL06dNH4bwGDRp8dpr6S6KiojBixAi4uLhAT08PQqEQrq6uACCPMykpCTdu3ED37t2hr68vv6+JiQm6dOmi8HhHjhyBQCDAt99+q/Bz4eDggOrVqytMoZGCo2SHFEpSUhJiYmLg5OQkP+br64tVq1bhhx9+wIkTJ3Dt2jVcv34dtra2SElJ+eJjvn//HgDQu3dvCIVChcvChQvBGMPHjx8BADt37sSgQYOwYcMGNGzYEFZWVvDz80NkZGS+jx8bGwvGGBwdHXPdJvs+YmJilHodAGD69On47bffcPXqVXTo0AHW1tZo1apVrhoPZTg4OOR7LGeMVlZWCtdlv0zzOp6amlromPLz/v17fPr0Cfr6+rnet8jISHntgizunN+bnp4erK2tv/g8Hz58gJOTE3R08v+1FRMTAz09Pdja2iocFwgEcHBw+KrXDkCu18/IyAgGBga5zs15f9nx7Pd///49Dh8+nOs1q1atGgDkqvnI+RqJRCIAKND/rWbNmuHAgQNIT0+Hn58fnJ2d4enpiaCgoC/eFyj4z+PYsWNx+vRphIeHQyKR4M8//0Tv3r3zvP+XTJw4ETNnzkT37t1x+PBh/Pvvv7h+/TqqV6+e5/f8pddHFmdeCyoKu8hCKpWibdu22LdvH6ZOnYrTp0/j2rVr8loq2XPLfvcU5Lnfv38vPzfnz8bVq1fzrAUiX0arsUihHD16FBkZGfKivri4OBw5cgSzZ8/Gjz/+KD9PLBbLE5QvsbGxAQCsXLky31VCsl8MNjY2WL58OZYvX46XL1/i0KFD+PHHHxEVFYXjx4/neV9LS0vo6Ojg3bt3uW57+/atQgzK0NPTw8SJEzFx4kR8+vQJp06dwk8//YR27drh1atX8g9EsVic6775/eLKK2mTHStIYlCUbGxsYG1tne/rbmpqCiAr7sjISIXRlPT09AIlmba2trh48SKkUmm+CY+1tTXS09Px4cMHhYSHMYbIyEj5CEdxYGNjA29vb8ybNy/P27P/IaEK3bp1Q7du3SAWi3H16lUsWLAAvr6+cHNzQ8OGDT9734L+PPr6+mLatGlYvXo1GjRogMjISIwaNapQ8W7btg1+fn6YP3++wvHo6GhYWFgo/XiyOGV/VGUXGRmpMLojS2Bz/p/N+f81LCwMd+7cwaZNmzBo0CD58SdPniicZ2lpCYFAkO9zZ2djYwOBQIALFy7IE7bs8jpGvoxGdojSXr58icmTJ8Pc3BzDhw8HwP9yZozl+o+4YcMGZGRkKBzL7y/Sxo0bw8LCAvfv30edOnXyvGQfApYpW7YsRo8ejTZt2uC///7LN25jY2PUr18f+/btU3huqVSKbdu2wdnZGRUrVlTuxcjBwsICvXv3xqhRo/Dx40f5ag43NzdERUUp/LJLS0vDiRMn8nyc06dPK5ybkZGBnTt3onz58vKVYurwudECkUiU5/HOnTsjJiYGGRkZeb5nlSpVAgB5Yrx9+3aF++/atUth9VF+OnTogNTU1M9uPCfbJmDbtm0Kx/fu3YukpKRitY1A586dERYWhvLly+f5uhUm2SnIaI9IJIKPjw8WLlwIALh169YXH7egP48GBgYYNmwYNm/ejKVLl6JGjRpo3Lix0t8HwH+n5Px9cvToUaWn4GTq168PkUiEnTt3Khy/evVqrqlAWeJz9+5dheOHDh3KFSOQOwHJvkoV4L976tSpgwMHDiAtLU1+PDExUb76T6Zz585gjOHNmzd5/lx4eXkV8Dsm2dHIDvmssLAw+ZxxVFQULly4gI0bN0JXVxf79++X//VsZmaGZs2aYfHixbCxsYGbmxtCQkLw119/5forzNPTEwDwxx9/wNTUFAYGBnB3d4e1tTVWrlyJQYMG4ePHj+jduzfs7Ozw4cMH3LlzBx8+fMDatWsRFxeHFi1awNfXF5UrV4apqSmuX7+O48ePo2fPnp/9fhYsWIA2bdqgRYsWmDx5MvT19bFmzRqEhYUhKCgo32Xgn9OlSxf5fkS2trZ48eIFli9fDldXV3h4eAAA+vbti1mzZqFfv36YMmUKUlNT8fvvv+dKBGVsbGzQsmVLzJw5E8bGxlizZg0ePnyYa/m5qn3uvfHy8sK+ffuwdu1a1K5dGzo6OqhTpw769euH7du3o2PHjhg3bhzq1asHoVCI169f4+zZs+jWrRt69OiBKlWq4Ntvv8Xy5cshFArRunVrhIWF4bfffoOZmdkXY+vfvz82btyIESNGIDw8HC1atIBUKsW///6LKlWqoF+/fmjTpg3atWuHadOmIT4+Ho0bN8bdu3cxe/Zs1KxZEwMHDlTr66eMX375BcHBwWjUqBHGjh2LSpUqITU1Fc+fP8exY8ewbt06pRNb2QfhwoUL0aFDB+jq6sLb2xtz587F69ev0apVKzg7O+PTp09YsWIFhEJhgTbLVObnceTIkVi0aBFu3ryJDRs2KBV/dp07d8amTZtQuXJleHt74+bNm1i8eHGhk30rKytMnDgRCxYsgKWlJXr06IHXr18jICAAjo6OCqOFdevWRaVKlTB58mSkp6fD0tIS+/fvx8WLFxUes3Llyihfvjx+/PFHMMZgZWWFw4cPIzg4ONfz//LLL+jUqRPatWuHcePGISMjA4sXL4aJiYnC6Hfjxo0xbNgwfPfdd7hx4waaNWsGY2NjvHv3DhcvXoSXlxf+97//Feo1KNU0VhpNijXZagTZRV9fn9nZ2TEfHx82f/58+VLx7F6/fs169erFLC0tmampKWvfvj0LCwtjrq6uuVbaLF++nLm7uzNdXd08l5x26tSJWVlZMaFQyMqUKcM6deokX+aamprKRowYwby9vZmZmRkzNDRklSpVYrNnz1ZYQZOfCxcusJYtWzJjY2NmaGjIGjRowA4fPqxwjjKrsZYsWcIaNWrEbGxsmL6+PitbtiwbMmQIe/78ucJ5x44dYzVq1GCGhoasXLlybNWqVfmuxho1ahRbs2YNK1++PBMKhaxy5cps+/btCuflt2JO9piy5f4ygwYNYsbGxrmeK/tqLMbyf28+fvzIevfuzSwsLJhAIFCIWyKRsN9++41Vr16dGRgYMBMTE1a5cmU2fPhw9vjxY/l5YrGYTZo0idnZ2TEDAwPWoEEDduXKlTx/RvKSkpLCZs2axTw8PJi+vj6ztrZmLVu2ZJcvX1Y4Z9q0aczV1ZUJhULm6OjI/ve//7HY2FiFx3J1dWWdOnXK9Ryy1z+7vH4e8no9GWPMx8eHVatWLdfxvJ7vw4cPbOzYsczd3Z0JhUJmZWXFateuzWbMmMESExPzfe7ssWZ//8RiMfvhhx+Yra2t/D2KiIhgR44cYR06dGBlypSR/1/u2LEju3DhQq7HzO/1+NLPY3bNmzdnVlZWCkv1Pyev1VixsbFsyJAhzM7OjhkZGbEmTZqwCxcuMB8fH4WVU7IVUtmXwef3mFKplM2dO5c5OzszfX195u3tzY4cOcKqV6/OevTooXD/R48esbZt2zIzMzNma2vLxowZw44ePZprNdb9+/dZmzZtmKmpKbO0tGTffPMNe/nyZZ7/t/bv38+8vLzkvyd+/fVXNnbsWGZpaZnrNfn7779Z/fr15b+nypcvz/z8/BS2LyAFJ2Asx5IaQgghpJCioqLg6uqKMWPGYNGiRZoO54siIiJQuXJlzJ49Gz/99FORPrdEIpGvCDx58mSRPndpQ9NYhBBCvtrr16/x7NkzLF68GDo6Ohg3bpymQ8rlzp07CAoKQqNGjWBmZobw8HAsWrQIZmZmGDJkiNqff8iQIWjTpg0cHR0RGRmJdevW4cGDBwq7rRP1oGSHEELIV9uwYQN++eUXuLm5Yfv27Qor7ooLY2Nj3LhxA3/99Rc+ffoEc3NzNG/eHPPmzSuSHn8JCQmYPHkyPnz4AKFQiFq1auHYsWNo3bq12p+7tKNpLEIIIYRoNVp6TgghhBCtRskOIYQQQrQa1eyAbyr39u1bmJqaFmqfFUIIIYQUPcYYEhISvthKhpId8FYBLi4umg6DEEIIIYXw6tWrz244SckOsnr3vHr1qkA7uRJCCCFE8+Lj4+Hi4iL/HM8PJTvI6m9iZmZGyQ4hhBBSwnypBIUKlAkhhBCi1SjZIYQQQohWo2SHEEIIIVqNkh1CCCGEaDVKdgghhBCi1SjZIYQQQohWo2SHEEIIIVqNkh1CCCGEaDVKdgghhBCi1SjZIYQQQohWo2SHEEIIIVqNemMRQggpVRhj+O/df9h6dytiUmKwrtM6GOsbazosokaU7BBCCCkVXsW9wvbQ7dh6dyvuf7gvPz64+mC0KtdKg5ERdaNkhxBCiNZKECdg34N92Hp3K85EnAEDAwCIdEXQ1dFFsiQZyZJkDUdJ1I2SHUIIIVolQ5qB0xGnsfXuVux7sE8hmWnm2gx+3n7oXbU3uu7oivMvziM1PVWD0ZKiQMkOIYQQrRD6PhRb7mxBYFgg3ia8lR/3sPKAX3U/fOv9Ldws3OTHRboiAIA4Q1zUoZIiRskOKfGuvLqCiScnwtfTF32q9YG9ib2mQyKEFJHIxEgEhQZhy90tuB15W37cytAK/ar1g191P9QrUw8CgSDXfQ30DACARnZKAUp2SIkXGBqIq6+v4urrqxh/YjxaubeCr5cvelTuAXMDc02HRwhRsRRJCg6GH8SWO1tw8ulJZLAMAIBQR4jOFTvDr7ofOnp0hL6u/mcfR5bsiNNpZEfbUbJDSryfmv6EClYVEBgWiGtvriH4WTCCnwVjxJER6FyxM/p79kenip3kv9gIISWPlElx/sV5bL2zFbvv70ZCWoL8tgbODeDn7Yc+1frA2si6wI8p0uPTWDSyo/0o2SElnqOpI8Y1GIdxDcbhyccn2BG2A9tDt+Nh9EPsfbAXex/shZnIDD2r9ISvpy9auLeAng796BNSEoRHh2Pr3a3YencrXsa9lB93s3DDQO+B+Nb7W1S0rlioxzbQzRzZoZodrUe/8YlWqWBVAT83+xkzms7Anfd3EBQahKCwILyKf4VNtzdh0+1NsDO2Q99qfeHr5Yv6ZernOZdPCNGc6ORo7AzbiS13t+Dam2vy42YiM/Sp2gd+1f3QuGxj6Ai+rgkAjeyUHpTsEK0kEAhQw6EGajjUwILWC3Dp5SUEhgZi9/3diEqKwsprK7Hy2kq4W7jD18sX/T37o5pdNU2HTUipJU4X48ijI9h6dyuOPj6KdGk6AEBXoIv2FdrDr7ofulTsAkOhocqekwqUSw9KdojW0xHooKlrUzR1bYrfO/yO4GfBCAwNxIGHBxDxKQLzLszDvAvz4G3vDV9PX/Tz7AdXC1dNh02I1mOM4errq9hyZwt23tuJ2NRY+W21HGvBz9sP/Tz7qW2FpXzpORUoaz1KdkipItQVoqNHR3T06IiktCQcfnQYgaGBOP7kOO6+v4u77+/ix9M/oknZJujv2R/fVP0Gtsa2mg6bEK3yLPYZtt3dhq13t+LJxyfy42VMy+Bb728x0HtgkYy00shO6UHJDim1jPWN0c+zH/p59sPHlI/Ye38vAsMCEfI8BBdfXsTFlxcx9p+xaFu+LXy9fNGtUjeYikw1HTYhJdKn1E/YfW83ttzdgosvL8qPGwuN0atqL/h5+6G5W3Po6ugWWUyymh0qUNZ+lOwQAr4B2dDaQzG09lC8jn+NnWE7ERQWhJvvbuKfJ//gnyf/wFDPEF0qdYGvpy/aV2gv/0VJCMmbJEOCE09PYMudLTgUfkieVAggQOtyreFX3Q89KvfQWMdxGtkpPSjZISQHZzNnTGo0CZMaTUJ4dDiCwoIQGBqIxx8fY9e9Xdh1bxcsDCzQu0pv+Hr5oplrsyL9a5SQ4owxhpvvbmLrna0ICgvCh+QP8tuq2VbDoOqD4OvlizJmZTQYJSffVJBGdrQeJTuEfEYlm0rwb+6P2T6zcfPdTQSFBmHHvR14m/AWG25twIZbG+Bo4oh+nv3g6+WL2o61aSk7KZVexb3C9tDt2HJnCx5EP5AftzO2wwCvARjoPRA1HGoUq/8fsgJlGtnRfpTsEFIAAoEAdZzqoI5THSxqswjnX5xHYGgg9jzYg3eJ77Ds6jIsu7oMHlYe6O/ZH75evqhkU0nTYROiVgniBOx7sA9b7m7B2YizYGAA+IhJ98rdMdB7INqWb1tsN/GkdhGlh4AxxjQdhKbFx8fD3NwccXFxMDMz03Q4pAQRp4tx4ukJBIYG4lD4IaSkp8hvq+VYC76evujr2RfOZs4ajJIQ1cmQZuB0xGlsubMF+x7sU/iZ93H1gV91P/Sq0qtE9KU78PAAeuzsgYbODXF5yGVNh0MKoaCf38Uz3SakhBDpidC1Uld0rdQVCeIEHAw/iMDQQJx8ehL/vfsP/737D1OCp8DHzQf9Pfujd9XesDK00nTYhCgt9H0ottzZgu2h2/Eu8Z38eEXrivDz9sMA7wFws3DTXICFQAXKpQeN7IBGdojqfUj6gD339yAwLFBhma1QR4j2FdrD18sXXSp20dgqFEIKIjIxEoGhgdhyZwvuvL8jP25laIX+nv0x0Hsg6pWpV6zqcJRxNuIsWm5piaq2VXFv5D1Nh0MKoaCf35TsgJIdol4vPr3Azns7ERgaqPCBYSw0RrfK3eDr6Yu25dtCqCvUYJSEcMmSZBx8eBBb7m7ByacnIWVSADxR71KpC/y8/dDBowP0dfU1HOnXu/LqChr93QjlLMvh6dinmg6HFAIlO0qgZIcUlXtR9+RL2SM+RciPWxta45uq36C/V380KdvkqxscEqIMKZPi/Ivz2HJnC/bc34OEtAT5bQ2dG8Kvuh/6VOujdVOw/737D7X/qI0ypmXweuJrTYdDCoGSHSVQskOKGmMM195cQ2BoIHbe24n3Se/ltzmbOctXdFW3r15ipwhI8fcw+iG23tmKbaHb8DLupfy4u4U7BnoPxLfe38LD2kODEarX/Q/3UW1NNVgbWiN6arSmwyGFQMmOEijZIZqULk3H2YizCAwLxL4H+xAvjpffVtmmMnw9fdHfqz8qWFXQYJREW0QnR2NH2A5subMF199elx83F5mjT7U+8Kvuh8YujUtFkv0s9hnK/14eJvomSJie8OU7kGKHkh0lULJDiovU9FQce3wMgaGBOPLoiMLOrvXK1IOvpy/6VOsDR1NHDUZJShpxuhhHHh3BlrtbcOzxMaRL0wEAugJddPDogIHeA9GlYhcYCg01HGnRehP/Bs7LnKGnowfJTImmwyGFQMmOEijZIcVRXGoc9j/cj6CwIJx6dkpeKKoj0EELtxbw9fJFzyo9YWFgodlASbHEGMOV11ew5c4W7Lq3C7GpsfLbajvWxkDvgejv1R92xnYajFKzYpJjYLPYBgCQPjOd2r6UQJTsKIGSHVLcvU98j133diEwLBBXX1+VH9fX1Ucnj07w9fJFJ49Ope4vc5Lbs9hn2HpnK7be3YqnsVkrjJzNnPGt17cYWH0gqtpW1WCExUdiWiJMF5gCAJJ+SoKR0EjDERFlUbKjBEp2SEnyLPYZdoTtwPbQ7bj/4b78uKm+KXpU6QFfT1+0Kteq2G7RT1QvNiUWu+/vxpY7W3Dp1SX5cWOhMXpX7Y2B3gPR3K05jVzkkC5Nh3AO3/IhZmqM1q02Kw0o2VECJTukJGKMITQqFIGhgQgKC1JYTWNrZIs+1frA18sXDZ0blopi09JGkiHB8SfHseXuFhwOPyyv79IR6KB1udbw8/ZD98rdaePKL9D9RRdSJsXbiW+pFq4EomRHCZTskJJOyqS48uoKAkMDsev+LkQnZy2jdbNwky9l97Tz1GCU5GsxxnDz3U1subMFQWFBCu+zp50nBlUfBF8vXziZOmkwypLFeL4xkiXJeDb2Gdwt3TUdDlESJTtKoGSHaBNJhgSnnp1CYFggDjw8gMS0RPltnnae8qXsJa2PUWn2Mu4ltt/dji13t+Bh9EP5cXtjewzwGoCB1QfSnkyFZLXQCrGpsXgw6gEq21TWdDhESZTsKIGSHaKtkiXJOPLoCAJDA3Hs8TFIpFnLaxu5NIKvpy++qfZNqV6RU1wliBOw98FebLmzBeeenwMD/1VtoGeAHpV7YKD3QLQp34Zqs76S0xInvEt8h1vDb6GGQw1Nh0OURMmOEijZIaVBbEos9j7Yi6CwIJyNOCv/8NQV6KJ1udbw9fJF98rdYSai/wOFlSHNQLIkGUmSJCSlJSl8/bljSZLcx2+9u4WU9BT5Yzd3a46B3gPRu2pveo9UqNyKcoj4FIGrQ66ivnN9TYdDlFTQz2/6k4CQUsLS0BI/1PoBP9T6AW8T3mJn2E4EhgXixtsbOPH0BE48PQEDPQN0qdgF/T37o4NHBxjoGWg6bJVijCElPUWpRER+LGdikkeykpqeqtJ4K1lXkrdtcLVwVeljE06kJwIAlb93pHihkR3QyA4p3R7HPJY3Jw2PCZcfNxeZo1eVXvD18i2yZcuMMUikkrxHPQp6LL9Rk8zERTaipU4CCGAkNIKxvjGMhcYw1jfm1zO/NhbyS85zsh9zs3BDTYeaVIejZjXX18TtyNs48e0JtC3fVtPhECXRNJYSKNkhhCcatyJvITA0EDvCduBNwhv5bQ4mDuhbrS98vXxRy7GWQvLwueTiS4lIXscyWEaRfL8GegZfTDhkSUmeyUo+x4yFxjDQM6AkpYRosKEB/n3zLw72O4iulbpqOhyiJJrGIoQoRSAQoJZjLdRyrIVFbRbhwosLCAwNxO77uxGZGIkV/67Ain9XFFk8ejp6X04ulExEZMeMhEa0wR4BAPlULU1jaTdKdgghuegIdODj5gMfNx+s7LgSJ56cQGBYIA6FH0KyJBkAn6r5YnKh94VRk88cE+oKNfwqkNJAVrMjThd/4UxSklGyQwj5LH1dfXSp1AVdKnVBanoqEtMSaaqGaA0a2SkdKNkhhBSYgZ6B1q3QIqWbSDdzZCeDRna0mY6mAyCEEEI0hUZ2SgeNJjsLFixA3bp1YWpqCjs7O3Tv3h3h4eEK5zDG4O/vDycnJxgaGqJ58+a4d++ewjlisRhjxoyBjY0NjI2N0bVrV7x+/boovxVCCCElkCzZoZod7abRZCckJASjRo3C1atXERwcjPT0dLRt2xZJSUnycxYtWoSlS5di1apVuH79OhwcHNCmTRskJCTIzxk/fjz279+PHTt24OLFi0hMTETnzp2RkVE0S1gJIYSUTLJpLBrZ0W4ardk5fvy4wvWNGzfCzs4ON2/eRLNmzcAYw/LlyzFjxgz07NkTALB582bY29sjMDAQw4cPR1xcHP766y9s3boVrVu3BgBs27YNLi4uOHXqFNq1a1fk3xchhJCSQT6yQzU7Wq1Y1ezExcUBAKysrAAAERERiIyMRNu2WbtaikQi+Pj44PLlywCAmzdvQiKRKJzj5OQET09P+Tk5icVixMfHK1wIIYSUPtQuonQoNskOYwwTJ05EkyZN4OnpCQCIjIwEANjb2yuca29vL78tMjIS+vr6sLS0zPecnBYsWABzc3P5xcXFRdXfDiGEkBKACpRLh2KT7IwePRp3795FUFBQrtty7uXBGPvi/h6fO2f69OmIi4uTX169elX4wAkhhJRYtPS8dCgWyc6YMWNw6NAhnD17Fs7OzvLjDg4OAJBrhCYqKko+2uPg4IC0tDTExsbme05OIpEIZmZmChdCCCGlD43slA4aTXYYYxg9ejT27duHM2fOwN3dXeF2d3d3ODg4IDg4WH4sLS0NISEhaNSoEQCgdu3aEAqFCue8e/cOYWFh8nMIIYSQvFC7iNJBo6uxRo0ahcDAQBw8eBCmpqbyERxzc3MYGhpCIBBg/PjxmD9/Pjw8PODh4YH58+fDyMgIvr6+8nOHDBmCSZMmwdraGlZWVpg8eTK8vLzkq7MIIYSQvNDITumg0WRn7dq1AIDmzZsrHN+4cSMGDx4MAJg6dSpSUlIwcuRIxMbGon79+jh58iRMTU3l5y9btgx6enro06cPUlJS0KpVK2zatAm6utTVmBBCSP5o6XnpIGCMMU0HoWnx8fEwNzdHXFwc1e8QQkgpcvDhQXTf2R0NnBvgypArmg6HKKmgn9/FokCZEEII0QRqF1E6ULJDCCGk1KJNBUsHSnYIIYSUWlSzUzpQskMIIaTUokagpQMlO4QQQkotWnpeOlCyQwghpNSiTQVLB0p2CCGElFo0slM6ULJDCCGk1JIlOxksA+nSdA1HQ9SFkh1CCCGllqxAGaCpLG1GyQ4hhJBSS1azA9Dyc21GyQ4hhJBSS09HD7oC3keR6na0FyU7hBBCSjVqGaH9KNkhhBBSqlHLCO1HyQ4hhJBSjZafaz9KdgghhJRqshVZVKCsvSjZIYQQUqrRyI72o2SHEEJIqUYFytqPkh1CCCGlGhUoaz9KdgghhJRq8pEdqtnRWpTsEEIIKdVkBco0sqO9KNkhhBBSqlHNjvajZIcQQkipRjU72o+SHUIIIaUaLT3Xfl+d7MTHx+PAgQN48OCBKuIhhBBCihRtKqj9lE52+vTpg1WrVgEAUlJSUKdOHfTp0wfe3t7Yu3evygMkhBBC1IlGdrSf0snO+fPn0bRpUwDA/v37wRjDp0+f8Pvvv2Pu3LkqD5AQQghRJypQ1n5KJztxcXGwsrICABw/fhy9evWCkZEROnXqhMePH6s8QEIIIUSdaOm59lM62XFxccGVK1eQlJSE48ePo23btgCA2NhYGBgYqDxAQgghRJ1oU0Htp6fsHcaPH48BAwbAxMQErq6uaN68OQA+veXl5aXq+AghhBC1oqXn2k/pZGfkyJGoX78+Xr58iTZt2kBHhw8OlStXDvPmzVN5gIQQQog60ciO9lN6GuuXX35BlSpV0KNHD5iYmMiPt2zZEqdOnVJpcIQQQoi6Uc2O9lM62QkICEBiYmKu48nJyQgICFBJUIQQQkhRoaXn2k/pZIcxBoFAkOv4nTt35Ku0CCGEkJJCVrNDS8+1V4FrdiwtLSEQCCAQCFCxYkWFhCcjIwOJiYkYMWKEWoIkhBBC1IVGdrRfgZOd5cuXgzGG77//HgEBATA3N5ffpq+vDzc3NzRs2FAtQRJCCCHqQgXK2q/Ayc6gQYMAAO7u7mjUqBGEQqHagiKEEEKKChUoaz+ll577+PhAKpXi0aNHiIqKglQqVbi9WbNmKguOEEIIUTdqF6H9lE52rl69Cl9fX7x48QKMMYXbBAIBMjIyVBYcIYQQom60qaD2UzrZGTFiBOrUqYOjR4/C0dExz5VZhBBCSElBNTvaT+lk5/Hjx9izZw8qVKigjngIIYSQIkU1O9pP6X126tevjydPnqgjFkIIIaTI0dJz7af0yM6YMWMwadIkREZGwsvLK9eqLG9vb5UFV9xkZGRAIpFoOgyiBKFQCF1dXU2HQQgpxmQ1O1ImRbo0HXo6Sn80kmJO6Xe0V69eAIDvv/9efkwgEMh3VtbGAmXGGCIjI/Hp0ydNh0IKwcLCAg4ODlRfRgjJk2xkB+CjOyb6Jp85m5RESic7ERER6oijWJMlOnZ2djAyMqIPzRKCMYbk5GRERUUBABwdHTUcESGkOJLV7AB8+TklO9pH6WTH1dVVHXEUWxkZGfJEx9raWtPhECUZGhoCAKKiomBnZ0dTWoSQXHR1dKGno4d0aTrV7WgppQuUAWDr1q1o3LgxnJyc8OLFCwC8ncTBgwdVGlxxIKvRMTIy0nAkpLBk7x3VWxFC8kPLz7Wb0snO2rVrMXHiRHTs2BGfPn2S1+hYWFhg+fLlqo6v2KCpq5KL3jtCyJfQ8nPtpnSys3LlSvz555+YMWOGwpRAnTp1EBoaqtLgCCGEkKJALSO0m9LJTkREBGrWrJnruEgkQlJSkkqCIoQQQooStYzQbkonO+7u7rh9+3au4//88w+qVq2qiphKpebNm2P8+PGaDoMQQkol2lhQuym9GmvKlCkYNWoUUlNTwRjDtWvXEBQUhAULFmDDhg3qiJEQQghRK1nNDhUoayelk53vvvsO6enpmDp1KpKTk+Hr64syZcpgxYoV6NevnzpiJIQQQtSKRna0W6GWng8dOhQvXrxAVFQUIiMj8erVKwwZMkTpxzl//jy6dOkCJycnCAQCHDhwQOH2wYMHQyAQKFwaNGigcI5YLMaYMWNgY2MDY2NjdO3aFa9fvy7Mt1VsbNu2DXXq1IGpqSkcHBzg6+sr3xgPAM6dOweBQIDTp0+jTp06MDIyQqNGjRAeHq7wOHPnzoWdnR1MTU3xww8/4Mcff0SNGjXkt+c1dda9e3cMHjy4wLEAwKFDh+Dh4QFDQ0O0aNECmzdvhkAgUNhx+vLly2jWrBkMDQ3h4uKCsWPHUo0XIaTYoAJl7VaoZEfGxsYGdnZ2hb5/UlISqlevjlWrVuV7Tvv27fHu3Tv55dixYwq3jx8/Hvv378eOHTtw8eJFJCYmonPnziW6bUVaWhrmzJmDO3fu4MCBA4iIiFBIQGRmzJiBJUuW4MaNG9DT01No4bF9+3bMmzcPCxcuxM2bN1G2bFmsXbtW5bE8f/4cvXv3Rvfu3XH79m0MHz4cM2bMUHiM0NBQtGvXDj179sTdu3exc+dOXLx4EaNHj1Y6HkIIUQcqUNZyTEnR0dFs5MiRrEqVKsza2ppZWloqXAoLANu/f7/CsUGDBrFu3brle59Pnz4xoVDIduzYIT/25s0bpqOjw44fP17g546Li2MAWFxcXK7bUlJS2P3791lKSkqBH68wfHx82Lhx4/K87dq1awwAS0hIYIwxdvbsWQaAnTp1Sn7O0aNHGQB5nPXr12ejRo1SeJzGjRuz6tWrf/Y5u3XrxgYNGpRvnDljmTZtGvP09FQ4Z8aMGQwAi42NZYwxNnDgQDZs2DCFcy5cuMB0dHTU/royVnTvISGk5Oq+ozuDP9j6G+s1HQpRwuc+v7NTumbn22+/xdOnTzFkyBDY29urfcO2c+fOwc7ODhYWFvDx8cG8efPko0k3b96ERCJB27Zt5ec7OTnB09MTly9fRrt27fJ8TLFYDLE4a6gyPj5erd+Dsm7dugV/f3/cvn0bHz9+hFQqBQC8fPlSYcVb9g7zsr5PUVFRKFu2LMLDwzFy5EiFx61Xrx7OnDmj0ljCw8NRt27dXM+T3c2bN/HkyRNs375dfowxBqlUioiICFSpUkWpmAghRNVoU0HtpnSyc/HiRVy8eBHVq1dXRzwKOnTogG+++Qaurq6IiIjAzJkz0bJlS9y8eRMikQiRkZHQ19eHpaWlwv3s7e0RGRmZ7+MuWLAAAQEB6g6/UJKSktC2bVu0bdsW27Ztg62tLV6+fIl27dohLS1N4VyhUCj/WpZ0ypKR7MdkGGMK13V0dHIdy95SoSCxsMxu9597HqlUiuHDh2Ps2LG5vt+yZcvm/UIQQkgRopod7aZ0slO5cmWkpKSoI5Zc+vbtK//a09MTderUgaurK44ePYqePXvme7+8PoCzmz59OiZOnCi/Hh8fDxcXF9UE/ZUePnyI6Oho/Prrr/KYbty4ofTjVKpUCdeuXcPAgQPlx3I+jq2tLd69eye/npGRgbCwMLRo0aLAsVSuXDlXHVXOc2rVqoV79+6hQoUKSn8fhBBSFGhkR7spXaC8Zs0azJgxAyEhIYiJiUF8fLzCRZ0cHR3h6uqKx48fAwAcHByQlpaG2NhYhfOioqJgb2+f7+OIRCKYmZkpXIqLsmXLQl9fHytXrsSzZ89w6NAhzJkzR+nHGTNmDP766y9s3rwZjx8/xty5c3H37l2FJLBly5Y4evQojh49iocPH2LkyJEKK6gKEsvw4cPx8OFDTJs2DY8ePcKuXbuwadMmAFkjS9OmTcOVK1cwatQo3L59G48fP8ahQ4cwZswY5V8gQghRA2oEqt2UTnYsLCwQFxeHli1bws7ODpaWlrC0tISFhUWu6SRVi4mJwatXr+T1KbVr14ZQKERwcLD8nHfv3iEsLAyNGjVSayzqYmtri02bNmH37t2oWrUqfv31V/z2229KP86AAQMwffp0TJ48GbVq1ZKvojIwMJCf8/3332PQoEHw8/ODj48P3N3d5aM6BY3F3d0de/bswb59++Dt7Y21a9fKV2OJRPwvJW9vb4SEhODx48do2rQpatasiZkzZ8rfR0II0TRajaXdBCxngcUX1KtXD3p6ehg3blyeBco+Pj4FfqzExEQ8efIEAFCzZk0sXboULVq0gJWVFaysrODv749evXrB0dERz58/x08//YSXL1/iwYMHMDU1BQD873//w5EjR7Bp0yZYWVlh8uTJiImJwc2bNxUalX5OfHw8zM3NERcXl2uUJzU1FREREXB3d1dIFEqiNm3awMHBAVu3blXr88ybNw/r1q3Dq1ev1Po8BaVN7yEhRD1mnZ2FOefnYFTdUVjVMf/tUEjx8rnP7+yUrtkJCwvDrVu3UKlSpa8KEOC1HdlHEmR1NIMGDcLatWsRGhqKLVu24NOnT3B0dESLFi2wc+dOeaIDAMuWLYOenh769OmDlJQUtGrVCps2bSpwoqOtkpOTsW7dOrRr1w66uroICgrCqVOnFEbBVGXNmjWoW7curK2tcenSJSxevJj20CGElChUoKzdlE526tSpg1evXqkk2WnevHmulTvZnThx4ouPYWBggJUrV2LlypVfHY82EQgEOHbsGObOnQuxWIxKlSph7969aN26tcqfS1YT9PHjR5QtWxaTJk3C9OnTVf48hBCiLvIC5QyaxtJGSic7Y8aMwbhx4zBlyhR4eXkpLH8GFPd+IZpjaGiIU6dOFclzLVu2DMuWLSuS5yKEEHWgkR3tpnSyI1sOnr01gUAgkC/3LsltGgghhJROVKCs3ZROdiIiItQRByGEEKIxtPRcuymd7Li6uqojDkIIIURjaFNB7aZ0sgMAb968waVLlxAVFaXQngBAni0BCCGEkOKMana0m9LJzsaNGzFixAjo6+vD2tpaYZ8dgUBAyQ4hhJASh2p2tJvSyc6sWbMwa9YsTJ8+HTo6Sm/ATAghhBQ7spEdSna0k9LZSnJyMvr160eJTgl37tw5CAQChV5YhBBSWslqdqhAWTspnbEMGTIEu3fvVkcshBBCiEbQyI52U3oaa8GCBejcuTOOHz+e56aCS5cuVVlwhBBCSFGgAmXtpnSyM3/+fJw4cULeLiJngbK2YwxITtbMcxsZAcq8xIwxLF68GOvWrcO7d+9QsWJFzJw5E71791ZfkIQQUgJRgbJ2UzrZWbp0Kf7++28MHjxYDeEUf8nJgImJZp47MREwNi74+T///DP27duHtWvXwsPDA+fPn8e3334LW1tb9QVJCCElUPZNBWUdAYj2UDrZEYlEaNy4sTpiISqUlJSEpUuX4syZM2jYsCEAoFy5crh48SLWr1+PYcOGaThCQggpPmQFylImRbo0HUJd4RfuQUoSpZOdcePGYeXKlfj999/VEU+xZ2TER1g09dwFdf/+faSmpqJNmzYKx9PS0lCzZk0VR0YIISWbbGQH4KM7lOxoF6WTnWvXruHMmTM4cuQIqlWrlqtAed++fSoLrjgSCJSbStIU2c7WR48eRZkyZRRuE4lEePr0qSbCIoSQYklWswPwuh0TfQ3VKxC1UDrZsbCwQM+ePdURC1GhqlWrQiQS4eXLl/Dx8cl1OyU7hBCSRUegA6GOEBKphIqUtVCh2kWQ4s/U1BSTJ0/GhAkTIJVK0aRJE8THx+Py5cswMTGhhq6EEJKDSE8ESZqElp9roUI1AgWADx8+IDw8HAKBABUrVqQVPsXQnDlzYGdnhwULFuDZs2ewsLBArVq18NNPP+Vq4EoIIaWdgZ4BEtMSaWRHCymd7CQlJWHMmDHYsmWL/ANTV1cXfn5+WLlyJYyUqaIlaiVrzJpfc1bGWBFHRAghxVf25edEuyjdLmLixIkICQnB4cOH8enTJ3z69AkHDx5ESEgIJk2apI4YCSGEELWTLT+nkR3to/TIzt69e7Fnzx40b95cfqxjx44wNDREnz59sHbtWlXGRwghhBQJahmhvQrV9dze3j7XcTs7OyRrqo8CIYQQ8pWoZYT2UjrZadiwIWbPno3U1KwfhpSUFAQEBMh36iWEEEJKGqrZ0V5KT2OtWLEC7du3h7OzM6pXrw6BQIDbt2/DwMAAJ06cUEeMhBBCiNpRzY72UjrZ8fT0xOPHj7Ft2zY8fPgQjDH069cPAwYMgKGhoTpiJIQQQtRONrJDyY72KdQ+O4aGhhg6dKiqYyGEEEI0RlazQwXK2kfpmp0FCxbg77//znX877//xsKFC1USFCGEEFLUaGRHeymd7Kxfvx6VK1fOdbxatWpYt26dSoIihBBCihoVKGsvpZOdyMhIODo65jpua2uLd+/eqSQoQggpzVJSgLdvNR1F6UMFytpL6WTHxcUFly5dynX80qVLcHJyUklQhBBSGjEGbN8OODsDZcoA5csDI0YAe/YAHz9qOjrtR5sKai+lC5R/+OEHjB8/HhKJBC1btgQAnD59GlOnTqV2EYQQUkjv3vHE5tChrGPPngHr1/OLQADUrg20bg20aQM0agQYGGguXm1EIzvaS+mRnalTp2LIkCEYOXIkypUrh3LlymHMmDEYO3Yspk+fro4YixfGgKQkzVyUbNzJGMOiRYtQrlw5GBoaonr16tizZw8A4Ny5cxAIBDh9+jTq1KkDIyMjNGrUCOHh4ep41Qgh+WAM2LwZqFqVJzpCITBnDhATw6+PHctvYwy4cQP49VegVSvAygpo1w5YvBi4dQvI7MtMvgLV7GgvpUd2BAIBFi5ciJkzZ+LBgwcwNDSEh4cHRCKROuIrfpKTARMTzTx3YiJgbFzg03/++Wfs27cPa9euhYeHB86fP49vv/0Wtra28nNmzJiBJUuWwNbWFiNGjMD333+f5zQlIUT1Xr8Ghg8Hjh3j12vXBjZuBLy8+PUuXfgFAN68AU6fBoKDgVOngMhI4ORJfgEAGxueBLVpw0d/XF2L/vsp6ahdhPYq1D47AGBiYoK6deuqMhaiQklJSVi6dCnOnDkjb+NRrlw5XLx4EevXr8ewYcMAAPPmzYOPjw8A4Mcff0SnTp2QmpoKAxofJ0RtGAP+/huYOBGIjwf09YGAAGDyZEAvn9/KZcoAfn78whhw7x5Pek6dAs6dA6KjgZ07+QUAPDyyprxatAAsLIrquyu5aOm59ip0slNqGRnxERZNPXcB3b9/H6mpqWjTpo3C8bS0NNSsWVN+3dvbW/61bJVdVFQUypYt+5XBEkLy8vIlMHRo1ohM/fp8NKdKlYI/hkAAeHryy/jxQFoa8O+/WaM+164Bjx/zy9q1gI4OUKdO1qhPw4ZAaRmMV4asZoemsbQPJTvKEgiUmkrSFGnmBP7Ro0dRpkwZhdtEIhGePn0KABAKhfLjAoFA4b6EENVhDPjjDz56k5jIi4vnzAEmTAB0db/usfX1gaZN+eWXX4C4OD7ac+oUT4DCw3kCdO0aMG8e/7upWbOs5MfLi/9qK+1oZEd7UbKjpapWrQqRSISXL1/Kp6mykyU7hBD1i4gAfvgBOHOGX2/cmE9jVayonuczNwe6deMXAHj1KmvK69QpICoKOH6cXwDAzo4nPbJpL2dn9cRV3NHSc+1VoGSnVq1aOH36NCwtLfHLL79g8uTJMFJiSoUUPVNTU0yePBkTJkyAVCpFkyZNEB8fj8uXL8PExASuVL1IiNpJpXwaado0vqDS0BCYPx8YM+brR3OU4eICfPcdv0ilQFhY1qhPSAhPfgID+QUAKlXKGvVp3pwnT6UBFShrrwIlOw8ePEBSUhIsLS0REBCAESNGULJTAsyZMwd2dnZYsGABnj17BgsLC9SqVQs//fQTTVURomZPnwJDhvBkAuDTRn/9BVSooNm4dHQAb29+mTgREIuBK1eykp8bN/i0V3g4sGoVT8rq1csa9alfn0+baSNaeq69BIx9efOWhg0bwsTEBE2aNEFAQAAmT54Mk3yWX8+aNUvlQapbfHw8zM3NERcXBzMzM4XbUlNTERERAXd3d1qhVELRe0iKklQKrFwJTJ/O2z4YG/O9cUaO5IlGcRcby+t9ZMXOjx8r3m5szEd7ZMlP1araU+8T/DQYbbe1hbe9N+6MuKPpcEgBfO7zO7sCjexs2rQJs2fPxpEjRyAQCPDPP/9AL4/1kQKBoEQmO4QQogqPHgHffw/Itqpq2RLYsAFwd9dsXMqwtAR69OAXAHjxImvU5/RpvsT96FF+AQAHh6zEp3VroCR3DaKaHe1VoGSnUqVK2LFjBwBAR0cHp0+fhp2dnVoDI4SQkiIjA1i2DJg5E0hN5fuO/vYbMGxYyR/1cHXl03FDhvBRq7t3s0Z9zp/nmxtu28YvAB/pkSU/Pj6Aqalm41cG1exoL6VXY1GtByGEZHnwgBf+/vsvv96mDfDnn9q5g7GODlCjBr9MmcITu8uXs5KfmzeB+/f55fff+QaJDRpkrfSqV4+3wyiuaOm59irU0vOnT59i+fLlePDgAQQCAapUqYJx48ahfPnyqo6PEEKKpfR0Pnrj78+LfM3MgKVL+TRWSR/NKSgDAz5V17IlsGAB78x+5kzWtNezZ8DFi/zi789HeZo3z5ryqly5eL1WtKmg9lI62Tlx4gS6du2KGjVqoHHjxmCM4fLly6hWrRoOHz6ca8deQgjRNmFhfDTnxg1+vUMHvmFgad2fRsbKCujdm18AnuzI9vY5fZonQ4cP8wvAW2DIRn1at+b1P5pEIzvaq0CrsbKrWbMm2rVrh19//VXh+I8//oiTJ0/iv//+U2mARYFWY2k3eg+JqkgkwMKFfJdiiYT3m1q+nPerKk4jFMVRRgZw+3bWqM/Fi3xELDtPz6xRn2bNir7n8vvE93BY4gABBMiYlSHfVZ4UXwVdjaV0smNgYIDQ0FB4eHgoHH/06BG8vb2RmlryMmJKdrQbvYdEFe7c4aM5t27x6126AOvWlezVR5qUksITHtnIT86/k4VC3sNLlvzUqZN/k1RV+ZT6CZYLLQEA4p/F0NfV0g2FtEhBkx2ld32wtbXF7du3cx2/ffs2rdAihGidtDReb1KnDk90rKyA7duBgwcp0fkahoY8kVm4kBc2f/jAO7b/8AMv7pZI+GqvmTN50mNjw5fDr17Nl/gr92d6wcimsQBafq5tlE52hg4dimHDhmHhwoW4cOECLl68iF9//RXDhw/HsGHD1BEjKWE2bdoECwuLfG8/d+4cBAIBPn36VKDzlXlsQlTpv/+AunWBgABekNyzJ3DvHuDrS9NWqmZjA/Tpw1eyRURkdWzv2ZNPF8bFAQcOAKNH83YWsiXxQUG83YUqZB/Jobod7aL0oODMmTNhamqKJUuWYPr06QAAJycn+Pv7Y+zYsSoPkGi/vn37omPHjpoOgxA5sZh3JP/1V15rYmPDRxS++YaSnKIgEPC2GhUqACNG8Pfgv/+ylrhfusSbm/79N78AQPXqWfv7NG3KO7srS0egA31dfaRlpNGKLC2jdLIjEAgwYcIETJgwAQkJCQB400lS8qWlpUFfA01vDA0NYWhoWOTPS0herl/ntTn37vHrffrwHlG2tpqNqzTT1eUjbHXrAj/9xJuqXryYlfzcuZN1WbKE9+5q3DhrlVft2gVvvCrSFSEtI41GdrTMV3VqMTU1pUSnGEtISMCAAQNgbGwMR0dHLFu2DM2bN8f48eMBAG5ubpg7dy4GDx4Mc3NzDB06FABw+fJlNGvWDIaGhnBxccHYsWORlJQkf9y0tDRMnToVZcqUgbGxMerXr49z587lG0dMTAzq1auHrl275lnAnnNq6s6dO2jRogVMTU1hZmaG2rVr44ZsjW+mEydOoEqVKjAxMUH79u3x7t27wr9QhIBvkPfjj3wTvHv3ADs7YM8eXkdCiU7xYmwMtGvH9zm6fZvv4hwYyPc4cnHhdVZnzwIzZvDGpTY2QK9evKD8yZPP1/vQ8nPtVALa0hUvjDEkpSVp5KLkwjlMnDgRly5dwqFDhxAcHIwLFy7k2hpg8eLF8PT0xM2bNzFz5kyEhoaiXbt26NmzJ+7evYudO3fi4sWLGD16tPw+3333HS5duoQdO3bg7t27+Oabb9C+fXs8ztkxEMDr16/RtGlTVK5cGfv27SvQaqgBAwbA2dkZ169fx82bN/Hjjz9CmG3b1eTkZPz222/YunUrzp8/j5cvX2Ly5MlKvTaEZHflClCzJi+WlUp5Tc69e/wDkhR/9vZA//68q/yLF1kd27t355s9fvoE7NsH/O9/gIcHUK4cMHQoT2RjYhQfS9YyggqUtYuaF/J93vnz57F48WLcvHkT7969w/79+9G9e3f57YwxBAQE4I8//kBsbCzq16+P1atXo1q1avJzxGIxJk+ejKCgIKSkpKBVq1ZYs2YNnNW0u1eyJBkmC4p484dMidMTYaxvXKBzExISsHnzZgQGBqJVq1YAgI0bN8Ipx/KRli1bKiQKfn5+8PX1lY/+eHh44Pfff4ePjw/Wrl2LN2/eICgoCK9fv5Y/1uTJk3H8+HFs3LgR8+fPlz/Wo0eP0KZNG3Tr1g0rVqwo8J4VL1++xJQpU1C5cmV5DNlJJBKsW7dOvmP36NGj8csvvxTosQnJLjmZr/ZZtoz/te/gAKxfD3TtqunISGEJBEDFivwyahQvLL9xI2t/nytXgOfPeYPWDRv4KFFYGODmxu9PIzvaSaMjO0lJSahevTpWrVqV5+2LFi3C0qVLsWrVKly/fh0ODg5o06aNvFYIAMaPH4/9+/djx44duHjxIhITE9G5c2dkZGQU1bdRLD179gwSiQT16tWTHzM3N0elSpUUzqtTp47C9Zs3b2LTpk0wMTGRX9q1awepVIqIiAj8999/YIyhYsWKCueEhITg6dOn8sdJSUlBkyZN0L17d/z+++9Kbc41ceJE/PDDD2jdujV+/fVXhccFACMjI4XWJI6OjohS1XIMUmpcuMCLWpcu5YmOnx8fzaFER7vI+nP9/DMQEsJ3cT56FJgwga/ySkritT4y8s7nVKCsVZQa2ZFIJGjbti3Wr1+PihUrfvWTd+jQAR06dMjzNsYYli9fjhkzZqBnz54AgM2bN8Pe3h6BgYEYPnw44uLi8Ndff2Hr1q1o3bo1AGDbtm1wcXHBqVOn0K5duzwfWywWQ5xt6874+PgCx2wkNELi9MQCn69KRsKCLy+QTXnlTDJyToUZGyuOFEmlUgwfPjzPlXVly5bF3bt3oauri5s3b0I3R8WfSbbtTkUiEVq3bo2jR49iypQpSo20+fv7w9fXF0ePHsU///yD2bNnY8eOHejRowcAKExpyb5HZaf4SOmVlMSLXFeu5ElOmTJ8NKdTJ01HRoqCiQnQsSO/PHrEE5/o6KzbZf2xaGRHuyiV7AiFQoSFhRXJFtoRERGIjIxE27Zt5cdEIhF8fHxw+fJlDB8+HDdv3pQnYDJOTk7w9PTE5cuX8012FixYgICAgELFJRAICjyVpEnly5eHUCjEtWvX4OLiAoAndY8fP4aPj0++96tVqxbu3buHChUq5Hl7zZo1kZGRgaioKDRt2jTfx9HR0cHWrVvh6+uLli1b4ty5c7mm0D6nYsWKqFixIiZMmID+/ftj48aN8mSHkMI6d47vzfLsGb/+/fd89Q5t3VQ62djwf7MnO/KRHarZ0SpKT2P5+fnhr7/+UkcsCiIjIwEA9vb2Csft7e3lt0VGRkJfXx+Wlpb5npOX6dOnIy4uTn559eqViqPXPFNTUwwaNAhTpkzB2bNnce/ePXz//ffQ0dH5bLI6bdo0XLlyBaNGjcLt27fx+PFjHDp0CGPGjAHAk5ABAwbAz88P+/btQ0REBK5fv46FCxfi2LFjCo+lq6uL7du3o3r16mjZsuVn3xOZlJQUjB49GufOncOLFy9w6dIlXL9+HVWqVPm6F4SUaomJvH6jRQue6Li4AMeP84JWSnRKr7ySHVmBMo3saBelC5TT0tKwYcMGBAcHo06dOrmmQZYuXaqy4IC8p2G+NLL0pXNEIhFEIpFK4ivOli5dihEjRqBz584wMzPD1KlT8erVq8+uiPL29kZISAhmzJiBpk2bgjGG8uXLo2/fvvJzNm7ciLlz52LSpEl48+YNrK2t0bBhwzw3BtTT00NQUBD69u0rH+H5HF1dXcTExMDPzw/v37+HjY0NevbsWeiROEJOneItCF684NeHDwcWLeKrdEjp9tmRHarZ0SpKJzthYWGoVasWAL7aJjtVTm85ODgA4KM3jo6O8uNRUVHy0R4HBwekpaUhNjZWYXQnKioKjRo1UlksJZWpqSm2b98uv56UlISAgAB5W4/nz5/neb+6devi5MmT+T6uUChEQEBAvgnI4MGDMXjwYPl1PT097N27V37dzs5OocYm+/n6+voICgrK97lzPjYAdO/enWp2SC7x8cDkybz9AMBX22zYAGQuTiQk75EdqtnRSkonO2fPnlVHHLm4u7vDwcEBwcHBqFmzJgA+qhQSEoKFCxcCAGrXrg2hUIjg4GD06dMHAPDu3TuEhYVh0aJFRRJncXbr1i08fPgQ9erVQ1xcnHx5drdu3TQcGSHqdeIE30dFNkM9ahRv/WCimV0jSDFFNTulR6H32Xny5AmePn0q32m3INNLOSUmJuLJkyfy6xEREbh9+zasrKxQtmxZjB8/HvPnz4eHhwc8PDwwf/58GBkZwdfXFwBfSj1kyBBMmjQJ1tbWsLKywuTJk+Hl5SVfnVXa/fbbbwgPD4e+vj5q166NCxcuwEb2P5wQLfPpEzBxIrBxI79erhyvy2neXJNRFUJyMt8Mxt2dtwcnakE1O6WH0slOTEwM+vTpg7Nnz0IgEODx48coV64cfvjhB1hYWGDJkiUFfqwbN26gRYsW8usTJ04EAAwaNAibNm3C1KlTkZKSgpEjR8o3FTx58qRCi4ply5ZBT08Pffr0kW8quGnTplzLokujmjVr4ubNm5oOg5AicfQoMGwY8PYt31hu7Fhg3jy+aVyJcuEC0KMH39pXIODzb5UqAZUrZ/1buTLfNpi6kn6VPEd2dGlTQW2kdLIzYcIECIVCvHz5UmGFTN++fTFhwgSlkp3mzZt/ttZCIBDA398f/v7++Z5jYGCAlStXYuXKlQV+XkKI9vj4ERg/Hti6lV/38OCdsJs00WhYhbNlC6+mlkh4N8u0NCAigl+OH1c818wsdwJUqRJvFV4KFmCogizZ+fSJv+RCIRUoayulk52TJ0/ixIkTuTaJ8/DwwAvZcgdCCCkCBw8CI0bwRpA6OnxX3F9+AYwKvv9m8SCV8i1+Fyzg13v3BjZv5jsgPnzImz09fJj19bNnvAL72jV+yU5Hh8/f5TUaZGNDo0HZWFryl4MxnjTb29M0lrZSOtlJSkqCUR6/SaKjo0vFcm5CiOZFR/NpKtnCvcqVeZ1OgwaajatQkpOBgQN5p0qAt+r+5ReetBgZ8ZbrOTfwFIt5+25ZEpQ9GYqP57c9ecLn9rKztFQcBZL9W748H9YoZXR1ASsrPmMYHc2THSpQ1k5KJzvNmjXDli1bMGfOHAB8qkkqlWLx4sUK9TeEEKIOe/bw1VVRUTwfmDoVmD0b+Mz2UcXX27e8GdfNmzzZ2LCBN+n6EpEIqFaNX7JjDHj/XnEUSPbv8+dAbCzvhHnliuL99PR4wpPXaJCVlcq+3eLIxiYr2QFo6bm2UjrZWbx4MZo3b44bN24gLS0NU6dOxb179/Dx40dcunRJHTESQgiiooDRo4Hdu/n1atX4aE7dupqNq9Bu3QK6dAHevOGfuPv3f32hkUDAW7c7OORegpaSAjx+nHtK7OFDPl0WHs4vhw4p3s/GJu/RIHd3niSVcDY2/NuWJTtUs6OdlP5JrVq1Ku7evYu1a9dCV1cXSUlJ6NmzJ0aNGqWw+R8hhKgCY8CuXTzRiY7mUw/Tp/MSlxI7c37gADBgAJ/CqlIFOHKE19mok6Eh4O3NL9kxxkeYciZADx/yjYqio4GLF/klO6GQV4PnLJCuVKlE9eDIuSKLana0U6HScgcHB9q+X0s1b94cNWrUwPLlyzUdCiGIjARGjuSDHgD/nN64EcjcxL3kYQz47Tdg2jT+dZs2PJPTZHIgEPDW72XK5N5eOimJtwbPORoUHs5Hiu7f55ec7O3zHg1ydeXZajGSM9mhkR3tVKhkJzY2Fn/99RcePHgAgUCAKlWq4LvvvoOVls/tapNz586hRYsWiI2NhUUJ+iuMlA6MAdu3A+PG8VUyenp8JGf6dL4iu0RKSwP+9z++Lh7gX//+e/GeCjI2BmrW5JfspFLg9eu8R4PevuV1Q+/fAyEhivcTiYCKFfMeDcq2f1pRkiU7MTGZIVLNjlZS+n9ZSEgIunXrBjMzM9SpUwcA8Pvvv+OXX37BoUOH4OPjo/IgSckmkUggLIUrPUjhvH3Lm3UeOcKv16zJR3OqV9dsXF/l40egVy/g3DleVb18OZ+XK6nLwHV0gLJl+aVtW8Xb4uPzHg169IivIgsN5ZecnJxyjwZVrgw4O/PnUxNra/5vzpEdSna0i9I/QaNGjUKfPn0QERGBffv2Yd++fXj27Bn69euHUaNGqSNGUkhisRhjx46FnZ0dDAwM0KRJE1y/fh3Pnz+Xr5yztLSEQCBQaK4plUoxdepUWFlZwcHBIdemjnFxcRg2bBjs7OxgZmaGli1b4s6dO/Lb/f39UaNGDfz9998oV64cRCIRNeokX8QYsGkTLzw+coSXhMydC/z7bwlPdB494mviz53joxdHjgBjxpTcROdLzMyAOnV4TdKcObyi/O5dPiX27Blw7BiwdCnPaH18+JQXwLPcM2eANWv4kF67dnzay8SEZ7z9+gH+/sCOHby4OylJJeHmO41FS8+1itIjO0+fPsXevXsV2jHo6upi4sSJ2LJli0qDK44YY0hOTtbIcxsZGSnVf2zq1KnYu3cvNm/eDFdXVyxatAjt2rXD48ePsXfvXvTq1Qvh4eEwMzODYbb+O5s3b8bEiRPx77//4sqVKxg8eDAaN26MNm3agDGGTp06wcrKCseOHYO5uTnWr1+PVq1a4dGjR/KpzCdPnmDXrl25flYIycurV/yz759/+PU6dfhojqenZuP6amfO8BGdT5/4B/eRI1rwTRWSri5fweXuDnTooHjbp09ZtUDZR4MeP+a1Qbdv80tOZcvmvVzeyanAySQVKJcOSic7tWrVwoMHD1CpUiWF4w8ePECNGjVUFVexlZycDBMNtU5OTEyEcQEb/SQlJWHt2rXYtGkTOmT+Yvnzzz8RHByMv//+G3Uz1+va2dnlqtnx9vbG7NmzAfCdsVetWoXTp0+jTZs2OHv2LEJDQxEVFSXfRPK3337DgQMHsGfPHgwbNgwA71C/detW2NraquJbJ1qKMd6oc9IkPvshEgEBAfx6cS5lKZANG3hdTno60LAhX4FlZ6fpqIonCwugfn1+yS49nbfKyGvzxOho4OVLfgkOVryfiUlWLVD2qTEPj1yNValAuXQo0K+Tu3fvyr8eO3Ysxo0bhydPnqBB5nalV69exerVq/Hrr7+qJ0qitKdPn0IikaBx48byY0KhEPXq1cODBw/kyU5evHMsTXV0dERUVBQA4ObNm0hMTIS1bKI7U0pKCp4+fSq/7urqSokO+awXL4ChQ7M+pxo04LW72VrulUwZGXy1laxPoK8vz+hK5K6HGqanxxMUDw+gc2fF22JicidB4eF85+jERL5RY85GyHk0VnVyagigOm0qqOUKlOzUqFEDAoFAoe5i6tSpuc7z9fVF3759VRddMWRkZITExESNPXdByd6rnNNejLEvToXlLCaW7ZIN8HoeR0dHnDt3Ltf9so8QFXQEipQ+Uimwfj3f+TgxkecAc+fyZp4lfsYzMZEnN4cP8+sBAcDMmdpbn6NJ1tZAo0b8kl1aGq8Nyms0KDY2V2NVVwBVEYb7CdUgFlPNjrYqULITERGh7jhKDIFAUCI+yCtUqAB9fX1cvHgRvr6+APiqqBs3bmD8+PHQz1y/m5GRodTj1qpVC5GRkdDT04Obm5uqwyZa7tkz3tT77Fl+vUkTPuhRsaJm41KJV6/4jsh37vD5uE2beFEtKVr6+llTV926ZR1njM9VZa8J2roViIpCVZ2HuC+thpgYyKfnaWRHuxQo2XF1dVV3HETFjI2N8b///Q9TpkyBlZUVypYti0WLFiE5ORlDhgxBcnIyBAIBjhw5go4dO8LQ0LBAtUitW7dGw4YN0b17dyxcuBCVKlXC27dvcezYMXTv3l2+HQEh2UmlwOrVwI8/8k2DDQ2BX3/lq6/VuKq46Fy7xj9YIyP56qIDB0poV1ItJhDwpqrZG6s+egQcPgxX4xgggedCFq5Us6ONClUC+ObNG1y6dAlRUVHy6Q2ZsWPHqiQw8vV+/fVXSKVSDBw4EAkJCahTpw5OnDgBS0tLWFpaIiAgAD/++CO+++47+Pn5YdOmTV98TIFAgGPHjmHGjBn4/vvv8eHDBzg4OKBZs2awly0hJSSbJ0+AIUOA8+f5dR8fPppTvrxm41KZXbuAQYOA1FTAy4tPYdEfiCVDZu1hGcOsZMe+XNbITkGm/UnJIGBKboCyceNGjBgxAvr6+rC2tlb4QRAIBHj27JnKg1S3+Ph4mJubIy4uDmZmZgq3paamIiIiAu7u7jCgAsMSid5DzcjI4BsEz5jBVw8bGwOLFgEjRmjJaA5jwLx5vCYHADp1AoKCNLYTMCmEyZOBJUuww3kS+r/+DTt3Au26xsFioQUAIHVGqnwpOimePvf5nZ3SIzuzZs3CrFmzMH36dOhoxW8sQoiqhYcD338PXL7Mr7dsyVdiu7trNi6VEYt58dG2bfz6hAnA4sVaUGFdymSuO7fT4b0ioqOzCpQBPpVFyY52UDpbSU5ORr9+/SjRIYTkkpHBP/Nr1OCJjqkpX3l16pQWJTofPvCGmdu28eRm3Tq+IzAlOiVP5jSWNbKSHX3drOZrVKSsPZTOWIYMGYLdu3erIxZCSAl2/z5fBTx1Ki9fadsWCAsDhg3TopXX9+7xje8uXQLMzfny5eHDNR0VKazMZMcig2+yEx3NyzFke+3Q8nPtofQ01oIFC9C5c2ccP34cXl5eufZkWbp0qcqCI4QUf+npfDTH359vcWJuzgc6vvtOi5IcADhxAujTh2/1XL48b/1QubKmoyJfIzPZMU3LGtkBeMsIcYaYRna0iNLJzvz583HixAl5u4icBcraihpZllz03qlPaChPamQb1XbqxKetypTRbFwqt3o1MHYsX0PfrBmwb19Wu2xScmW+h0YpismOgZ4B4sXxtPxciyid7CxduhR///23QpdsbSYbuUpOTlZolklKDlnj1pyjkKTwJBK+T86cOfxrCwtgxQpg4EAtG81JT+fFx6tW8euDB/NsTl//s3cjJURmgbIoORYCSBEdzSs7qGWE9lE62RGJRAr9lrSdrq4uLCws5L2hlO08TjRH1qE+KioKFhYW1H1dRW7f5qM5sibUXbvyGl1HR01GpQZxcUDfvnz6CuDZ3dSpWpbNlXJWVgAAgVQKC3xCTAy/Ti0jtI/Syc64ceOwcuVK/P777+qIp1hycHAAAHnCQ0oWCwsL+XtICi88nC8fX76cD3hYWQErVwL9+2vh539EBG88ef8+3+552zagZ09NR0VUTV+fLxlMSIA1YvA2mic7suXmNLKjPZROdq5du4YzZ87gyJEjqFatWq6pgX379qksuOJCIBDA0dERdnZ2kEgkmg6HKEEoFNKIzld4/RrYuRMIDAT++y/reK9evIxFKzfNvnwZ6N6dLzF3cgIOHQJq19Z0VERdrK2BhATYIBpPkj2QnJw1skPJjvZQOtmxsLBAz1L6F46uri59cBKt9/EjsGcPT3DOn+cbBQOAnh5fTj58OJ+60krbt/PdENPSgFq1eKKjddXWRIG1NfD8Oex1Y4AMICYm2zQWFShrDaWTnY0bN6ojDkKIBiUl8ZZOgYF865jsA5hNmwK+vkDv3vJ6Tu0jlQKzZwNz5/LrPXrwjtjGxpqNi6hf5oosV5MYII6vyKICZe1TqEaghJCSTyIBTp7kCc7BgzzhkalenSc4/foBZctqLsYikZLCV1nt2sWvT5sGzJ+vJQ28yBdlZvDOhlnJDhUoax+lkx13d/fPrkYqiY1ACSktpFLg4kXer3L3bj5kL1OuHE9w+vcHqlbVXIxFKjIS6NYNuHYNEAqBP/7giQ8pPTJHdhxFWXvtUIGy9lE62Rk/frzCdYlEglu3buH48eOYMmWKquIihKgIY8CdO3wEJyiIFx3L2Nvz0Zv+/YF69bRwVdXn3LkDdOkCvHrFl5bt3883DCSlS2ayY6+b1TLCoAzV7GibQi09z8vq1atx48aNrw6IEKIaT57w5CYwEHj4MOu4mRlfTeXrCzRvzguPS53Dh3mGl5QEVKrEWz9UqKDpqIgm5NEMVFSWRna0jcp+zXXo0AHTp0+nAmZCNOjdO156EhjIZ2ZkRCI+iOHrC3ToABgYaC5GjWIMWLYMmDyZf92qFZ/Ps7TUdGREUzKTHcuMrGSHana0j8qSnT179sAqczdKQkjR+fSJt2oKDATOnuV1OQCvr23Thg9g9OjBR3RKNYkEGD2a1+UAvB37qlW8VoeUXnk0A3Wi1VhaR+lkp2bNmgoFyowxREZG4sOHD1izZo1KgyOE5C0lhc+8BAYCx47xbWFkGjbkIzjffKOlm/4VRmwsXzt/5gwvTFq6FBg3rpQVKZE8Za7GMkrNSnbK0aaCWkfpZKd79+4K13V0dGBra4vmzZujcuXKqoqLEJJDejpw+jRPcPbvBxISsm6rVi1rqXi5cpqLsVh6/Ji3fnj0CDAx4YVMnTtrOipSXGSO7IgSYwAwREcLaFNBLaR0sjN79mx1xEEIyQNjwJUrPMHZtYt3MJBxdeVTVL6+gJeX5mIs1kJCeE+rjx/5hkGHDwPe3pqOihQnmcmOrkQMYyQhOtqElp5rodK4DoOQYi8sLGup+PPnWcdtbHgj7v79+XQV7Xv3GRs38t4WEglQvz5w4ABADWFJTsbGvCFoWhqsEYPIaBOIdGlkR9sUONnR0dH57GaCAG+YmZ6e/tVBEVIaPX+etVQ8LCzruIkJLzD29eWLh6ie9gukUmD6dGDRIn69b1+e+BgaajYuUjwJBHx05907WCMGL9NcgQwa2dE2BU529u/fn+9tly9fxsqVK8FkHQMJIQUSFcWnp4KCeLNtGX19oGNHnuB06gQYGWkuxhIlKQn49ls+igMAs2bxnlc0BEY+x8YGePcOTvoxuJUGSFJo6bm2KXCy061bt1zHHj58iOnTp+Pw4cMYMGAA5syZo9LgCNFG8fH8szgwEDh1CsjI4McFAqBlSz5F1bMnbf2itNeveTv2W7f4xkJ//QUMGKDpqEhJIGsGahoDxABpyTSyU1iMMYgzxPIi7+KiUDU7b9++xezZs7F582a0a9cOt2/fhqenp6pjI0RrpKYC//zDE5wjR/h1mbp1+QhOnz6Ak5PmYizRbtzgic67d4CtLc8mGzXSdFSkpMhMdlwM+fJzcRLV7BRW3z19cerZKTwa8wg2RjaaDkdOqWQnLi4O8+fPx8qVK1GjRg2cPn0aTZs2VVdshJRoGRl8k7+gIGDvXiAuLuu2SpWymm56eGguRq2wdy8wcCDffKhaNZ5NurlpOipSksiagerz/lgpCTSyU1hXXl9BbGoswqPDYVO2BCY7ixYtwsKFC+Hg4ICgoKA8p7UIKe0YA65f5yM4O3fyptoyZcpkLRWvUYP2s/tqjAG//gr89BO/3r49f9FL/VbRRGnyZqB8ZCc5gTYVLKy0DL7DaXEbFStwsvPjjz/C0NAQFSpUwObNm7F58+Y8z9u3b5/KgiOkpHjwIGup+NOnWcetrPhOxr6+QJMmVCerMmIxX1Yu+z00diywZEkp7WpKvpqsGaiAJztJcQaAHhUoF4Ys2ZH9W1wU+DeDn5/fF5eeE1KavHoF7NjBk5zbt7OOGxkB3bvzUZy2bfnKKqJC0dG8gvvCBUBXF/j9d2DkSE1HRUqyzJYRsmagiZ9EgA2N7BSGfGSnmCWKBU52Nm3apMYwCCkZoqOBPXt4gnPhQtZxPT0+i+Lry+tkjY01F6NWe/CAt3p49oxPV+3aBbRrp+moSEknawYq4clO/EcDwKb4TcWUBCV+GouQ0ioxETh0iCc4J07wHlUyPj58BKd3b/nvS6IuwcF8TjAuDnB354XIVatqOiqiDTL/8xqn8ALluI9UoFwYUiZFupT/giyx01iElCZpaTyxCQzkiU5yctZtNWvyEZy+fQEXF83FWKqsWweMHs2XuDVpAuzbx5eYE6IKmcmOQRIf2YmLpk0FC0OSIZF/XdxeO0p2CMkklQLnz/MEZ88eIDY267YKFbKWileurLkYS52MDGDSJGDFCn594EDgzz/5poGEqEpmsqOXnAAh0hAbzX++xBliMMaoXrWAso/m0DQWIcUIY3zD3cBAXmz85k3WbY6OfPTG1xeoU4eWihe5+HieXR47xq/Pm8d7XtEbQVTNwoL/XDEGK3zEh/dZ/VnSMtLkXdDJ5ykkOzSyQ4jmPXqU1XTz0aOs4+bmvP7G15fX4+jqai7GUu3FC16IHBbGG3hu2cLfGELUQVeX7xMREwNrxOB9WlavltT0VEp2Cih7slPcanaK9a4f/v7+EAgEChcHBwf57Ywx+Pv7w8nJCYaGhmjevDnu3bunwYhJcfbmDbB0KW/PUKkS4O/PEx0DA96qYf9+4P17YMMG3qOKEh0NuXoVqFePJzoODkBICCU6RP2yt4zIyNovgoqUC46msb5CtWrVcOrUKfl13WyfQIsWLcLSpUuxadMmVKxYEXPnzkWbNm0QHh4OU1NTTYRLipnYWN5NIDAQOHeOT1sBPJFp04aP4HTvDtCPSzGxYwcweDDfNLBGDV4dTlXgpChkJjtuJtFAigD6OgZIk6YWuw/t4oymsb6Cnp6ewmiODGMMy5cvx4wZM9CzZ08AwObNm2Fvb4/AwEAMHz68qEMlxURyMnD4ME9w/vkHkGQtEEDjxrwM5JtvADs7zcVIcmAM+OUXPtwG8M2Ktm8HTEw0GhYpRWQjO0Z8RZZQIEIaUmlkRwnFeRqr2Cc7jx8/hpOTE0QiEerXr4/58+ejXLlyiIiIQGRkJNq2bSs/VyQSwcfHB5cvX/5ssiMWiyEWZ2Wd8fHxav0eiPpJJHwblqAgPh2VlJR1m5cXH8Hp14/6QxZLqanA99/zNw8AJk/mPa9oHpEUJVkzUBFPdnSZAYC4YjdCUZzRNFYh1a9fH1u2bEHFihXx/v17zJ07F40aNcK9e/cQmdlh0d7eXuE+9vb2ePHixWcfd8GCBQgICFBb3KToSKW8xubnn4EPH7KOu7llLRX39NRYeORL3r/n84hXr/JtqNetA4YM0XRUpDTKbBkhawaqw2hjQWXRNFYhdejQQf61l5cXGjZsiPLly2Pz5s1o0KABAOTa/6AgeyJMnz4dEydOlF+Pj4+HC9UFlDiPHwNDh/L6VYBPS/Xpw5OcBg1ohXKxFxrKV1y9fAlYWvLiqhYtNB0VKa0yR3ZsMpuB6mQYADrFb4SiOFOYxpIWr2msYr0aKydjY2N4eXnh8ePH8joe2QiPTFRUVK7RnpxEIhHMzMwULqTkSE8HFi0CvL15omNkBCxbxldbrVwJNGxIiU6xd+wY0KgRT3Q8PPjIDiU6RJMykx1LKU92WDqN7CirOI/slKhkRywW48GDB3B0dIS7uzscHBwQHBwsvz0tLQ0hISFo1KiRBqMk6nTnDlC/PjBtGi/1aN2ar1AeP57PgpBijjHepbxLF950rEULnuhUrKjpyEhpl5nsmKXx/lhMQi0jlJU92fkYX7xet2Kd7EyePBkhISGIiIjAv//+i969eyM+Ph6DBg2CQCDA+PHjMX/+fOzfvx9hYWEYPHgwjIyM4Ovrq+nQiYqlpgIzZvCdjP/7j294unEjcPIk7wlJSgCJBBg5Ehg3jhdbDRkCHD/ON3MjRNMykx2jVD6yI02jkR1lZU92PiUWr2SnWP8t/Pr1a/Tv3x/R0dGwtbVFgwYNcPXqVbi6ugIApk6dipSUFIwcORKxsbGoX78+Tp48SXvsaJlLl/jnYng4v96rF7BqFd9vjpQQnz7xgqrgYD7HuHgxMHEizTeS4iNHM9B0cebIDtXsFFj2ZCdVUrxqdop1srNjx47P3i4QCODv7w9/2d4cRKskJAA//QSsXs1nPxwc+NeZ2yqRkuLpU16I/PAhYGzMN0Dq2lXTURGiKHM1ljDhIwSQQpLCkx0a2Sk4xWSneCWJxTrZIaXXP/8Aw4cDr17x699/D/z2G1+0Q0qQCxeAHj2AmBjA2Znv9lijhqajIiS3zJEdgVQKc8ThUwpNYylLTAXKhBRMTAzg5wd07MgTHXd3PvPx11+U6JQ4mzcDrVrxN7VOHeDaNUp0SPGlry/fsdsG0UA6FSgrKzWt+O6gTMkOKRYYA3buBKpUAbZuBXR0eElHaChfcUVKEKmUzz8OHsyLknv35nsEODpqOjJCPi9zdMfVOAagpedKSxJnT3aKV5JI01hE49684Yt0Dh3i16tV4yM59etrNi5SCMnJwMCBwL59/PqMGbznlQ79XUVKAGtr4MULuJnGZI3sFLMP7eIsOVuyI2HF63WjZIdojKzVw5QpQHw8IBTyz8bp0/mIMilh3r7lhcc3b/I3cMMGnvgQUlJkFim7GMUAGTSyo6yUbNNY6ax4TWNRskM04skT3urh3Dl+vX59PppTrZpGwyKFdesW3yjwzRv+gbF/P9CkiaajIkQ5mdNYTqIYqtkphBRx9mSneL1uNLZMilR6Ot9ixcuLJzqyVg+XLlGiU2IdOMATmzdveNHVv/9SokNKpsxkx16PanYKI/vIToageCU7NLJDisydO3xzwJs3+fXWrYE//vjKHZBjY3n2tHEjkJHBsydj46xLzut5HfvSOSIRbX6XF8b4fgDTpvGv27YFdu0CzM01HRkhhZOZ7NgKooF0/ouJanYKLkVCyQ4pxVJTgblzgYUL+ciOhQWwdClfrFPoHCIxEVixgic6cXEqjDYPOjqFS5IKeo5QqN741SEtDfjf/4C//+bXR47k7wc1KCMlWfZmoOlVANDIjjKy75rMBFSzQ0qRS5eAH37gm+cCKmj1kJoKrF0LLFgAfPjAj3l6AgEBvJlkUlLWJTn589e/dI5sSFYq5ds5JyR89euRJ6FQfYmUkRGgq6vaeGNi+BsZEsITweXLgTFjVPschGiCrBmoJAZI43vuxKTEaDKiEkWhRYSOFOnSdOjpFI80o3hEQbSOyls9SCR8quqXX3htCABUqMCv9+2rnqXN6emqSZryO5aRkfW9xcWpb4RKJFJN0mRsDIjFfDvrJ08AU1O+OVKHDuqJm5Cilrkay1gcAyTzkZ17Ufc0GVGJIk5Py3FdDD394pFmFI8oiFY5fpy3enj5kl//qlYPGRlAUBDg7897LAGAiwswaxYwaJB6p4D09Hj9iTpqUBjjSY4qkqb8zmGMP5dYzC8fP6oufjc33vrB01N1j0mIpmVvBhrPf7bfJLzBx5SPsDK00mRkJYJYkjPZSYOxvrGGolFEyQ5RmZgYYMIEvgMywAuP//ijkDsgM8aXL8+aBdzL/MvKzo5vxDNsGGBgoLK4NUIg4HvR6Ourpw8GY3zKT9WjUUlJQIMGvFbHzk71cROiSZnJjjA+BhCbQifODVLz5wh9HwofNx8NB1f8iXO0iPiUKIaVkYaCyYGSHfLVGOOLcMaM4WU0AgEwfjwwZw6f+VD6wU6eBH7+Gbhxgx+zsACmTgXGji3EA5ZSAgFgaMgvmUPzhJAvyEx2dMSpMEIykiO9APPnuPv+LiU7BZCzH1ZMnBjlisnfRJTskK+SV6uHDRv4H/9Ku3iRj9ycP8+vGxvzrGnyZJ7wEEKIOpmY8KlxiQQ2iMHL995ApcMIjQrVdGQlQs5kJza++KzIok0FSaFIpXyKqmpVnugIhbys5r//CpHo3LzJi1ybNuWJjkjE58OePeNr1inRIYQUBYFAPrrjbhYDvPcCANx9f1eTUZUYuZOd4rPXDo3sEKWprNXD/fvAzJlZTSP19Piugz//DDg7qzJkQggpGBsbIDISbqYxCHnvDQAIiwqDlEmhI6Dxgc+RSHPU7CQUn2SH3jlSYOnpfFXVV7d6ePYM8PPjK3n27eN/TX37Ld+MZ906SnQIIZqTObJT1jgG+OgBoUCEJEkSImIjNBxY8Zcr2UmkZIeUMHfu8OmpKVP4Ip/WrYGwMF5SU+A969684bvuVqrEl2wxBvToAdy9y6+XL6/Ob4EQQr4sM9kpI4oGpHpw0OX77VDdzpfl7HQen0Q1O6SEEIv5TFOdOry0xsKCrzo+eVKJnlYfPgCTJvFNANet40NE7doB16/zkR3aq4UQUlxkbwYKwCaDT2VR3c6XpUtzJjvFZ2SHanZIvnK2eujZk7d6cHQs4APExQFLlvC5rsREfqxJE2DePKBZM7XETAghXyUz2bHR4cmOSbIXYEgjOwWRjhzJTjIlO6QYS0zkrR5WreIzTfb2vNVDr14FfICkJGDlSmDRIt6VHABq1+Yrq9q1ow7ihJDiK3NfKispT3aEH72BMkDoe0p2viSdZSY36SJAT4zEFJrGIsXU8eO82HjlSp7ofP898OBBARMdsZjfsXx5YPp0nuhUqQLs3cunrNq3p0SHEFK8ZY7sWAt4svMwhC8/f/zxMVIkKRoLqyTIkI3sZDZRTUwtPiM7lOwQALzVw6BBfLubly9566PgYL6k/IvdDNLT+YkVK/Jdjt+/B8qVA7ZsAUJD+fwXJTmEkJIgM9mxFcTAxAR4+8gBFkIbSJkU9z/c13BwxZtUwJMdvQxTAEASJTukuJC1eqhalecmslYPYWEF6GkllQI7dvA7//ADz5KcnHgR8sOHwMCBSizVIoSQYkDWMuJjNLp2BQABTFNoc8GCkI3sCFlmsiMWgzHg7rNIMFljYg2hZKcUe/uWr/zu2xeIiuI5y+XLvJ74sy2oGOPbJtesCfTvDzx+zOe5lyzhOw4OH67ebuSEEKIumckOYmLQpw//MvYhX5FFRcqfJ9XhyY5Ih09jpYjT0HrcblTf6ogOq0dpMjRKdkojxoA//+TJzcGDPC+ZPbuArR5OnwYaNgS6deP745iZ8Y6fz54BEyfyxpOEEFJSyZKd+Hi0aymBqSmQ+JRGdr6EMQaWOY1lpMtHdsLLjccZ4XgAwImYtZoKDQCtxip1njwBhg0Dzp7l1+vV4+U2X9zq5soV3qRTdkcjI16fM2UKYGWl1pgJIaTIWFry+XzGYJD8Ed262WPbGRrZ+ZIMlgEI+FSVidA06waztxqKSBGN7JQS2Vs9nD2b1erh8uUvJDq3bwNdugCNGvE76usDY8YAT58CCxZQokMI0S66ulmrMmRTWR+qAUyAqKQovE98r7HQUtNTcfzJcTz48EBjMeQnexNQx7QkDUaSNxrZKQXu3uX9NW/c4NdbteIdy8uV+8ydwsOBWbN49TLAfwEMHsyPlS2r7pAJIURzrK2Bjx+BmBi0bQuYGRoh/mMFwPoxQqNCYW9iX2ShSDIkOPXsFHbc24EDDw8gXhwPU31TPBz9EE6mTkUWx5dkT3a8Qo8jpL4Gg8kDjexoMVmrh9q1eaIja/UQHPyZROfFC765TtWqWYlO//68Q/mGDZToEEK0n6xuJzoaIhEvUcT7oqvbyZBm4EzEGQw7PAwOSxzQMbAjttzZgnhxPHQEOkhIS8Dkk5PVHocysic7JsVnL0E5Sna01OXLfLHU3Ll8CqtnT56vfPddPlvevHsHjB4NeHgAGzfyZeVdu/IOoIGBfA8dQggpDbKtyALAp7KiZMmOeut2br69ibLLy6LVllb4878/8THlI+yM7TCq7ihc+O4Crv1wDToCHQSFBeFMxBm1xqIMWbIjzAAM0jUcTB4o2dEyiYm8brhJE77zsb09sGcP38Q4z55WMTHAtGl81+PVqwGJhM9zXbnCl2p5exf590AIIRqVI9lp0wYwSuC/C688U+/IztRTU/E24S2sDK0wtNZQnBp4Cm8mvsGqjqvQyLkJbhyujXZW/wMAjD42WmFE5YuCgrJG7FVMFod+BiDKUMtTfBWq2dEiJ07wLW5evODXv/uOFyXnWUMcH88rlJcu5V8DfEn5vHlAixZFFjMhhBQFqVSKxMRExMXFIT4+Xv6vlZUV6tWrp3hyZn8sWbIjEgHtanhhP4Cn8feRIc2Aro7qN0x9FPMIZyLOQEegg1vDb6GseVbZAGN8bciaNQAM5sJ4+i48iH6A5VeXY2rjqV9+8Hv3AF9f/nXlyrn+kE1PBy5eBGrV4juKKEucnpXs6FOyQ9QhJoZvcbNlC7/u5sYLkNu0yePklBQ+gvPrr/L/yKhenSc5HTtSWwdCSLHCGENycrI8OcmeqChzLCEhId/nOHfuHHx8fLIO5BjZAYDvepTD/stGyNBPRnj0E1S1q6Ty7/WPm38AADpU6JAr0Zk0KTPRAYBUCyTtWwz0GIyAkAD09+wPF3OXzz+4/M4AAgL4cD/4YP7Wrfwj4Nkzvgn++vVA587KxZ6clm1kpxhOY1GyU4IxxqeoRo/mOyALBMC4cbxOJ9cOyGlpfEOdOXN4fQ4AVKoE/PIL0Ls3oEMzmoQQ1UpNTS10ciI7Fh8fj4wM1Q0VCIVCmJubw8zMDKmpqXj79i1WrFjxxWSnXRtd6B6phgyH69h9/i5m91ZtspOanopNtzcBAIbXHi4/zhjvq7xsGb++YQPw6hUQEOAH1NqAZNeLmHhyInZ/szv/B4+Pz/prGAD27YPk+m1svlMD8+cDERH8sEDAd9bv0oV3+1m+vOC7iyQmZyU7etI8TsjQL9gDqQklOyXU27fAyJG8rAbgi6f++iuPHZAzMoBt2wB/f+D5c37M1ZVvmTxwIKBHPwKEEEUSieSrRlFkx9LSVLcsR0dHR56kmJmZyb9W9phIJIIgcwT73r178PT0xMGDB/Hq1Su4uGSOjmRbjSWjrw+UM/bGY1zHkWuhmN37G5V9bwCw78E+xKTEwNnMGR08OsiP+/sDCxfyr1ev5tuIAIBAIID/2tXA8FrYc38PTj49ibbl2+b94Fu38oLOKlWQUc0Lunt24YxPAIam7AcA2NkBU6fy3UUWLuSdf7Zu5St3163LXI32BUmJyQB4sqMDAQDFXli6SWWUeDVUjz7pShjGeFIzeTIQF8dbPfz0E8/8RaJsJ0qlfJhy1izelBMAHByAn3/mTTsVTiaEaIOMjAwkJCQUOjmRfZ2SkqLSuExNTQudnMi+NjIykicpqlKtWjX4+PggJCQE69evx9y5c/kNeYzsAECLal54/BIIjbqLjAzV9jlef3M9AOCHmj9AT4d/NM+fzwffAT6yM3Jk1vmzZwM6Ot6YdWkM0HA5fLeNxpsZoRDp5fjdzhjPkgCEVBuJOZda4iR2o13KAbSyuoWOM2pixAi+0SwALFrEV+9+9x3/6Ojene8+8vvvWaVMeUl9zoeH9DIE0HMqA+C14glSzTaFpmSnBClQqwfGgH/+4UnNrVv8mJUV8OOPwKhRWT/RhBQQYwxJSUn48OEDoqKikJqaqumQSgWxWFzgREX2b2JiokpjMDIy+urRFBMTE+gU42nyUaNGISQkBH/++SdmzZoFfX39XAXKMr2aeOOPQEBsHooLF4DmzVUTw4MPD3D+xXnoCHQwpBYfulmyhHfoAfhoy/jxue83cyYgnuOPeXE7EGP6GJ0X/IbgmTPkt8fGAlfmn0PHBw+QCGN0OuWDpHQrHDDoh56pQTjeKAB6Ew/ketwGDfjHh78/sHgxX8R1+jSwdi1PhPIifv4UAMAyDKBra4+cyQ5jmv0ZoGSnBEhPB1as4D/YKSm81+a8eXyJucJfFiEhfJjn8mV+3dSUVy5PmACYm2skdlI8paSkICoqSp7A5Pev7GtV/6VP1EskEn3VKIrsa71SMM3dvXt3ODk54e3bt9i7dy/69++fNbLz8SP/AzJzRKlWGb7XDqyeYdvuBDRvbprPoypHVpjcuWJnOJs5Y+VKPnoP8JGdqZ9ZbDV3pjkezVmC3dIBOCWeh/H+A1Cnght27uQrdFcYLcLiRsBKT0MkOXnDQGAKy8brgDY7oHfkIO8AXatWrsc1MODrWGSjPPfvA7168cQrr3jSX0UA5kCG1Ah6dg65T2CaXfwiYIyxL5+m3eLj42Fubo64uDiYFWbNnRrlbPXQsiXvWK6wA/L16/xPgOBgft3AgK9RnDr18+OORGuIxeIvJi7ZE5ikJOV71xgYGMDOzg7GuarfiTrICmkLk7CIaJpaKQEBAfD390fjxo1x8eJFvv28gQG/MTaWbz+fyXq+Ez5K3sFy7xV8uNXgq6eyUiQpKLO0DGJTY3HU9yiSbnfkmxiC/1qXzax9DmMM5ee0RAQ7BzzsBhxZC1TbDX3PLUhzuZnrfBN9E5x40hCN/g7m1ciHDn328cViHsuSJfz6smW5R5o2+zbF4EoX4frGFQurtka/uL8Ubhd8rAjpivAvfzNKKujnt/an7SWUWMxHbxYs4CM75uZ8SxyFHZBDQ/lwj6xKWSgEhg7lP5VOxadnClGeRCJBdHR0nqMsef0bL9srSQn6+vqws7ODra0t7OzsFL7O619jY2OV10wQUhwMHToUc+fOxaVLl3Dnzh1Ur16dL2lNSuJFytmSndrOXgiOeIdY/bs4f77BV29Ltuf+HsSmxqKseVk0c2qHKi358XHj+OLZghAIBDg8YhW819aAtPJBoNIhQMCQBt6IvPknc/QbuAjtK7TH4AODcfb5WbQrdwnHXQVofPgw/2u6Tp18H18k4nu2GRvzkaYJE3jBdvYaIvbhFVAJkMAMuvaOQFzOIDW7Hp2SnWLo8mVeQ/wgs7Ftjx68vky+A/KTJ7w6LSiID7Hq6PCVVbNnA+7uGoub5C89PR0xMTH5ThPl/Dc2Nlbp59DT08s3WcnrmKmpKSUvhABwcnJCjx49sHv3bqxevRp//PEHn8pKSuJ1OxUqyM+t7uCN4IiTgF0odu36+j1YZYXJQ2sNxaqVunj9GnBx4X/oKvPfs5pdNUxqNAGLLy8GBAyNyjRE351h+OZqAhz/+gOozYeLjvgeQZegLjgTcQbtBunh+KZ0NPH3B44c+eJz+PvzP8QXLuQloPr6/LMKGRnAx7cAgDSBBXTtHIBHivdlAs3uNEjJTjGSmMhLblat4jmMvT1Pcnr1yjzh1SueVm/cyH+4AOCbb/gGUVWqaCzu0kgqleLjx48FqneJiorCx48foeyMsY6ODmxtbQucwFhYWFDyQkghjRo1Crt378b27duxaNEiWFhbAy9f5ipS9rLPrNuxv4u9e4GVKwu/g8e9qHu49OoSdAW66O76PRp34ccX/xQLQ0NLpR9vQasF8HH1gaedJ1xPXAVO9eN/JffoIT/HSGiEw/0Po9uObjj17BTafwv8s+0oml6/DtStm+9ji1MS8fHeDSzwtYb1W2DLVuD3oYDde6Br3XdIhwQAkK5jDl0Dw9wPQMkOAb7Q6uH9e57mr13LNwcE+G7Hc+fybp/kqzHGEBsbW+CC3ejoaEilee2clT+BQABra+svThfJEhlLS8tivYqFEG3SrFkzVKtWDffu3cPmzZsxLp8VWV52PNkROITiwweG+fMFmDmzcJvPy0Z1ulbqij+XOiE+HljjEIC+//MH7ozgf+0q8TtAV0cXnSp24ldWD+T/DhvGSxyyMRIa4VC/Q+i2oxuCnwWjw7fAsSVj0WzHFYXzUtNTceLBEew+ugiHEm4gUcgw9www/QIwRXbSz/yftMw8SVcggpGxRe7gdGgaq1T7+JEvmNq8mV93c+NbdbdtC14Y99NivhQrmW/YBB8fXszTuLGmQi4RGGOIj48vUMHuhw8f8OHDB6SnK/+f0crKqkCJi62tLaytraGryo05CCEqIxAIMGrUKIwcORJr1qzBmJo1eafsHMlOFdsq0BXoIsMgFjB7g9mznfHyJf9bNEdO8VnJkmRsucN3Ne7iOBzD1gDt8Q/+F+nPT1i3jic6q1Ypn0mFhgIXLvDlukOH5nmKodAQB/sdRPe/2+Jk5EV0KH8Vx/5Zi3qtB+PE0xPYHboDhx8cRALL3GoicwPkGa2AJ04GWBtiCvEnAVIyb44VpQKIh55AHy3Kt879hLqq22CyMCjZ0ZD8Wj3MmQOYIBGYt4JvcBCXWeVVty5Pclq3pv5VORw+fBi7du3KlcAUZvdWc3PzAhfs2tjYQKjMbzdCSLH27bffYtq0aXj06BFOV66MNkCuZMdAzwAVrSviQfQDjJgVij+mOuOvv3jLhT17AMsCzj7turcLceI4uFu449iqNnBIf4UdwoGABECTJsClS7yflUjEl0Ep83s/cxNB9OgBlMl/52JDoSEODglGjxkVcNzoDTpcGQ3d21ORmJa1X5NLHND7hTF6Nx6KWw1cMfbUJGyskornHRpid++9mDrSkneiYIsATINQRx+6Rsb49g6wrXq2J9MTFzx+NaBkRwPevuXFXQcO8OtVq/J+Jw1rpvJsfv584MMHfqOnJ5+u6tqVkpx8PHjwANu2bcvzNhMTkwKvOLKxsaElu4SUYqamphg0aBBWrVqF1eHhPNkJD+crsrJt4+Ft740H0Q/g3iAUhw51QL9+wJkzQKNGwNGjObYGyYdsCqu97VD8uTsD59AP5pIYXpoQHAxs386rf5ct45XABa1YjovjLYIA/kHzBQZ6Btg/5CR6/lIN/3hIgbRElP0E9L4PfPPBFvW+nwmd+UMBAwM0AlDOthL67OmDs8/PovHGhjj021GIxeWx8x3/41JPRx/Q1cUHY8WWEQJdzW5GSslOEcq31cNkCUSBG4E+c4DXmbtOVqjAi5H79qUmnV/QsmVLLF68OFcCY2trC0PDPArlCCEkHyNHjsSqVatw+NEjvARQdvduYPduwNYWqFYNqFoVXm5J2Ang7svrmNofuHgR6NSJt1do0IDvBtKwYf7Pcff9XVx9fRV6Onq4+dd3mIcZaIzLgJkZfy4DA77BWloaX9+9cCEf4QkI+PI3sGULX0VWrRoveygAg4pVsV/fDzv3b0HlaKCuoAwE03/iMeT4A7CDRwdc/O4iOgd1RnhMOBpvaoC5fgew87fMRqA6fL6rTIIekFm0DABMNwOpYikMRBr6PGOExcXFMQAsLi5Obc/x5AljLVsyxlMexurWZezurXTGtm1jrHz5rBtcXBj780/G0tLUFgshhJD8tWjRggFg0ytXZszNLev3c+blUEUw+IN5/w+Mff89YxER7PVrxmrW5KeIRIzt2JH/4488MpLBH6zRst6sEw5nPfaePblPXr486/a5cz8fuFTKWOXK/NzVq5X7pt+/Z2zoUMbWrmUsNfWLp7+Jf8Nqra/F4A8mmiNiGF+WwR+s6oQJjDHGnttYM/hD4fLoWbJyMRVAQT+/aQdlqHcH5YwMYPlyxVYPc+cwjHM7AN3ZM4F79/iJdnZ8mGf48KydOwkhhBS5vXv3onfv3rC1tcWrV68gSk/nG5/dvw/cu4fnT27A3fsMhBlA0jxAT0cP4u++Q9SQsRj8swhnr7wCDN6gcat38KobiYTUKMTExeBj3EfEJsQi4n0E0sXpsH7WGm0izyMdaUhydUWSqyuSk5MhkUhQs2ZNtGjRAi1atIBLUBAwbRoPbvHirF4SYjHw6BH/HLl/nze0OnIEMDHh9RKmqmlnkZ+ktCQM2DcAB8MPyo9ViJyGx2t/RaRTLTTvdQvh2TbxP98xFk3rWqg0hoJ+flOyA/UlO6GhfBTw+nV+vWULhm1+J+G4+ues/g8WFrytw5gx/AeUEEKIRqWnp8PNzQ1v3rxBp06dYGpqiqSkJCQnJyMpKQlJSUkIexMGJmbQTQMy0pG9PEXlypUrhxaWlmhx8yaaAyjTujXfd+3Jk6w917KbMIFvuV8EMqQZmBI8BcuuLgMA2L4ejKg/N+JuvR8QmvoXvu2VdW7k2Fewt3RW6fNTu4hiYMwYnuiYmwPbRlxEpyszIPjuPL/R2Jg3F5k8WWErckIIIZqlp6eHESNGYObMmTh69Ohnz82VaugCEAI6QgGgJ4RUIAIEhtDRNYK9nQncyprB1NgC4kNv4CO+A309A1jMnA4jFxcYGxvD2NgYGRkZuHz5Ms6dO4cbN27g2bNneAZA1m3K49QpNAfQAkBzU1M4ennxlS5VqwJeXgWu1VEFHYEOxlcZj2XbVwGOEkjf8r3fPjjXxID9QL03QNVRQLouIH50H6iv2mSnoLQm2VmzZg0WL16Md+/eoVq1ali+fDmaNm2q0ZjWrgU2j/sP/hk/w2DhP/ygSMQLzn78kU9dEUIIKXYmT54MAwMDiMVieRJibGwMIyMjGBsbIyY9BpciL8Ha3BrO1s5wtioDl0cv4bRkPSyu3IIglQFIQ4Yuw2NdN9xM88LdKG98+OCFGk5RGCv+DgAgCdoBYe9uuZ6/a9euAPjIxcWLF3Hu3DmcPXsW/928iceM4TGAPwEgIQGVYmLQQk8PzZ2c0KxaNVhmZOQ94vOV4uPjce/ePYSFhSlc5L35dAxg4PwG0dHReGlXDakAXD4CRilAvAhINNZXeUwFpRXTWDt37sTAgQOxZs0aNG7cGOvXr8eGDRtw//59lC1b9ov3V1vNztChfE05wPcT//57XrzjrJnMlhBCiJoxxutmli0Dbt4EPtOk91GXSah46DelHj4uLg4XLlzA2bNnce7cOdy6dUvpVjSqpqurC3NzG3z8+P6z5608uhKjO45W6XOXqpqd+vXro1atWli7dq38WJUqVdC9e3csWLAg1/lisRhicdYGR/Hx8XBxcVF9sjN/PvDzz8CAAbyDWvnyqntsQgghxRtjvL9WaChw9y7Sb4Ui9vxdWEaF456ND7zfHIdA/+s2Jo2NjcX58+flIz937txRUfC5CQQClCtXDp6enqhWrRo8PT3h6emJihUrQigUYv/+/fD390dYWFie958aOBUL+y9UaUylJtlJS0uDkZERdu/ejR7Zmp2NGzcOt2/fRkhISK77+Pv7IyCP/QpUnuwkJgLPn/ONAQkhhBAAyMgAE+hAoKP6jWJTUlIgkUi+fGIh6Ovrw+ALq4UZY0hISAAApIkZ9JgEH9JjYSg0hKOVo8pb5pSaAuXo6GhkZGTA3t5e4bi9vT0iIyPzvM/06dMxceJE+XXZyI7KmZhQokMIIUSRri7UtR++oaGhRjdTFQgEuZIOC9jkc3bRKfHJjowgxzbajLFcx2REIhG1BSCEEEJKiRLfh8DGxga6urq5RnGioqJyjfYQQgghpPQp8cmOvr4+ateujeDgYIXjwcHBaNSokYaiIoQQQkhxoRXTWBMnTsTAgQNRp04dNGzYEH/88QdevnyJESNGaDo0QgghhGiYViQ7ffv2RUxMDH755Re8e/cOnp6eOHbsGFxdXTUdGiGEEEI0rMQvPVcFdTYCJYQQQoh6FPTzu8TX7BBCCCGEfA4lO4QQQgjRapTsEEIIIUSrUbJDCCGEEK1GyQ4hhBBCtJpWLD3/WrIFafHx8RqOhBBCCCEFJfvc/tLCckp2AHmHVrU0AyWEEEKIWiUkJMDc3Dzf22mfHQBSqRRv376Fqalpvs1D1UXWcf3Vq1e0x48a0etcNOh1Lhr0OhcNep2LTmFfa8YYEhIS4OTkBB2d/CtzaGQHgI6ODpydnTUag5mZGf1nKgL0OhcNep2LBr3ORYNe56JTmNf6cyM6MlSgTAghhBCtRskOIYQQQrQaJTsaJhKJMHv2bIhEIk2HotXodS4a9DoXDXqdiwa9zkVH3a81FSgTQgghRKvRyA4hhBBCtBolO4QQQgjRapTsEEIIIUSrUbJDCCGEEK1GyQ4hhBBCtBolO2pEC92KBr3ORNvQzzQhqkXJjpp8+PABycnJ8uv0y0s9oqKi5I1cAXqd1UkqlQIAMjIyNByJdouLi1N4jelnWj2ioqLw4cMHpKWlAcj6+Saq9eTJEwQHB2s6DOqNpWoSiQSjRo3CuXPn4OzsDFdXVyxZsgRWVlaaDk2rpKenY/jw4Th9+jTs7Ozg4eGBFStWwMbGRtOhaR2JRIJx48ZBIpHgzz///GyzPVJ4EokE48ePx+3bt2FsbIymTZvip59+gq6urqZD0yoSiQRjxozBuXPnYGFhAQsLCxw8eJA2DlSDu3fvokaNGrCyssLNmzfh6uqqsVjot5YKxcbGomPHjnjy5Ak2btyI/v37486dO+jatSvCw8M1HZ7WSE9Px+DBg3H//n1s3rwZ/fv3x927d9GzZ088ePBA0+FplX///RetW7fGnj17sHnzZly6dAkCgYBGd1QsODgYVatWxb179zBlyhS4uLhg+/bt8Pf3B0CjO6qyZ88eVKlSBQ8fPsTatWsxZMgQPH78GJMmTdJ0aFopLS0N7dq1g1AoxKJFizQaCyU7KnTt2jVERkZi7dq1aNy4MYYOHYrt27fjypUrWL16NaKiojQdolZ49+4drl27hlGjRsHHxwcTJkxAcHAwnj17hrVr1+L9+/eaDlFrXL16FR4eHti8eTO6dOki/1Cg0QbViY+Px65du9CuXTsEBweje/fuWLt2Lfr164fr168jOTkZAoFA02FqhXPnzsHX1xenTp1CixYtMHToUDRt2pRGddTkv//+g6WlJbZv344//vgD165d01gslOyo0Pv37/H69WtUqlRJfiw2NhYWFhYIDg7GhQsXNBid9oiJicHr16/RoEEDAIBYLIaDgwOmT5+OkydP4vz58xqOsOSTjST06tULEydORIcOHTBs2DA8e/YMf/31FwA+wka+XkZGBpo0aYIffvgBQqEQjDHo6+sjNTUVKSkpMDIyopGdrySrx/n5558xdOhQ6OnxCo4XL14gNDQUTk5O+PfffzUZotbI/rMqEong6uqKli1bom7duggICADAE/yiRslOIR07dgyA4hvr4uICa2trLFy4UH5sw4YNGDJkCCQSCU6dOpXrPuTz/vjjD/z5558KCYyHhwccHBywbds2AJDXkIwaNQqmpqb4559/IBaLNRJvSSZ7rUNCQuQjCWXKlEHVqlUBAHXq1EG/fv0QEBCAjIwM6Onp0c9yIWR/nQHA0tISfn5+qFGjBoCsD+a4uDiUK1cOAGhkpxCyv86y3xH2/2/vXoOiKsM4gP93BZFLoRlY3IRItEJDBCEgwgqIRoYcygwFGVGn1CLF5EvZOEwXNR1Jo9QaMbVpDKcmrRAa0ZQcYAhvXOOWU4oUiEwKsSxPH4j1ApYavsvu/n8zftizZ9fn/e/Zsw9n33N2zBi4u7sDADZu3AgvLy/Y2dlh7969iImJwapVq7jvuEnXbs8ajcawDf/000/4888/AQC7du1Cbm4uYmJiEB0djaqqKrWFCt2Uffv2iaurq2g0GiksLBQRke7ubhERaW1tlTVr1ohGo5GQkBBxcHAQX19f0el08v7774urq6sxSzcpn332mTg7O8sjjzwifn5+4uTkJG+99ZaIiFy4cEFWrFghPj4+cu7cORER6ejoEBGR7du3y8iRIw236b8NlPXbb78tIpe37T5FRUUybtw4Wb58uYiI6PV65fWaqn/bpq/MuaenR0REgoKC5OOPP75qGf23G92es7Oz5YcffjBku3PnTrG1tZXGxkaj1G1q/i3nv/76S0REZs2aJd9//72IiGzdulVsbW3F2tpacnJylNfLZucmHD58WJ566ilZsmSJxMTESEBAwIDrHTp0SDZu3Ch5eXmGZatXr5bQ0FBpa2tTVa7J2rVrlzz88MPy0UcfiYjIb7/9Jps2bRJ7e3u5cOGCiIjk5+dLYGCgLFq0SEQufxgUFBSIs7OzHD9+3DjFm5h/y7q9vb3f+pcuXZK1a9eKo6Oj4UOhoKDA8LrQwG4254aGBnFycpKqqirDsrq6OhFhg/lvbiTnaxv4PpWVlWJlZXXVfpsGdqPb89y5cyUxMVECAwPFyclJMjIyZNSoUfLee+8pr5lfY90A+edQ/ZgxYxAVFYVly5YhIyMDFRUVhvkLV16jITw8HEuWLEFkZCSA3hnpR48ehb+/PxwdHdUPwET05azT6RAUFISkpCQAgIuLC/z8/ODq6oqKigoAQFhYGBISErB9+3Z8+eWX0Ol0AIDCwkI8+OCDmDhxonEGYSJuJOuBzmyztbVFXFwcJk+ejJkzZyIgIADx8fFobW1VWr+puNWcc3Nz4e7ujvHjx6OsrAxBQUEIDg5Gd3c3T/0fwM3kfL3J9V999RUef/xxhIWFqSnaBN1Mzh0dHWhvb8e3336LqVOnoqysDK+//jrS09Px2muvobGxUXnxdB2lpaX9jsT0/VWg0+kkLS1NnJycpLOzc8DHV1VVSU1NjSQlJYmXl5ccPXr0ttdsikpLS+X8+fOG221tbf3++jp27Jjcc8890traaljW3t4uK1askDvuuEMee+wxee6558TW1lY++OADEeGh/4HcatZXOnnypEyaNEk0Go0sWrTIcMiaLrvVnPu22ZdfflmeffZZWbp0qWi1WklJSbnufsaS/d/t+ZdffpHa2lqZP3++uLi4SHZ2tohw33GtW825uLhYysvLr1qvs7NT1qxZo/wIJZudAeTk5Iibm5t4e3uLh4eHrFy5Us6ePSsivW+CvjdCfX29uLu7S1pamuG+K61bt068vb0lPDxcampq1A7CBFyb8xtvvCFNTU2G+698M6xfv15CQ0NFRPp9uH7xxRfy5ptvyosvviiVlZVqijcxg5X14cOHZezYsRIcHCy1tbVqijchg5GzXq+XsWPHikajkYiIiH4fFnTrOXd1dRmW19TUSFpamri5ucm0adOkurpa3QBMxK3mPBQbczY71ygpKZEJEybIhg0b5Pjx45KVlSVOTk7y0ksvSUtLi4hcPrrT09MjWVlZYmVlJfX19SLSu9Pq+87yzJkzUlpaapyBDHE3krNerxedTiciIjNmzJDFixcbs2STNZhZnzlzhkcor2Owcm5ra5N33nlH9u/fr7R+UzFYOV+6dEkOHjxoONGErmZu+2g2O//oOyrz4Ycfipub21UTLjdt2iTBwcGSkZHR73EtLS0SEhIicXFxUlpaKlFRUbJjxw5OIryOm81Zr9dLT0+PeHt7y759+0REpLq6WmbNmiWnT59WW7yJYdZqMGc1mLMa5pozZ7r9o+86Fg0NDfDx8TFcdAoAkpOTMWXKFHz33XcoLy8HcPnHEO+66y4sWLAAX3/9NQIDAzF8+HDEx8dzEuF13GzOWq0WJSUlsLOzg7+/P1599VVMmjQJLS0tcHZ2NsoYTAWzVmMwc3ZycjLKGEwBt2c1zDVni/1Ezs/PxyuvvILMzMyrLmEdGhqKH3/8EU1NTQB6mxp7e3vExcVBo9EgLy8PQO+M/q6uLmRlZSElJQXh4eE4ceIE9u7dC1tbW6OMaSj6vzkDvRdwPHXqFMaPH4/8/HwUFhYiLy+Pl3i/BrNW43bmPGLECOXjGaq4PathKTlbXLNz9uxZxMbGYs6cOWhtbcUnn3yCqKgow4scFRUFT09Pw1WQ+7rcyMhIaLVa1NbWGp7r/PnzqKmpwbZt23Dw4EE89NBD6gc0RA1mztbW1rj77ruRnZ2N8vJyTJkyRf2AhjBmrQZzVoM5q2FxORv7ezSVLl68KHPnzpXnn3/eMKFYRCQwMFCSk5NFpHfy8aeffiparbbfxLXZs2fLtGnTlNZsigYj54iICMPt5uZmNYWbIGatBnNWgzmrYYk5W9SRHTs7O9jY2CA5ORleXl6GHzKcPn36VRecmjlzJuLi4jB//nwcOnQIIoKmpib8/PPPmD17tjGHYBIGI+c5c+YYno/zGK6PWavBnNVgzmpYYs4aEcv6JT+dTgdra2sAvVeD1Gg0SExMhK2tLbZs2WJY1tnZiZiYGFRUVMDPzw+nTp2Ch4cHdu/ebfghObo+5qwOs1aDOavBnNWwtJwtrtkZSHh4OObNm4fk5GSICHp6ejBs2DCcO3cOJ06cQElJCTw9PZGQkGDsUk0ac1aHWavBnNVgzmqYc84W3+zU19cjJCQE33zzjWFSVVdXF4YPH27kyswLc1aHWavBnNVgzmqYe84WNWfnSn093pEjR+Dg4GB4cVetWoXU1FQ0NzcbszyzwZzVYdZqMGc1mLMalpKz1X+vYp76TqMrLi5GfHw88vPzsXDhQly6dAk7duwYUhdDMmXMWR1mrQZzVoM5q2ExOd/ek72Gto6ODrn//vtFo9GIjY2NvPvuu8YuySwxZ3WYtRrMWQ3mrIYl5Gzxc3YiIyMxbtw4rF+/nlcvvY2YszrMWg3mrAZzVsPcc7b4Zkev12PYsGHGLsPsMWd1mLUazFkN5qyGueds8c0OERERmTeLPRuLiIiILAObHSIiIjJrbHaIiIjIrLHZISIiIrPGZoeIiIjMGpsdIiIiMmtsdoiIiMissdkhIpMgInjyyScRHR3d776srCw4Ojri9OnTRqiMiIY6NjtEZBI0Gg22bduGoqIibN682bC8oaEB6enpyMzMhIeHx6D+nzqdblCfj4iMg80OEZkMd3d3ZGZmYvny5WhoaICIICUlBU888QSmTp2Kp59+Gg4ODhgzZgwSExPxxx9/GB6bm5uLsLAwjBw5EqNHj8b06dNRV1dnuL+xsREajQa7d+9GREQERowYgZ07dxpjmEQ0yPhzEURkcp555hm0tbUhPj4eGRkZKCkpQUBAABYsWICkpCR0dHQgPT0d3d3dOHDgAABgz5490Gg0mDhxIi5evIiVK1eisbERx44dg1arRWNjI7y8vODp6Yl169Zh8uTJsLGxgYuLi5FHS0T/F5sdIjI5zc3N8PX1RUtLC3JyclBWVoaioiLs37/fsM6vv/4Kd3d3VFdXw8fHp99z/P7773B2dsbJkyfh6+traHY2bNiA1NRUlcMhotuMX2MRkclxdnbGwoUL8cADD2DGjBkoLS1FQUEBHBwcDP8mTJgAAIavqurq6pCQkID77rsPd955J7y8vACg36TmgIAAtYMhotvOytgFEBHdCisrK1hZ9e7Cenp6EBsbi9WrV/db79577wUAxMbGwt3dHVu3boWLiwt6enrg6+uLrq6uq9a3t7e//cUTkVJsdojI5Pn7+2PPnj3w9PQ0NEBXamlpQWVlJTZv3oxHH30UAHDkyBHVZRKRkfBrLCIyeYsXL0ZrayteeOEFFBcXo76+Hnl5eZg3bx70ej1GjRqF0aNHY8uWLaitrcWBAwewbNkyY5dNRIqw2SEik+fi4oLCwkLo9XpER0fD19cXqampcHR0hFarhVarxeeff47S0lL4+vpi6dKlWLt2rbHLJiJFeDYWERERmTUe2SEiIiKzxmaHiIiIzBqbHSIiIjJrbHaIiIjIrLHZISIiIrPGZoeIiIjMGpsdIiIiMmtsdoiIiMissdkhIiIis8Zmh4iIiMwamx0iIiIya38DwSsclo8Gt3sAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.dates as mdates\n",
    "\n",
    "\n",
    "sns.lineplot(x=\"date\", \n",
    "             y=\"text\",\n",
    "             hue=\"language\", \n",
    "             palette=palette,\n",
    "             data=date_df2)\n",
    "\n",
    "plt.title(\"Dates of submitted comments by language\")\n",
    "plt.xlabel(\"Year\")\n",
    "plt.ylabel(\"Number of comments\")\n",
    "\n",
    "plt.gcf().autofmt_xdate()\n",
    "plt.gca().fmt_xdata = mdates.DateFormatter('%Y-%m')\n",
    "\n",
    "save_plot(\"time_plot.png\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9ee4484b-30a7-4055-a7a9-41343ac44ebb",
   "metadata": {},
   "source": [
    "## Toxicity Classification"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "40266753-306c-4ccb-aeaa-34bddf01c33c",
   "metadata": {},
   "source": [
    "We will now attempt to build a Toxicity Classifier using an LLM. The process, prompts and decisions of this classifier can be found in the `prompts.pdf` report. Note that due to the project's restrictions we are only allowed to use data annotated by the LLM model."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0f3cb51e-ecf0-452d-8dbb-e0c56f014bc8",
   "metadata": {},
   "source": [
    "### Importing the data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7d9bc918-44c5-461b-8bb6-42bae62c14ef",
   "metadata": {},
   "source": [
    "We will import the LLM's responses and build a dataframe out of them. The annotated dataset was derived from a sample of `gold.csv`, in order to avoid feeding the model operational data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "f9ea4800-2809-480f-b8e7-35e68e153601",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(365,\n",
       " '\"Εαν σας άρεσε κάντε Like ρεεε μη ντρέπεστε! +1 όσοι είδαν το μπλουζάκι PDT p.s το AR το βάλαμε έτσι για συντομογραφία, εννοείται πως άλλο AR και άλλο τρίσφαιρο!\" - 1')"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "with open(os.path.join(\"data\", \"chatgpt_annot.txt\"), \"r\", encoding=\"utf8\") as file:\n",
    "    raw_annot = file.read()\n",
    "\n",
    "records = raw_annot.split(\"\\n\")\n",
    "len(records), records[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "b955aaa6-9b30-46f6-a279-ae4a029d2329",
   "metadata": {},
   "outputs": [],
   "source": [
    "annotations = [record.split(\"-\") for record in records]\n",
    "\n",
    "comments = []\n",
    "values = []\n",
    "for annotation in filter(lambda x: len(x) != 1, annotations):\n",
    "    comment = annotation[0]\n",
    "    value = int(annotation[1])\n",
    "    comments.append(comment)\n",
    "    values.append(value)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "7f12bd53-b2ae-4791-8123-811ff3d22791",
   "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>comments</th>\n",
       "      <th>toxicity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>\"Εαν σας άρεσε κάντε Like ρεεε μη ντρέπεστε! +...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>\"Φήμες λένε ότι ο Gianuba βάζει καρδούλες ️\"</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>\"Congrats on your channel , que venha o 1 milh...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>\"ΤΟΣΑ ΤΕΛΕΙΑ ΒΙΝΤΕΟ ΘΑ ΚΑΝΕΙ Ο GIANOUBA22\"</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>\"Η φάση που ο gianuba 22 φοράει μπλούζα pdt nu...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178</th>\n",
       "      <td>Ψάχνουμε άτομα για raid Παρασκευή 9:00, Σάββατ...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>179</th>\n",
       "      <td>Μήπως παίζατε καθόλου στον παλιό Neltharion πρ...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>180</th>\n",
       "      <td>Σε Horde ή Alliance θα είναι το guild;</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>181</th>\n",
       "      <td>Υπάρχει κανένα active ελληνικό guild από πλευρ...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>182</th>\n",
       "      <td>Θα παίξει κανείς στον καινούργιο realm; Αν ψηθ...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>183 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                              comments  toxicity\n",
       "0    \"Εαν σας άρεσε κάντε Like ρεεε μη ντρέπεστε! +...         1\n",
       "1        \"Φήμες λένε ότι ο Gianuba βάζει καρδούλες ️\"          1\n",
       "2    \"Congrats on your channel , que venha o 1 milh...         1\n",
       "3          \"ΤΟΣΑ ΤΕΛΕΙΑ ΒΙΝΤΕΟ ΘΑ ΚΑΝΕΙ Ο GIANOUBA22\"          1\n",
       "4    \"Η φάση που ο gianuba 22 φοράει μπλούζα pdt nu...         1\n",
       "..                                                 ...       ...\n",
       "178  Ψάχνουμε άτομα για raid Παρασκευή 9:00, Σάββατ...         3\n",
       "179  Μήπως παίζατε καθόλου στον παλιό Neltharion πρ...         4\n",
       "180            Σε Horde ή Alliance θα είναι το guild;          1\n",
       "181  Υπάρχει κανένα active ελληνικό guild από πλευρ...         1\n",
       "182  Θα παίξει κανείς στον καινούργιο realm; Αν ψηθ...         1\n",
       "\n",
       "[183 rows x 2 columns]"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "toxicity_df = pd.DataFrame({\"comments\": comments, \"toxicity\": values})\n",
    "toxicity_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2d1703f6-b0f5-4ad4-9945-1ae5685bf1fe",
   "metadata": {},
   "source": [
    "Our dataset is comprised of 182 comments. These are of course way too few for any NLP task, but due to resource constraints and those placed by ChatGPT's frontend is the best sample we can acquire.\n",
    "\n",
    "Our sample attempted to be as representative as possible, including data from all language categories as uniformally as possible, as well as purposefully including many toxic comments.\n",
    "\n",
    "The overall distribution of labels produced by ChatGPT can be seen below:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "3d461df4-506b-4ce7-975b-8e7db19712bc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeoAAAH+CAYAAABTKk23AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy81sbWrAAAACXBIWXMAAA9hAAAPYQGoP6dpAAA+sUlEQVR4nO3deVxU9eL/8feIMoAi7iy5gOaSu6kppOKSlkvLtdLUUnO5ejWXrFDzJugtSOuSlmWZC1oPl2supeVCpVipqaW5oZaiWEmE4Qqh4vn94Y/5OgLK6AxzlNfz8ZjHo/nMZ8685wP55pw5M2MxDMMQAAAwpWLuDgAAAPJHUQMAYGIUNQAAJkZRAwBgYhQ1AAAmRlEDAGBiFDUAACZGUQMAYGIUNQAAJkZRo1DExcXJYrFox44d+c45evSoLBaL3nzzzetuKzg4WBaLRW3bts3z9gULFshischisWjjxo3X3db+/fsVFRWlo0eP3uAZ3Jqc5xYXF+fQ/TZu3JjreXzxxReKiopyar6C6t+/v4KDg93y2FL+67hkyRLVq1dP3t7eslgs2rVrl6KiomSxWNwT1Anee+89h39fblZGRoaioqJu+P8L3IOixm3J19dXmzZt0uHDh3PdNnfuXJUuXbpA29m/f78mTZrk8qIODAzUli1b1LVrV4fud++992rLli269957bWNffPGFJk2a5OyIt4W81vHPP//UM888oxo1amjt2rXasmWLatWqpUGDBmnLli1uTHtrCruoJ02aRFGbFEWN21KrVq101113ae7cuXbjhw8f1qZNm9SzZ083Jcub1WpVy5YtVbFiRYfuV7p0abVs2bLAf3jc6fJax0OHDunixYt6+umnFR4erpYtW8rHx0eVK1dWy5YtnfbYGRkZTtsW4AiKGrelYsWKqW/fvpo/f74uX75sG587d66qVKmiBx544IbbiIuL05NPPilJateune1w+dV7MXPnzlWjRo3k5eWlcuXK6R//+IcSExNtt7/++usqVqyYVq1aZbft/v37y8fHR3v27JGU/yHbAwcOqFevXvL395fValXVqlXVt29fZWVlScp96Lt///569913JcmW12Kx6OjRo+rQoYPq1Kmja79nxzAM3X333QXam1+4cKFCQ0NVqlQplSpVSo0bN9acOXOue593331Xbdq0UaVKlVSyZEk1aNBAU6dO1cWLF+3m7dy5U926dVOlSpVktVoVFBSkrl276tdff7XNWbp0qVq0aCE/Pz/5+PioevXqGjBggO32a9exf//+atWqlSSpZ8+edi+J5Hfoe8mSJQoNDVXJkiVVqlQpPfjgg9q5c6fdnP79+6tUqVLas2ePOnXqJF9fX3Xo0OG66/Dtt9+qQ4cO8vX1lY+Pj8LCwvT555/bzckvU85LQzlHdoKDg7Vv3z4lJCTYfsY5Lznk/E58/PHHGjNmjAICAuTt7a3w8PBcz6Nt27Z5vkR09UsYR48etf3hM2nSJNvj9e/f/7rPF4WHosZta8CAAfr999+1bt06SVJ2drbmz5+v/v37q1ixG/9qd+3aVdHR0ZKulM2WLVvsDqvGxMRo4MCBqlevnpYvX67p06dr9+7dCg0N1c8//yxJGjt2rDp37qx+/frp2LFjkqR58+Zp/vz5euedd9SgQYN8H/+nn35S8+bNtXXrVk2ePFlr1qxRTEyMsrKydOHChTzv88orr+iJJ56QJFveLVu2KDAwUKNGjdLBgwf11Vdf2d1nzZo1Onz4sIYPH37d9Zg4caL69OmjoKAgxcXFacWKFXbPKz+HDx9W79699dFHH2n16tUaOHCg3njjDQ0ZMsQ25/z58+rYsaP++OMPvfvuu4qPj9e0adNUtWpVnT171vZ8evbsqerVq2vx4sX6/PPPNXHiRF26dCnfx37llVdsf7hER0dry5Yteu+99/KdHx0drV69eqlu3br63//+p48++khnz55V69attX//fru5Fy5c0COPPKL27dvr008/ve7LDQkJCWrfvr1Onz6tOXPmaNGiRfL19dXDDz+sJUuWXHf98rJixQpVr15dTZo0sf2MV6xYYTfn5Zdf1pEjRzR79mzNnj1bv//+u9q2basjR4449FiBgYFau3atJGngwIG2x3vllVcczg0XMYBCMG/ePEOSsX379nznJCUlGZKMN95447rbqlatmtG1a1fDMAwjPDzceOKJJwzDMIzPP//csFgsRlJSkrF06VJDkrFhw4brbiu/eenp6Ya3t7fRpUsXu/Hk5GTDarUavXv3to2lpaUZlStXNu677z7jxx9/NHx8fIynn346z+c2b94821j79u2NMmXKGKmpqfnm27BhQ658w4cPN/L6Xzc7O9uoXr268eijj9qNd+7c2ahRo4Zx+fLlfB/nyJEjhoeHh9GnT5985xiGYfTr18+oVq1avrdnZ2cbFy9eNBYsWGB4eHgYf/31l2EYhrFjxw5DkrFy5cp87/vmm28akoxTp07lOyevdcxZo6VLl9rNjYyMtFun5ORko3jx4saIESPs5p09e9YICAgwevToYfc8JRlz587NN8vVWrZsaVSqVMk4e/asbezSpUtG/fr1jcqVK9vW/tpMOXL+/0hKSrKN1atXzwgPD881N+f53nvvvXY/06NHjxolSpQwBg0aZBsLDw/PcxvX/hz//PNPQ5IRGRlZoOeLwsUeNW5rAwYM0GeffaaTJ09qzpw5ateunVPOSt6yZYsyMzNzHf6rUqWK2rdvb7fXWr58eS1ZskQ//vijwsLCVLVqVb3//vvX3X5GRoYSEhLUo0cPh1+3zk+xYsX03HPPafXq1UpOTpZ0ZW937dq1GjZs2HXPgI6Pj1d2dvYN97rzsnPnTj3yyCMqX768PDw8VKJECfXt21fZ2dk6dOiQJOnuu+9W2bJlNXbsWL3//vu59l4lqXnz5pKkHj166H//+59+++03h7Ncz7p163Tp0iX17dtXly5dsl28vLwUHh6e54lUjz/++A23e/78eX3//fd64oknVKpUKdu4h4eHnnnmGf366686ePCgM5+KJKl37952P9Nq1aopLCxMGzZscPpjwb0oatzWnnjiCXl5eemtt97SqlWrNHDgQKds9+TJk5KuHBa8VlBQkO32HC1atFC9evX0999/61//+pdKlix53e2np6crOztblStXdkreHAMGDJC3t7ftD4V3331X3t7edq/z5uXPP/+UJIfzJCcnq3Xr1vrtt980ffp0ffPNN9q+fbvtcHRmZqYkyc/PTwkJCWrcuLFefvll1atXT0FBQYqMjLS9lt2mTRutXLnSVqaVK1dW/fr1tWjRIocy5eePP/6QdOUPghIlSthdlixZorS0NLv5Pj4+BTqJLz09XYZh5Pu7IinX74szBAQE5DnmiseCexV3dwDgVvj4+Oipp55STEyMSpcure7duztlu+XLl5cknThxItdtv//+uypUqGA3FhkZqT179qhp06aaOHGiunXrpurVq+e7/XLlysnDw8PuRCpn8PPzU79+/TR79my9+OKLmjdvnnr37q0yZcpc9345e/W//vqrqlSpUuDHW7lypc6fP6/ly5erWrVqtvFdu3blmtugQQMtXrxYhmFo9+7diouL0+TJk+Xt7a1x48ZJkh599FE9+uijysrK0tatWxUTE6PevXsrODhYoaGhBc6Vl5yf2SeffGKXNT8FfQ922bJlVaxYsXx/V65+bC8vL0lSVlaWrFarbd61fyQUREpKSp5jOb+7OY93+vTpXPNu5vHgPuxR47b3r3/9Sw8//LAmTpxo+4ewoHL+sczZ88sRGhoqb29vffzxx3bjv/76q77++mu7M4Dj4+MVExOjf//734qPj5efn5969uyZ7wlhkmxn6S5dutThfzTzy5xj5MiRSktL0xNPPKFTp07pueeeu+E2O3XqJA8PD82cOdOhLDlldnXpGIahDz/88Lr3adSokd566y2VKVNGP/74Y645VqtV4eHhmjJliiTlOpv5Zjz44IMqXry4Dh8+rGbNmuV5uRklS5ZUixYttHz5crufyeXLl/Xxxx+rcuXKqlWrliTZXpbZvXu33TaufdeAdGUN8vsZS9KiRYvszvA/duyYNm/ebHeWd3BwsA4dOmR7F4F0Ze9+8+bNuR5Lyv93Cu7FHjUK1ddff53nh4t06dLF9t979uzRJ598kmtO8+bN89wTaty4sVauXHlTeerXry9JmjVrlnx9feXl5aWQkBCVL19er7zyil5++WX17dtXvXr10smTJzVp0iR5eXkpMjJS0pU97pz370ZGRqpYsWJasmSJ2rRpo4iICE2bNi3fx46NjVWrVq3UokULjRs3Tnfffbf++OMPffbZZ/rggw/k6+ub5/1yziSfMmWKOnfuLA8PDzVs2FCenp6SpFq1aumhhx7SmjVr1KpVKzVq1OiG6xAcHKyXX35Z//nPf5SZmalevXrJz89P+/fvV1paWr5nPHfs2FGenp7q1auXIiIi9Pfff2vmzJlKT0+3m7d69Wq99957euyxx1S9enUZhqHly5fr1KlT6tixo6QrZ53/+uuv6tChgypXrqxTp05p+vTpKlGihMLDw2/4HAryHCdPnqwJEyboyJEjeuihh1S2bFn98ccf2rZtm0qWLHnTHyQTExOjjh07ql27dnrxxRfl6emp9957T3v37tWiRYtsf9B06dJF5cqV08CBAzV58mQVL15ccXFxOn78eK5t5hyBWLJkiapXry4vLy+7dxGkpqbqH//4hwYPHqzTp08rMjJSXl5eGj9+vG3OM888ow8++EBPP/20Bg8erJMnT2rq1Km5Dun7+vqqWrVq+vTTT9WhQweVK1dOFSpUcOun0OEqbj2VDUVGzlmt+V2SkpJsZ/Tmd8k50/fqs77zU9Czvg3DMKZNm2aEhIQYHh4euc4onj17ttGwYUPD09PT8PPzMx599FFj3759hmFcOas3PDzc8Pf3N06cOGG3zTfeeMOQZKxYscIwjLzPVjYMw9i/f7/x5JNPGuXLlzc8PT2NqlWrGv379zf+/vtvwzDyPus7KyvLGDRokFGxYkXDYrHkOlvYMAwjLi7OkGQsXrz4hs//agsWLDCaN29ueHl5GaVKlTKaNGlilzmvs75XrVplNGrUyPDy8jLuuusu46WXXjLWrFljl/vAgQNGr169jBo1ahje3t6Gn5+fcd999xlxcXG27axevdro3Lmzcddddxmenp5GpUqVjC5duhjffPONbc6tnPWdY+XKlUa7du2M0qVLG1ar1ahWrZrxxBNPGF9++aXd8yxZsqRDa/fNN98Y7du3N0qWLGl4e3sbLVu2NFatWpVr3rZt24ywsDCjZMmSxl133WVERkYas2fPzvVzPHr0qNGpUyfD19fXkGRb95zn+9FHHxkjR440KlasaFitVqN169bGjh07cj3e/PnzjXvuucfw8vIy6tatayxZsiTPn+OXX35pNGnSxLBarYYko1+/fg49f7iOxTCu+XQEALe9xx9/XFu3btXRo0dVokQJd8eBE23cuFHt2rXT0qVLbe+px52NQ9/AHSIrK0s//vijtm3bphUrVig2NpaSBu4AFDVwhzhx4oTCwsJUunRpDRkyRCNGjHB3JABOwKFvAABMjLdnAQBgYhQ1AAAmRlEDAGBiFLWufJLSmTNncn2PLwAA7kZRSzp79qz8/Pxs34sLAIBZUNQAAJgYRQ0AgIlR1AAAmJhbi3rTpk16+OGHFRQUJIvFkusbkAzDUFRUlIKCguTt7a22bdtq3759dnOysrI0YsQIVahQQSVLltQjjzzi9O/4BQDAXdxa1OfPn1ejRo00Y8aMPG+fOnWqYmNjNWPGDG3fvl0BAQHq2LGj3Ulfo0eP1ooVK7R48WJ9++23OnfunLp166bs7OzCehoAALiMaT5C1GKxaMWKFXrsscckXdmbDgoK0ujRozV27FhJV/ae/f39NWXKFA0ZMkSnT59WxYoV9dFHH6lnz56SpN9//11VqlTRF198oQcffLBAj33mzBn5+fnp9OnTub6nFQAAdzLta9RJSUlKSUlRp06dbGNWq1Xh4eHavHmzJOmHH37QxYsX7eYEBQWpfv36tjl5ycrK0pkzZ+wuAACYkWmLOiUlRZLk7+9vN+7v72+7LSUlRZ6enipbtmy+c/ISExMjPz8/26VKlSpOTg8AgHOYtqhzWCwWu+uGYeQau9aN5owfP16nT5+2XY4fP+6UrAAAOJtpizogIECScu0Zp6am2vayAwICdOHCBaWnp+c7Jy9Wq1WlS5e2uwAAYEamLeqQkBAFBAQoPj7eNnbhwgUlJCQoLCxMktS0aVOVKFHCbs6JEye0d+9e2xwAAG5nxd354OfOndMvv/xiu56UlKRdu3apXLlyqlq1qkaPHq3o6GjVrFlTNWvWVHR0tHx8fNS7d29Jkp+fnwYOHKgXXnhB5cuXV7ly5fTiiy+qQYMGeuCBB9z1tAAAcBq3FvWOHTvUrl072/UxY8ZIkvr166e4uDhFREQoMzNTw4YNU3p6ulq0aKH169fL19fXdp+33npLxYsXV48ePZSZmakOHTooLi5OHh4ehf58AABwNtO8j9qdeB81AMCsTPsaNQAAoKgBADA1ihoAABOjqAEAMDG3nvV9p0pOTlZaWpq7Y9xQhQoVVLVqVXfHAABcB0XtZMnJyapT5x5lZma4O8oNeXv76MCBRMoaAEyMonaytLQ0ZWZmqMWASJUODHZ3nHydOXFU38+dpLS0NIoaAEyMonaR0oHBKle1trtjAABuc5xMBgCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJiYqYv60qVL+ve//62QkBB5e3urevXqmjx5si5fvmybYxiGoqKiFBQUJG9vb7Vt21b79u1zY2oAAJzH1EU9ZcoUvf/++5oxY4YSExM1depUvfHGG3rnnXdsc6ZOnarY2FjNmDFD27dvV0BAgDp27KizZ8+6MTkAAM5h6qLesmWLHn30UXXt2lXBwcF64okn1KlTJ+3YsUPSlb3padOmacKECerevbvq16+v+fPnKyMjQwsXLnRzegAAbp2pi7pVq1b66quvdOjQIUnSTz/9pG+//VZdunSRJCUlJSklJUWdOnWy3cdqtSo8PFybN292S2YAAJypuLsDXM/YsWN1+vRp1alTRx4eHsrOztZrr72mXr16SZJSUlIkSf7+/nb38/f317Fjx/LdblZWlrKysmzXz5w544L0AADcOlPvUS9ZskQff/yxFi5cqB9//FHz58/Xm2++qfnz59vNs1gsdtcNw8g1drWYmBj5+fnZLlWqVHFJfgAAbpWpi/qll17SuHHj9NRTT6lBgwZ65pln9PzzzysmJkaSFBAQIOn/9qxzpKam5trLvtr48eN1+vRp2+X48eOuexIAANwCUxd1RkaGihWzj+jh4WF7e1ZISIgCAgIUHx9vu/3ChQtKSEhQWFhYvtu1Wq0qXbq03QUAADMy9WvUDz/8sF577TVVrVpV9erV086dOxUbG6sBAwZIunLIe/To0YqOjlbNmjVVs2ZNRUdHy8fHR71793ZzegAAbp2pi/qdd97RK6+8omHDhik1NVVBQUEaMmSIJk6caJsTERGhzMxMDRs2TOnp6WrRooXWr18vX19fNyYHAMA5TF3Uvr6+mjZtmqZNm5bvHIvFoqioKEVFRRVaLgAACoupX6MGAKCoo6gBADAxihoAABOjqAEAMDGKGgAAE6OoAQAwMYoaAAATo6gBADAxihoAABOjqAEAMDGKGgAAE6OoAQAwMYoaAAATo6gBADAxihoAABOjqAEAMDGKGgAAE6OoAQAwMYoaAAATo6gBADAxihoAABOjqAEAMDGKGgAAE6OoAQAwMYoaAAATo6gBADAxihoAABOjqAEAMDGKGgAAE6OoAQAwMYoaAAATo6gBADAxihoAABOjqAEAMDGKGgAAE6OoAQAwMYoaAAATo6gBADAxihoAABOjqAEAMDGKGgAAE7vloj5z5oxWrlypxMREZ+QBAABXcbioe/TooRkzZkiSMjMz1axZM/Xo0UMNGzbUsmXLnB4QAICizOGi3rRpk1q3bi1JWrFihQzD0KlTp/T222/r1VdfdXpAAACKMoeL+vTp0ypXrpwkae3atXr88cfl4+Ojrl276ueff3Z6QAAAijKHi7pKlSrasmWLzp8/r7Vr16pTp06SpPT0dHl5eTk9IAAARVlxR+8wevRo9enTR6VKlVK1atXUtm1bSVcOiTdo0MDZ+QAAKNIcLuphw4apRYsWSk5OVseOHVWs2JWd8urVq+u1115zekAAAIoyhw99T548Wffcc4/+8Y9/qFSpUrbx9u3b68svv3RqOAAAijqHi3rSpEk6d+5crvGMjAxNmjTJKaEAAMAVDhe1YRiyWCy5xn/66Sfb2eAAAMA5CvwaddmyZWWxWGSxWFSrVi27ss7Ozta5c+c0dOhQl4QEAKCoKnBRT5s2TYZhaMCAAZo0aZL8/Pxst3l6eio4OFihoaEuCQkAQFFV4KLu16+fJCkkJERhYWEqUaKEy0IBAIArHH57Vnh4uC5fvqxDhw4pNTVVly9ftru9TZs2TgsHAEBR53BRb926Vb1799axY8dkGIbdbRaLRdnZ2U4LBwBAUedwUQ8dOlTNmjXT559/rsDAwDzPAAcAAM7hcFH//PPP+uSTT3T33Xe7Ig8AALiKw++jbtGihX755RdXZAEAANdweI96xIgReuGFF5SSkqIGDRrkOvu7YcOGTgsHAEBR53BRP/7445KkAQMG2MYsFovtE8s4mQwAAOdxuKiTkpJckQMAAOTB4aKuVq2aK3IAAIA8OHwymSR99NFHuv/++xUUFKRjx45JuvIRo59++qlTwwEAUNQ5XNQzZ87UmDFj1KVLF506dcr2mnSZMmU0bdo0Z+cDAKBIc7io33nnHX344YeaMGGCPDw8bOPNmjXTnj17nBoOAICizuGiTkpKUpMmTXKNW61WnT9/3imhAADAFQ4XdUhIiHbt2pVrfM2aNapbt64zMgEAgP/P4bO+X3rpJQ0fPlx///23DMPQtm3btGjRIsXExGj27NmuyAgAQJHlcFE/++yzunTpkiIiIpSRkaHevXvrrrvu0vTp0/XUU0+5IiMAAEWWw0UtSYMHD9bgwYOVlpamy5cvq1KlSs7OBQAAdJNFnaNChQrOygEAAPLgcFGfPHlSEydO1IYNG5SamqrLly/b3f7XX385LRwAAEWdw0X99NNP6/Dhwxo4cKD8/f1lsVhckcvmt99+09ixY7VmzRplZmaqVq1amjNnjpo2bSpJMgxDkyZN0qxZs5Senq4WLVro3XffVb169VyaCwCAwuBwUX/77bf69ttv1ahRI1fksZOenq77779f7dq105o1a1SpUiUdPnxYZcqUsc2ZOnWqYmNjFRcXp1q1aunVV19Vx44ddfDgQfn6+ro8IwAAruRwUdepU0eZmZmuyJLLlClTVKVKFc2bN882FhwcbPtvwzA0bdo0TZgwQd27d5ckzZ8/X/7+/lq4cKGGDBlSKDkBAHAVhz/w5L333tOECROUkJCgkydP6syZM3YXZ/rss8/UrFkzPfnkk6pUqZKaNGmiDz/80HZ7UlKSUlJS1KlTJ9uY1WpVeHi4Nm/enO92s7KyXJobAABncbioy5Qpo9OnT6t9+/aqVKmSypYtq7Jly6pMmTIqW7asU8MdOXJEM2fOVM2aNbVu3ToNHTpUI0eO1IIFCyRJKSkpkiR/f3+7+/n7+9tuy0tMTIz8/PxslypVqjg1NwAAzuLwoe8+ffrI09NTCxcudPnJZJcvX1azZs0UHR0tSWrSpIn27dunmTNnqm/fvrZ512YwDOO6ucaPH68xY8bYrp85c4ayBgCYksNFvXfvXu3cuVO1a9d2RR47gYGBuT4//J577tGyZcskSQEBAZKu7FkHBgba5qSmpubay76a1WqV1Wp1QWIAAJzL4UPfzZo10/Hjx12RJZf7779fBw8etBs7dOiQqlWrJunKF4QEBAQoPj7edvuFCxeUkJCgsLCwQskIAIArObxHPWLECI0aNUovvfSSGjRooBIlStjd3rBhQ6eFe/755xUWFqbo6Gj16NFD27Zt06xZszRr1ixJVw55jx49WtHR0apZs6Zq1qyp6Oho+fj4qHfv3k7LAQCAuzhc1D179pQkDRgwwDZmsVhsrwtnZ2c7LVzz5s21YsUKjR8/XpMnT1ZISIimTZumPn362OZEREQoMzNTw4YNs33gyfr163kPNQDgjuBwUSclJbkiR766deumbt265Xu7xWJRVFSUoqKiCi8UAACFxOGiznl9GAAAuN5NfXvWb7/9pu+++y7PL+UYOXKkU4IBAICbKOp58+Zp6NCh8vT0VPny5e3er2yxWChqAACcyOGinjhxoiZOnKjx48erWDGH390FAAAc4HDTZmRk6KmnnqKkAQAoBA637cCBA7V06VJXZAEAANdw+NB3TEyMunXrprVr1+b5gSexsbFOCwcAQFHncFFHR0dr3bp1ts/6vvZkMgAA4DwOF3VsbKzmzp2r/v37uyAOAAC4msOvUVutVt1///2uyAIAAK7hcFGPGjVK77zzjiuyAACAazh86Hvbtm36+uuvtXr1atWrVy/XyWTLly93WjgAAIo6h4u6TJky6t69uyuyAACAa9zUR4gCAIDCcVNfyiFJf/75pw4ePCiLxaJatWqpYsWKzswFAAB0EyeTnT9/XgMGDFBgYKDatGmj1q1bKygoSAMHDlRGRoYrMgIAUGQ5XNRjxoxRQkKCVq1apVOnTunUqVP69NNPlZCQoBdeeMEVGQEAKLIcPvS9bNkyffLJJ2rbtq1trEuXLvL29laPHj00c+ZMZ+YDAKBIu6lvz/L39881XqlSJQ59AwDgZA4XdWhoqCIjI/X333/bxjIzMzVp0iSFhoY6NRwAAEWdw4e+p0+froceekiVK1dWo0aNZLFYtGvXLnl5eWndunWuyAgAQJHlcFHXr19fP//8sz7++GMdOHBAhmHoqaeeUp8+feTt7e2KjAAAFFk39T5qb29vDR482NlZAADANRx+jTomJkZz587NNT537lxNmTLFKaEAAMAVDhf1Bx98oDp16uQar1evnt5//32nhAIAAFc4XNQpKSkKDAzMNV6xYkWdOHHCKaEAAMAVDhd1lSpV9N133+Ua/+677xQUFOSUUAAA4AqHTyYbNGiQRo8erYsXL6p9+/aSpK+++koRERF8hCgAAE7mcFFHRETor7/+0rBhw3ThwgVJkpeXl8aOHavx48c7PSAAAEWZw0VtsVg0ZcoUvfLKK0pMTJS3t7dq1qwpq9XqinwAABRpN/191KVKlVLz5s2dmQUAAFzD4ZPJAABA4aGoAQAwMYoaAAATK1BR33vvvUpPT5ckTZ48me+dBgCgkBSoqBMTE3X+/HlJ0qRJk3Tu3DmXhgIAAFcU6Kzvxo0b69lnn1WrVq1kGIbefPNNlSpVKs+5EydOdGpAAACKsgIVdVxcnCIjI7V69WpZLBatWbNGxYvnvqvFYqGoAQBwogIVde3atbV48WJJUrFixfTVV1+pUqVKLg0GAABu4gNPLl++7IocAAAgDzf1yWSHDx/WtGnTlJiYKIvFonvuuUejRo1SjRo1nJ0PAIAizeH3Ua9bt05169bVtm3b1LBhQ9WvX1/ff/+96tWrp/j4eFdkBACgyHJ4j3rcuHF6/vnn9frrr+caHzt2rDp27Oi0cAAAFHUO71EnJiZq4MCBucYHDBig/fv3OyUUAAC4wuGirlixonbt2pVrfNeuXZwJDgCAkzl86Hvw4MH65z//qSNHjigsLEwWi0XffvutpkyZohdeeMEVGQEAKLIcLupXXnlFvr6++u9//6vx48dLkoKCghQVFaWRI0c6PSAAAEWZw0VtsVj0/PPP6/nnn9fZs2clSb6+vk4PBgAAbvJ91DkoaAAAXIvvowYAwMQoagAATIyiBgDAxBwq6osXL6pdu3Y6dOiQq/IAAICrOFTUJUqU0N69e2WxWFyVBwAAXMXhQ999+/bVnDlzXJEFAABcw+G3Z124cEGzZ89WfHy8mjVrppIlS9rdHhsb67RwAAAUdQ4X9d69e3XvvfdKUq7XqjkkDgCAczlc1Bs2bHBFDgAAkIebfnvWL7/8onXr1ikzM1OSZBiG00IBAIArHC7qkydPqkOHDqpVq5a6dOmiEydOSJIGDRrEt2cBAOBkDhf1888/rxIlSig5OVk+Pj628Z49e2rt2rVODQcAQFHn8GvU69ev17p161S5cmW78Zo1a+rYsWNOCwYAAG5ij/r8+fN2e9I50tLSZLVanRIKAABc4XBRt2nTRgsWLLBdt1gsunz5st544w21a9fOqeEAACjqHD70/cYbb6ht27basWOHLly4oIiICO3bt09//fWXvvvuO1dkBACgyHJ4j7pu3bravXu37rvvPnXs2FHnz59X9+7dtXPnTtWoUcMVGQEAKLIc3qOWpICAAE2aNMnZWQAAwDVuqqjT09M1Z84cJSYmymKx6J577tGzzz6rcuXKOTsfAABFmsOHvhMSEhQSEqK3335b6enp+uuvv/T2228rJCRECQkJrsgIAECR5fAe9fDhw9WjRw/NnDlTHh4ekqTs7GwNGzZMw4cP1969e50eEgCAosrhPerDhw/rhRdesJW0JHl4eGjMmDE6fPiwU8MBAFDUOVzU9957rxITE3ONJyYmqnHjxs7IBAAA/r8CHfrevXu37b9HjhypUaNG6ZdfflHLli0lSVu3btW7776r119/3TUpAQAoogq0R924cWM1adJEjRs3Vq9evXT8+HFFRESoTZs2atOmjSIiInTs2DH17t3bpWFjYmJksVg0evRo25hhGIqKilJQUJC8vb3Vtm1b7du3z6U5AAAoLAXao05KSnJ1jhvavn27Zs2apYYNG9qNT506VbGxsYqLi1OtWrX06quvqmPHjjp48KB8fX3dlBYAAOcoUFFXq1bN1Tmu69y5c+rTp48+/PBDvfrqq7ZxwzA0bdo0TZgwQd27d5ckzZ8/X/7+/lq4cKGGDBnirsgAADjFTX3gyW+//abvvvtOqampunz5st1tI0eOdEqwqw0fPlxdu3bVAw88YFfUSUlJSklJUadOnWxjVqtV4eHh2rx5c75FnZWVpaysLNv1M2fOOD0znCM5OVlpaWnujnFDFSpUUNWqVd0dA8AdyOGinjdvnoYOHSpPT0+VL19eFovFdpvFYnF6US9evFg//vijtm/fnuu2lJQUSZK/v7/duL+//3W/GzsmJoaPQL0NJCcnq06de5SZmeHuKDfk7e2jAwcSKWsATudwUU+cOFETJ07U+PHjVayYw+/ucsjx48c1atQorV+/Xl5eXvnOu/qPBenKIfFrx642fvx4jRkzxnb9zJkzqlKlyq0HhlOlpaUpMzNDLQZEqnRgsLvj5OvMiaP6fu4kpaWlUdQAnM7hos7IyNBTTz3l8pKWpB9++EGpqalq2rSpbSw7O1ubNm3SjBkzdPDgQUlX9qwDAwNtc1JTU3PtZV/NarXKarW6LjicqnRgsMpVre3uGADgFg637cCBA7V06VJXZMmlQ4cO2rNnj3bt2mW7NGvWTH369NGuXbtUvXp1BQQEKD4+3nafCxcuKCEhQWFhYYWSEQAAV3J4jzomJkbdunXT2rVr1aBBA5UoUcLu9tjYWKeF8/X1Vf369e3GSpYsqfLly9vGR48erejoaNWsWVM1a9ZUdHS0fHx8XP6ebgAACoPDRR0dHa1169apdu0rhyKvPZmssEVERCgzM1PDhg1Tenq6WrRoofXr1/MeagDAHcHhoo6NjdXcuXPVv39/F8S5sY0bN9pdt1gsioqKUlRUlFvyAADgSg6/Rm21WnX//fe7IgsAALiGw0U9atQovfPOO67IAgAAruHwoe9t27bp66+/1urVq1WvXr1cJ5MtX77caeEAACjqHC7qMmXK2D5XGwAAuNZNfYQoAAAoHK7/eDEAAHDTHN6jDgkJue77pY8cOXJLgQAAwP9xuKhHjx5td/3ixYvauXOn1q5dq5deeslZuQAAgG6iqEeNGpXn+LvvvqsdO3bcciAAAPB/nPYadefOnbVs2TJnbQ4AAMiJRf3JJ5+oXLlyztocAADQTRz6btKkid3JZIZhKCUlRX/++afee+89p4YDAKCoc7ioH3vsMbvrxYoVU8WKFdW2bVvVqVPHWbkAAIBuoqgjIyNdkQMAAOSBDzwBAMDECrxHXaxYset+0Il05buhL126dMuhAADAFQUu6hUrVuR72+bNm/XOO+/IMAynhAIAAFcUuKgfffTRXGMHDhzQ+PHjtWrVKvXp00f/+c9/nBoOAICi7qZeo/799981ePBgNWzYUJcuXdKuXbs0f/58Va1a1dn5AAAo0hwq6tOnT2vs2LG6++67tW/fPn311VdatWqV6tev76p8AAAUaQU+9D116lRNmTJFAQEBWrRoUZ6HwgEAgHMVuKjHjRsnb29v3X333Zo/f77mz5+f57zly5c7LRwAAEVdgYu6b9++N3x7FgAAcK4CF3VcXJwLYwAAgLzwyWQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYWHF3BwCA21VycrLS0tLcHeOGKlSooKpVq7o7Bm4SRQ0ANyE5OVl16tyjzMwMd0e5IW9vHx04kEhZ36YoagC4CWlpacrMzFCLAZEqHRjs7jj5OnPiqL6fO0lpaWkU9W2KogaAW1A6MFjlqtZ2dwzcwTiZDAAAE6OoAQAwMYoaAAATo6gBADAxihoAABOjqAEAMDGKGgAAE6OoAQAwMYoaAAATo6gBADAxihoAABOjqAEAMDGKGgAAE6OoAQAwMYoaAAATo6gBADAxihoAABOjqAEAMDGKGgAAE6OoAQAwMYoaAAATo6gBADAxUxd1TEyMmjdvLl9fX1WqVEmPPfaYDh48aDfHMAxFRUUpKChI3t7eatu2rfbt2+emxAAAOJepizohIUHDhw/X1q1bFR8fr0uXLqlTp046f/68bc7UqVMVGxurGTNmaPv27QoICFDHjh119uxZNyYHAMA5irs7wPWsXbvW7vq8efNUqVIl/fDDD2rTpo0Mw9C0adM0YcIEde/eXZI0f/58+fv7a+HChRoyZIg7YgMA4DSm3qO+1unTpyVJ5cqVkyQlJSUpJSVFnTp1ss2xWq0KDw/X5s2b891OVlaWzpw5Y3cBAMCMbpuiNgxDY8aMUatWrVS/fn1JUkpKiiTJ39/fbq6/v7/ttrzExMTIz8/PdqlSpYrrggMAcAtum6J+7rnntHv3bi1atCjXbRaLxe66YRi5xq42fvx4nT592nY5fvy40/MCAOAMpn6NOseIESP02WefadOmTapcubJtPCAgQNKVPevAwEDbeGpqaq697KtZrVZZrVbXBQYAwElMvUdtGIaee+45LV++XF9//bVCQkLsbg8JCVFAQIDi4+NtYxcuXFBCQoLCwsIKOy4AAE5n6j3q4cOHa+HChfr000/l6+tre93Zz89P3t7eslgsGj16tKKjo1WzZk3VrFlT0dHR8vHxUe/evd2cHgCAW2fqop45c6YkqW3btnbj8+bNU//+/SVJERERyszM1LBhw5Senq4WLVpo/fr18vX1LeS0AAA4n6mL2jCMG86xWCyKiopSVFSU6wMBAFDITP0aNQAARR1FDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmBhFDQCAiVHUAACYGEUNAICJUdQAAJgYRQ0AgIlR1AAAmJipvz0LgHMlJycrLS3N3TFuqEKFCqpataq7YwCmQFEDRURycrLq1LlHmZkZ7o5yQ97ePjpwIJGyBkRRA0VGWlqaMjMz1GJApEoHBrs7Tr7OnDiq7+dOUlpaGkUNiKIGipzSgcEqV7W2u2MAKCBOJgMAwMQoagAATIyiBgDAxChqAABMjKIGAMDEKGoAAEyMogYAwMQoagAATIyiBgDAxChqAABMjKIGAMDEKGoAAEyMogYAwMQoagAATIyiBgDAxChqAABMjKIGAMDEKGoAAEyMogYAwMQoagAATIyiBgDAxChqAABMjKIGAMDEKGoAAEyMogYAwMQoagAATIyiBgDAxChqAABMjKIGAMDEKGoAAEyMogYAwMQoagAATIyiBgDAxChqAABMjKIGAMDEKGoAAEyMogYAwMQoagAATIyiBgDAxChqAABMjKIGAMDEKGoAAEysuLsDAACQnJystLQ0d8e4oQoVKqhq1aqF+pgUNQDArZKTk1Wnzj3KzMxwd5Qb8vb20YEDiYVa1hQ1AMCt0tLSlJmZoRYDIlU6MNjdcfJ15sRRfT93ktLS0ihqAEDRUzowWOWq1nZ3DNPhZDIAAEyMogYAwMQoagAATIyiBgDAxChqAABMjKIGAMDEKGoAAEyMogYAwMQoagAATIyiBgDAxO6Yon7vvfcUEhIiLy8vNW3aVN988427IwEAcMvuiKJesmSJRo8erQkTJmjnzp1q3bq1OnfurOTkZHdHAwDgltwRRR0bG6uBAwdq0KBBuueeezRt2jRVqVJFM2fOdHc0AABuyW3/7VkXLlzQDz/8oHHjxtmNd+rUSZs3b87zPllZWcrKyrJdP336tCTpzJkzt5zn3LlzkqS/jh3UpazMW96eq5xJuXK04YcffrBlNpuDBw9KYi2dhfV0LtbTeW63tTx37pxT+kKSfH19ZbFYrj/JuM399ttvhiTju+++sxt/7bXXjFq1auV5n8jISEMSFy5cuHDh4tbL6dOnb9hzt/0edY5r/yIxDCPfv1LGjx+vMWPG2K5fvnxZf/31l8qXL3/jv2xu4MyZM6pSpYqOHz+u0qVL39K2ChvZ3YPs7kF29yC7PV9f3xvOue2LukKFCvLw8FBKSordeGpqqvz9/fO8j9VqldVqtRsrU6aMU3OVLl36tvslzEF29yC7e5DdPchecLf9yWSenp5q2rSp4uPj7cbj4+MVFhbmplQAADjHbb9HLUljxozRM888o2bNmik0NFSzZs1ScnKyhg4d6u5oAADckjuiqHv27KmTJ09q8uTJOnHihOrXr68vvvhC1apVK/QsVqtVkZGRuQ6t3w7I7h5kdw+yuwfZHWcxDMMo1EcEAAAFdtu/Rg0AwJ2MogYAwMQoagAATIyiBgDAxChqB2zatEkPP/ywgoKCZLFYtHLlyhveJyEhQU2bNpWXl5eqV6+u999/3/VB8+Bo9o0bN8piseS6HDhwoHACXyUmJkbNmzeXr6+vKlWqpMcee8z22cDXY4a1v5nsZln7mTNnqmHDhrYPdwgNDdWaNWuuex8zrLnkeHazrHleYmJiZLFYNHr06OvOM8vaX60g2c2y9lFRUbkyBAQEXPc+hbXmFLUDzp8/r0aNGmnGjBkFmp+UlKQuXbqodevW2rlzp15++WWNHDlSy5Ytc3HS3BzNnuPgwYM6ceKE7VKzZk0XJcxfQkKChg8frq1btyo+Pl6XLl1Sp06ddP78+XzvY5a1v5nsOdy99pUrV9brr7+uHTt2aMeOHWrfvr0effRR7du3L8/5ZllzyfHsOdy95tfavn27Zs2apYYNG153npnWPkdBs+cww9rXq1fPLsOePXvynVuoa37rX4tRNEkyVqxYcd05ERERRp06dezGhgwZYrRs2dKFyW6sINk3bNhgSDLS09MLJZMjUlNTDUlGQkJCvnPMuvYFyW7mtS9btqwxe/bsPG8z65rnuF52M6752bNnjZo1axrx8fFGeHi4MWrUqHznmm3tHclulrWPjIw0GjVqVOD5hbnm7FG70JYtW9SpUye7sQcffFA7duzQxYsX3ZTKMU2aNFFgYKA6dOigDRs2uDuOpP/7WtJy5crlO8esa1+Q7DnMtPbZ2dlavHixzp8/r9DQ0DznmHXNC5I9h5nWfPjw4erataseeOCBG84129o7kj2HGdb+559/VlBQkEJCQvTUU0/pyJEj+c4tzDW/Iz6ZzKxSUlJyfTGIv7+/Ll26pLS0NAUGBrop2Y0FBgZq1qxZatq0qbKysvTRRx+pQ4cO2rhxo9q0aeO2XIZhaMyYMWrVqpXq16+f7zwzrn1Bs5tp7ffs2aPQ0FD9/fffKlWqlFasWKG6devmOddsa+5IdjOtuSQtXrxYP/74o7Zv316g+WZae0ezm2XtW7RooQULFqhWrVr6448/9OqrryosLEz79u1T+fLlc80vzDWnqF0sr6/fzGvcbGrXrq3atWvbroeGhur48eN688033VrUzz33nHbv3q1vv/32hnPNtvYFzW6mta9du7Z27dqlU6dOadmyZerXr58SEhLyLTwzrbkj2c205sePH9eoUaO0fv16eXl5Ffh+Zlj7m8lulrXv3Lmz7b8bNGig0NBQ1ahRQ/Pnz7f7WuSrFdaac+jbhQICAvL8+s3ixYvn+Rea2bVs2VI///yz2x5/xIgR+uyzz7RhwwZVrlz5unPNtvaOZM+Lu9be09NTd999t5o1a6aYmBg1atRI06dPz3Ou2dbckex5cdea//DDD0pNTVXTpk1VvHhxFS9eXAkJCXr77bdVvHhxZWdn57qPWdb+ZrLnxd3/1khSyZIl1aBBg3xzFOaas0ftQqGhoVq1apXd2Pr169WsWTOVKFHCTalu3s6dO912yHjEiBFasWKFNm7cqJCQkBvexyxrfzPZ8+Kutb+WYRjKysrK8zazrHl+rpc9L+5a8w4dOuQ62/jZZ59VnTp1NHbsWHl4eOS6j1nW/may58UMv+9ZWVlKTExU69at87y9UNfc6aen3cHOnj1r7Ny509i5c6chyYiNjTV27txpHDt2zDAMwxg3bpzxzDPP2OYfOXLE8PHxMZ5//nlj//79xpw5c4wSJUoYn3zyiemzv/XWW8aKFSuMQ4cOGXv37jXGjRtnSDKWLVtW6Nn/9a9/GX5+fsbGjRuNEydO2C4ZGRm2OWZd+5vJbpa1Hz9+vLFp0yYjKSnJ2L17t/Hyyy8bxYoVM9avX59nbrOs+c1kN8ua5+faM6fNvPbXulF2s6z9Cy+8YGzcuNE4cuSIsXXrVqNbt26Gr6+vcfTo0TxzF+aaU9QOyHkbwbWXfv36GYZhGP369TPCw8Pt7rNx40ajSZMmhqenpxEcHGzMnDmz8IMbjmefMmWKUaNGDcPLy8soW7as0apVK+Pzzz93S/a8cksy5s2bZ5tj1rW/mexmWfsBAwYY1apVMzw9PY2KFSsaHTp0sBVdXrkNwxxrbhiOZzfLmufn2rIz89pf60bZzbL2PXv2NAIDA40SJUoYQUFBRvfu3Y19+/blm9swCm/N+ZpLAABMjJPJAAAwMYoaAAATo6gBADAxihoAABOjqAEAMDGKGgAAE6OoAQAwMYoagMOCg4M1bdq0As2Ni4tTmTJlXJoHuJPxgSfAHa5t27Zq3LhxgYu1IP7880+VLFlSPj4+N5ybmZmps2fPqlKlSpKkqKgorVy5Urt27XJaHuBOxpdyAHBYxYoVCzzX29tb3t7eLkwD3Nk49A3cwfr376+EhARNnz5dFotFFotFR48eVUJCgu677z5ZrVYFBgZq3LhxunTpkiRpwYIFKlWqlN3X+40YMUK1atXS+fPnJeU+9H3q1Cn985//lL+/v7y8vFS/fn2tXr1akv2h77i4OE2aNEk//fSTLU9cXJwGDBigbt262WW/dOmSAgICNHfuXBeuEGB+7FEDd7Dp06fr0KFDql+/viZPnixJys7OVpcuXdS/f38tWLBABw4c0ODBg+Xl5aWoqCj17dtXq1evVp8+fbR582Z9+eWX+uCDD/Tdd9+pZMmSuR7j8uXL6ty5s86ePauPP/5YNWrU0P79+/P8SsOePXtq7969Wrt2rb788ktJkp+fn2rVqqU2bdroxIkTtq83/OKLL3Tu3Dn16NHDhSsEmB9FDdzB/Pz85OnpKR8fHwUEBEiSJkyYoCpVqmjGjBmyWCyqU6eOfv/9d40dO1YTJ05UsWLF9MEHH6hhw4YaOXKkli9frsjISDVv3jzPx/jyyy+1bds2JSYmqlatWpKk6tWr5znX29tbpUqVUvHixW15JCksLEy1a9fWRx99pIiICEnSvHnz9OSTT6pUqVLOXBLgtsOhb6CISUxMVGhoqCwWi23s/vvv17lz5/Trr79KksqWLas5c+Zo5syZqlGjhsaNG5fv9nbt2qXKlSvbSvpmDRo0SPPmzZMkpaam6vPPP9eAAQNuaZvAnYCiBooYwzDsSjpnTJLd+KZNm+Th4aHff//d9tp0Xpx1oljfvn115MgRbdmyRR9//LGCg4PVunVrp2wbuJ1R1MAdztPTU9nZ2bbrdevW1ebNm3X1OzM3b94sX19f3XXXXbbrU6dO1apVq1S6dGmNGDEi3+03bNhQv/76qw4dOnRTeXKUL19ejz32mObNm6d58+bp2WefLehTBO5oFDVwhwsODtb333+vo0ePKi0tTcOGDdPx48c1YsQIHThwQJ9++qkiIyM1ZswYFStWTGfPntUzzzyjESNGqHPnzlq4cKH+97//aenSpXluPzw8XG3atNHjjz+u+Ph4JSUlac2aNVq7dm2+eZKSkrRr1y6lpaUpKyvLdtugQYM0f/58JSYmql+/fi5ZD+B2Q1EDd7gXX3xRHh4eqlu3ripWrKiLFy/qiy++0LZt29SoUSMNHTpUAwcO1L///W9J0qhRo1SyZElFR0dLkurVq6cpU6Zo6NCh+u233/J8jGXLlql58+bq1auX6tatq4iIiDz3miXp8ccf10MPPaR27dqpYsWKWrRoke22Bx54QIGBgXrwwQcVFBTk5JUAbk98MhkA08jIyFBQUJDmzp2r7t27uzsOYAq8PQuA212+fFkpKSn673//Kz8/Pz3yyCPujgSYBkUNwO2Sk5MVEhKiypUrKy4uTsWL808TkIND3wAAmBgnkwEAYGIUNQAAJkZRAwBgYhQ1AAAmRlEDAGBiFDUAACZGUQMAYGIUNQAAJkZRAwBgYv8PNbYtU6CgY1wAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 500x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.displot(toxicity_df.toxicity)\n",
    "plt.title(\"LLM toxicity classifier output\")\n",
    "plt.ylabel(\"Number of comments\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8a974f3d-6c84-4c3e-8b5f-ca62d226f899",
   "metadata": {},
   "source": [
    "### Data Transformation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eb37eb5c-8e9f-4536-ab21-a643f17478b0",
   "metadata": {},
   "source": [
    "We will now repeat the procedure for training the same models as above. We will not be including a validation set because of the critically low amount of data we have at our disposal."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "b34b4fdd-6b62-468b-9a86-60bf9e7783ee",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "\n",
    "data_train, data_test = train_test_split(toxicity_df, random_state=42)\n",
    "# new vectorizer because of new data\n",
    "vectorizer = TfidfVectorizer().fit(data_train.comments)\n",
    "x_train = vectorizer.transform(data_train.comments)\n",
    "y_train = data_train.toxicity\n",
    "x_test = vectorizer.transform(data_test.comments)\n",
    "y_test = data_test.toxicity"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "15b8f762-77f9-4362-9d87-7808c44c0e22",
   "metadata": {},
   "source": [
    "### Model Selection"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dc47ab8f-b4b0-4c20-908f-94f14166133e",
   "metadata": {},
   "source": [
    "#### Dummy Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "35774f77-ad7a-4fba-9b44-6a21333e9a76",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.1408450704225352\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           1       0.54      1.00      0.70        25\n",
      "           2       0.00      0.00      0.00        13\n",
      "           3       0.00      0.00      0.00         4\n",
      "           4       0.00      0.00      0.00         3\n",
      "           5       0.00      0.00      0.00         1\n",
      "\n",
      "    accuracy                           0.54        46\n",
      "   macro avg       0.11      0.20      0.14        46\n",
      "weighted avg       0.30      0.54      0.38        46\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.dummy import DummyClassifier\n",
    "\n",
    "\n",
    "majority = DummyClassifier(strategy=\"most_frequent\")\n",
    "majority.fit(x_train, y_train)\n",
    "majority_res = majority.predict(x_test)\n",
    "\n",
    "get_statistics(y_test, majority_res)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1c117497-cb30-4df8-afd2-61ef3da02068",
   "metadata": {},
   "source": [
    "#### Naive Bayes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "19d1cc4f-8530-4795-b4e9-c027868e8660",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Naive Bayes mean macro F1-score 0.2797, std: 0.1395\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\user\\anaconda3\\envs\\manis\\Lib\\site-packages\\sklearn\\model_selection\\_split.py:725: UserWarning: The least populated class in y has only 3 members, which is less than n_splits=5.\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "naive_x_train = x_train.toarray()\n",
    "naive_x_test = x_test.toarray()\n",
    "\n",
    "naive_model = MultinomialNB()\n",
    "res = cross_val_res(naive_model, naive_x_train, y_train, cv=5)\n",
    "print(f\"Naive Bayes mean macro F1-score {res[0]:.4f}, std: {res[1]:.4f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "3d774a28-0ff7-4c4a-bf88-800bc5c5462b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.21913043478260869\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           1       0.55      0.96      0.70        25\n",
      "           2       0.00      0.00      0.00        13\n",
      "           3       0.00      0.00      0.00         4\n",
      "           4       0.50      0.33      0.40         3\n",
      "           5       0.00      0.00      0.00         1\n",
      "\n",
      "    accuracy                           0.54        46\n",
      "   macro avg       0.21      0.26      0.22        46\n",
      "weighted avg       0.33      0.54      0.40        46\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "naive_model = MultinomialNB().fit(naive_x_train, y_train)\n",
    "naive_res = naive_model.predict(naive_x_test)\n",
    "\n",
    "get_statistics(y_test, naive_res)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0bfc5307-89fa-4663-a406-201c54a507ef",
   "metadata": {},
   "source": [
    "#### Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "id": "126ce160-f513-4fa9-8cc3-219224ce4e39",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Logistic Regression mean macro F1-score 0.1818, std: 0.4605\n"
     ]
    }
   ],
   "source": [
    "with warnings.catch_warnings():\n",
    "    # ignore warnings about deprecated methods in libraries\n",
    "    warnings.simplefilter(\"ignore\")\n",
    "\n",
    "    lr = LogisticRegression(max_iter=1000)\n",
    "    res = cross_val_res(lr, x_train, y_train)\n",
    "    print(f\"Logistic Regression mean macro F1-score {res[0]:.4f}, std: {res[1]:.4f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "id": "9aec7e92-9cd6-4c92-944e-3e59cd9ef7e2",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.21913043478260869\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           1       0.55      0.96      0.70        25\n",
      "           2       0.00      0.00      0.00        13\n",
      "           3       0.00      0.00      0.00         4\n",
      "           4       0.50      0.33      0.40         3\n",
      "           5       0.00      0.00      0.00         1\n",
      "\n",
      "    accuracy                           0.54        46\n",
      "   macro avg       0.21      0.26      0.22        46\n",
      "weighted avg       0.33      0.54      0.40        46\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "with warnings.catch_warnings():\n",
    "    warnings.simplefilter(\"ignore\")\n",
    "    lr = LogisticRegression(max_iter=1000).fit(x_train, y_train)\n",
    "    lr_res = lr.predict(x_test)\n",
    "\n",
    "get_statistics(y_test, lr_res)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0d6be997-d438-4452-b2dc-17365aecf9f0",
   "metadata": {},
   "source": [
    "#### Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "id": "6407054c-546e-4f02-b442-49d9bed8fda3",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\user\\anaconda3\\envs\\manis\\Lib\\site-packages\\sklearn\\model_selection\\_split.py:725: UserWarning: The least populated class in y has only 3 members, which is less than n_splits=10.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Random Forest mean macro F1: 0.3092, std: 0.4472\n"
     ]
    }
   ],
   "source": [
    "forest_model = RandomForestClassifier(n_estimators=200, \n",
    "                                      n_jobs=-1, \n",
    "                                      criterion=\"entropy\")\n",
    "res = cross_val_res(forest_model, x_train, y_train)\n",
    "print(f\"Random Forest mean macro F1: {res[0]:.4f}, std: {res[1]:.4f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "id": "bae3ffa1-5ac5-488c-b66d-47954275ad07",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:    0.4s\n",
      "[Parallel(n_jobs=-1)]: Done 200 out of 200 | elapsed:    0.5s finished\n",
      "[Parallel(n_jobs=8)]: Using backend ThreadingBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=8)]: Done  34 tasks      | elapsed:    0.0s\n",
      "[Parallel(n_jobs=8)]: Done 184 tasks      | elapsed:    0.0s\n",
      "[Parallel(n_jobs=8)]: Done 200 out of 200 | elapsed:    0.0s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.22117647058823525\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           1       0.56      0.96      0.71        25\n",
      "           2       0.00      0.00      0.00        13\n",
      "           3       0.00      0.00      0.00         4\n",
      "           4       0.50      0.33      0.40         3\n",
      "           5       0.00      0.00      0.00         1\n",
      "\n",
      "    accuracy                           0.54        46\n",
      "   macro avg       0.21      0.26      0.22        46\n",
      "weighted avg       0.34      0.54      0.41        46\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "forest_model = RandomForestClassifier(n_estimators=200,\n",
    "                                      n_jobs=-1,\n",
    "                                      criterion=\"entropy\", \n",
    "                                      verbose=1).fit(x_train, y_train)\n",
    "forest_pred = forest_model.predict(x_test)\n",
    "get_statistics(y_test, forest_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1402517e-99e0-4d5e-9d93-19fbb1bb166b",
   "metadata": {},
   "source": [
    "#### Adaboost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "id": "57a85ff7-b7ca-43ac-97f1-1d662e0d8e00",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AdaBoost mean macro F1: 0.2347, std: 0.2274\n"
     ]
    }
   ],
   "source": [
    "ada_model = AdaBoostClassifier(n_estimators=50)\n",
    "res = cross_val_res(ada_model, x_train, y_train, cv=3)\n",
    "print(f\"AdaBoost mean macro F1: {res[0]:.4f}, std: {res[1]:.4f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "id": "dea91b85-d109-4d90-a4d9-996f6e2d7433",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Macro F1: 0.24285714285714288\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           1       0.56      1.00      0.71        25\n",
      "           2       0.00      0.00      0.00        13\n",
      "           3       0.00      0.00      0.00         4\n",
      "           4       1.00      0.33      0.50         3\n",
      "           5       0.00      0.00      0.00         1\n",
      "\n",
      "    accuracy                           0.57        46\n",
      "   macro avg       0.31      0.27      0.24        46\n",
      "weighted avg       0.37      0.57      0.42        46\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ada_model = AdaBoostClassifier(n_estimators=100).fit(x_train, y_train)\n",
    "ada_pred = ada_model.predict(x_test)\n",
    "get_statistics(y_test, ada_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cb4ee783-5e74-4f34-b06b-3a4c04768ebb",
   "metadata": {},
   "source": [
    "### Classifiying the crawled data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2b88df92-ae84-47cc-bc06-3e7ce27f3a1e",
   "metadata": {},
   "source": [
    "We can clearly see that none of our classifiers can make any meaningful guesses, being marginally better than a simple majority classifier, which is entirely the fault of our low amount of data. Because of the restrictions placed upon this project, we cannot augment this data in any meaningful way, and thus will continue the analysis with faulty classifiers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "id": "1afc59de-c15d-43ae-b126-38440266fb31",
   "metadata": {
    "scrolled": true
   },
   "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>title</th>\n",
       "      <th>link</th>\n",
       "      <th>source</th>\n",
       "      <th>text</th>\n",
       "      <th>date</th>\n",
       "      <th>language</th>\n",
       "      <th>toxicity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>ANTIGUAS CANCIONES DE GRECIA PAIS NATAL DE MIS...</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>other</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>Încă  o zi petrecută cu muzica voastră  fantas...</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>el</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>FelicităriSuperb \\nSă fiți mereu bine \\nMomen...</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>greeklish</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>Lovely collection! Thank you &lt;3</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>greeklish</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...</td>\n",
       "      <td>https://www.youtube.com/watch?v=p5g82ta4sTk&amp;pp...</td>\n",
       "      <td>song</td>\n",
       "      <td>Beautiful thank you  so many memories awesome ...</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>en</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2692</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Copy by fantaros</td>\n",
       "      <td>2021-11-29</td>\n",
       "      <td>el</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2693</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Fist</td>\n",
       "      <td>2021-11-29</td>\n",
       "      <td>el</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2694</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Yui</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>el</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2695</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Uihhii</td>\n",
       "      <td>2022-11-29</td>\n",
       "      <td>el</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2696</th>\n",
       "      <td>😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...</td>\n",
       "      <td>https://www.youtube.com/watch?v=6ay2HZwz2sA&amp;pp...</td>\n",
       "      <td>gaming</td>\n",
       "      <td>Zzfitdt</td>\n",
       "      <td>2021-11-29</td>\n",
       "      <td>el</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2359 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  title  \\\n",
       "0     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "1     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "2     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "3     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "4     Μουσική ιστορία Νο.1 (μέρος πρώτο) - 100 χρυσά...   \n",
       "...                                                 ...   \n",
       "2692  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2693  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2694  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2695  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "2696  😱ΑΥΤΟΣ ΕΧΕΙ '0Views'(React Σε Montages Με '0'V...   \n",
       "\n",
       "                                                   link  source  \\\n",
       "0     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "1     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "2     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "3     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "4     https://www.youtube.com/watch?v=p5g82ta4sTk&pp...    song   \n",
       "...                                                 ...     ...   \n",
       "2692  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2693  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2694  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2695  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "2696  https://www.youtube.com/watch?v=6ay2HZwz2sA&pp...  gaming   \n",
       "\n",
       "                                                   text        date  \\\n",
       "0     ANTIGUAS CANCIONES DE GRECIA PAIS NATAL DE MIS...  2022-11-29   \n",
       "1     Încă  o zi petrecută cu muzica voastră  fantas...  2022-11-29   \n",
       "2      FelicităriSuperb \\nSă fiți mereu bine \\nMomen...  2022-11-29   \n",
       "3                       Lovely collection! Thank you <3  2022-11-29   \n",
       "4     Beautiful thank you  so many memories awesome ...  2022-11-29   \n",
       "...                                                 ...         ...   \n",
       "2692                                   Copy by fantaros  2021-11-29   \n",
       "2693                                               Fist  2021-11-29   \n",
       "2694                                                Yui  2022-11-29   \n",
       "2695                                             Uihhii  2022-11-29   \n",
       "2696                                            Zzfitdt  2021-11-29   \n",
       "\n",
       "       language  toxicity  \n",
       "0         other         1  \n",
       "1            el         1  \n",
       "2     greeklish         1  \n",
       "3     greeklish         1  \n",
       "4            en         1  \n",
       "...         ...       ...  \n",
       "2692         el         1  \n",
       "2693         el         1  \n",
       "2694         el         1  \n",
       "2695         el         1  \n",
       "2696         el         1  \n",
       "\n",
       "[2359 rows x 7 columns]"
      ]
     },
     "execution_count": 83,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "crawl_df[\"toxicity\"] = ada_model.predict(vectorizer.transform(crawl_df.text))\n",
    "crawl_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9f9b95fd-85f8-4977-90a9-e4829f048219",
   "metadata": {},
   "source": [
    "### Analysing the Toxicity of crawled data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "353410a7-0cc3-40c7-b0a8-456b08589c9d",
   "metadata": {},
   "source": [
    "We will now execute the scripts necessary for the analysis available at `report.pdf`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "id": "e62ee869-393a-45a7-9362-fa2ab5637907",
   "metadata": {},
   "outputs": [],
   "source": [
    "def export_to_latex(df, name, col_format, caption=None):\n",
    "    \"\"\"\n",
    "    Export a pandas DataFrame to a LaTeX file.\n",
    "\n",
    "    :param df: The DataFrame to be exported.\n",
    "    :type df: pd.DataFrame\n",
    "\n",
    "    :param name: The name of the LaTeX file (excluding the '.tex' extension).\n",
    "    :type name: str\n",
    "\n",
    "    :param col_format: A string specifying the column formatting for the LaTeX table.\n",
    "    :type col_format: str\n",
    "\n",
    "    :param caption: The caption for the LaTeX table (optional).\n",
    "    :type caption: str, optional\n",
    "\n",
    "    :return: This function does not return anything.\n",
    "    \"\"\"\n",
    "    path = os.path.join(RESOURCE_OUTPUT, name)\n",
    "    df.to_latex(buf=path,\n",
    "                index=False,\n",
    "                formatters={\"name\": str.upper},\n",
    "                float_format=\"{:.3f}\".format,\n",
    "                label=\"tab::\" + name.split(\".\")[0],\n",
    "                caption=caption,\n",
    "                escape=True,\n",
    "               encoding=\"utf-8\",\n",
    "               column_format=col_format)\n",
    "    print(f\"Dataframe exported to {path}\")\n",
    "\n",
    "\n",
    "def remove_emojis(text):\n",
    "    return emoji_pattern.sub(r'', text)\n",
    "\n",
    "long_col_format = '|p{10cm}|p{1cm}|'"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f7bfdace-8ae4-4798-9a13-21e34c78766f",
   "metadata": {},
   "source": [
    "#### Finding the most toxic language"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "id": "f23595dd-85cc-411d-ad92-6392fa846945",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataframe exported to results\\toxic_lang.tex\n"
     ]
    },
    {
     "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>language</th>\n",
       "      <th>toxicity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>el</td>\n",
       "      <td>1.003053</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>en</td>\n",
       "      <td>1.117949</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>greeklish</td>\n",
       "      <td>1.004819</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>other</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    language  toxicity\n",
       "0         el  1.003053\n",
       "1         en  1.117949\n",
       "2  greeklish  1.004819\n",
       "3      other  1.000000"
      ]
     },
     "execution_count": 85,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "toxic_lang_df = crawl_df.loc[:, [\"language\", \"toxicity\"]].groupby(\"language\").mean().reset_index()\n",
    "\n",
    "caption = \"Average toxicity by language.\"\n",
    "export_to_latex(toxic_lang_df, \"toxic_lang.tex\", caption=caption, col_format=\"|p{3.5cm}|p{1cm}|\")\n",
    "\n",
    "toxic_lang_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6e1e4b1d-b290-4d45-926e-0d5663bc6d9e",
   "metadata": {},
   "source": [
    "#### Finding the most toxic video"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "id": "c5f296c4-4925-4abe-a7a2-881026081cad",
   "metadata": {
    "scrolled": true
   },
   "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>toxicity</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>link</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=Iz1U4yxmRT4&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <td>1.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=2duliv41A1I&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <td>1.428571</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=m7gGlw93Mq0&amp;pp=ygUOZ3JlZWsgZm9ydG5pdGU%3D</th>\n",
       "      <td>1.285714</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=d-6Y4vE3g8U&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <td>1.250000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=ii2To2gvzkU&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <td>1.166667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=KLBYBvxiTvQ&amp;pp=ygUOZ3JlZWsgZm9ydG5pdGU%3D</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=JeNV28qWi_A&amp;pp=ygUOZ3JlZWsgZm9ydG5pdGU%3D</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=HILSvOQV_bc&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=H-TIQdWiuOg&amp;pp=ygUOZ3JlZWsgZm9ydG5pdGU%3D</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=zHQBNoNpmog&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>116 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    toxicity\n",
       "link                                                        \n",
       "https://www.youtube.com/watch?v=Iz1U4yxmRT4&pp=...  1.500000\n",
       "https://www.youtube.com/watch?v=2duliv41A1I&pp=...  1.428571\n",
       "https://www.youtube.com/watch?v=m7gGlw93Mq0&pp=...  1.285714\n",
       "https://www.youtube.com/watch?v=d-6Y4vE3g8U&pp=...  1.250000\n",
       "https://www.youtube.com/watch?v=ii2To2gvzkU&pp=...  1.166667\n",
       "...                                                      ...\n",
       "https://www.youtube.com/watch?v=KLBYBvxiTvQ&pp=...  1.000000\n",
       "https://www.youtube.com/watch?v=JeNV28qWi_A&pp=...  1.000000\n",
       "https://www.youtube.com/watch?v=HILSvOQV_bc&pp=...  1.000000\n",
       "https://www.youtube.com/watch?v=H-TIQdWiuOg&pp=...  1.000000\n",
       "https://www.youtube.com/watch?v=zHQBNoNpmog&pp=...  1.000000\n",
       "\n",
       "[116 rows x 1 columns]"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "toxic_videos_df = crawl_df.loc[:, [\"link\", \"toxicity\"]].groupby(\"link\").mean().sort_values(\"toxicity\", ascending=False)\n",
    "toxic_videos_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "id": "acb7f375-3681-4a87-8aa5-0ae77f6e7a63",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataframe exported to results\\toxic_videos.tex\n"
     ]
    },
    {
     "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>title</th>\n",
       "      <th>toxicity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Ρέμος: Αυτά όχι σε μένα, άσε τις μαγκιές</td>\n",
       "      <td>1.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Αγαπημένα Ελληνικά Τραγούδια / Greek Music Non...</td>\n",
       "      <td>1.428571</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ΟΤΙ ΒΡΩ ΣΤΟ ORTNIT ΤΟ ΤΡΩΩ CHALLNG! (ortnite G...</td>\n",
       "      <td>1.285714</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Δεν Μιλούν Για Τον Μπρούνο (Από το “Ενκάντο: Έ...</td>\n",
       "      <td>1.250000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Ένα τραγούδι η ζωή μας - 70 αγαπημένα τραγούδι...</td>\n",
       "      <td>1.166667</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               title  toxicity\n",
       "0           Ρέμος: Αυτά όχι σε μένα, άσε τις μαγκιές  1.500000\n",
       "1  Αγαπημένα Ελληνικά Τραγούδια / Greek Music Non...  1.428571\n",
       "2  ΟΤΙ ΒΡΩ ΣΤΟ ORTNIT ΤΟ ΤΡΩΩ CHALLNG! (ortnite G...  1.285714\n",
       "3  Δεν Μιλούν Για Τον Μπρούνο (Από το “Ενκάντο: Έ...  1.250000\n",
       "4  Ένα τραγούδι η ζωή μας - 70 αγαπημένα τραγούδι...  1.166667"
      ]
     },
     "execution_count": 87,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "toxic_videos_df = toxic_videos_df.merge(\n",
    "                    crawl_df.loc[:, [\"link\", \"title\"]].drop_duplicates(), \n",
    "                    on=\"link\", \n",
    "                    how=\"inner\").loc[:, [\"title\", \"toxicity\"]].head(5)\n",
    "\n",
    "toxic_videos_df.title = toxic_videos_df.title.apply(lambda x: remove_emojis(x))\n",
    "caption = \"The top 5 videos with the most toxic comments on average.\"\n",
    "export_to_latex(toxic_videos_df, \"toxic_videos.tex\", caption=caption, col_format=long_col_format)\n",
    "\n",
    "toxic_videos_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "48aac27c-20e0-403b-8fdf-5602a31bcdec",
   "metadata": {},
   "source": [
    "#### Finding videos where toxicity was uniform across time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "id": "da1a158f-a837-43ec-b436-5099e9f60064",
   "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></th>\n",
       "      <th>toxicity</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>link</th>\n",
       "      <th>date</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=2duliv41A1I&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <th>2022-12-29</th>\n",
       "      <td>4.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=Iz1U4yxmRT4&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <th>2023-11-24</th>\n",
       "      <td>4.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=d-6Y4vE3g8U&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <th>2023-06-29</th>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=q7elHOjAyEY&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <th>2023-07-29</th>\n",
       "      <td>1.307692</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=m7gGlw93Mq0&amp;pp=ygUOZ3JlZWsgZm9ydG5pdGU%3D</th>\n",
       "      <th>2021-11-29</th>\n",
       "      <td>1.285714</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"3\" valign=\"top\">https://www.youtube.com/watch?v=JeNV28qWi_A&amp;pp=ygUOZ3JlZWsgZm9ydG5pdGU%3D</th>\n",
       "      <th>2023-11-20</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2023-11-17</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2023-11-15</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=Iz1U4yxmRT4&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <th>2023-11-28</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>https://www.youtube.com/watch?v=zHQBNoNpmog&amp;pp=ygUjzrXOu867zrfOvc65zrrOsSDPhM-BzrHOs86_z4XOtM65zrE%3D</th>\n",
       "      <th>2023-06-29</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>327 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                               toxicity\n",
       "link                                               date                \n",
       "https://www.youtube.com/watch?v=2duliv41A1I&pp=... 2022-12-29  4.000000\n",
       "https://www.youtube.com/watch?v=Iz1U4yxmRT4&pp=... 2023-11-24  4.000000\n",
       "https://www.youtube.com/watch?v=d-6Y4vE3g8U&pp=... 2023-06-29  2.000000\n",
       "https://www.youtube.com/watch?v=q7elHOjAyEY&pp=... 2023-07-29  1.307692\n",
       "https://www.youtube.com/watch?v=m7gGlw93Mq0&pp=... 2021-11-29  1.285714\n",
       "...                                                                 ...\n",
       "https://www.youtube.com/watch?v=JeNV28qWi_A&pp=... 2023-11-20  1.000000\n",
       "                                                   2023-11-17  1.000000\n",
       "                                                   2023-11-15  1.000000\n",
       "https://www.youtube.com/watch?v=Iz1U4yxmRT4&pp=... 2023-11-28  1.000000\n",
       "https://www.youtube.com/watch?v=zHQBNoNpmog&pp=... 2023-06-29  1.000000\n",
       "\n",
       "[327 rows x 1 columns]"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# get the mean toxicity per day\n",
    "toxic_time_df = crawl_df.loc[:, [\"link\", \"date\", \"toxicity\"]].groupby([\"link\", \"date\"]).mean().sort_values(\"toxicity\", ascending=False)\n",
    "toxic_time_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "id": "d07de243-5172-4d99-8a62-3bb10c1055b7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataframe exported to results\\toxic_uniform.tex\n"
     ]
    },
    {
     "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>title</th>\n",
       "      <th>toxicity</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Greek Music Mix 2021 - Ελληνικα Τραγουδια Mix ...</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ΠΩΣ ΕΧΑΣΑ 100€ ΣΤΟ ORTNIT! *15.000 VBUCKS* (or...</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Ένα τραγούδι η ζωή μας - 70 αγαπημένα τραγούδι...</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ΚΑΘΕ KILL ΑΛΛΑΖΩ ΠΛΗΚΤΡΟΛΟΓΙΟ CHALLNG! (ortnit...</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Ελληνικό Έντεχνο - λαφρολαϊκό mix</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>108</th>\n",
       "      <td>ΕΠΙΚΗ ΠΡΩΤΗ ΝΙΚΗ ΣΤΟ ORTNIT ft Alex (LPDudes) ...</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>109</th>\n",
       "      <td>ΓΙΑ ΚΑΘΕ DATH ΤΡΩΩ ΚΑΥΤΕΡΗ ΚΟΤΟΜΠΟΥΚΙΑ! (ortni...</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>110</th>\n",
       "      <td>Νατάσα Θεοδωρίδου - Αν πας με άλλη (AI Cover)</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>111</th>\n",
       "      <td>Ρέμος: Αυτά όχι σε μένα, άσε τις μαγκιές</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>112</th>\n",
       "      <td>ΕΠΕΣΤΡΕΨΑ ΣΤΟ OG ORTNIT!</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>113 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 title  toxicity\n",
       "0    Greek Music Mix 2021 - Ελληνικα Τραγουδια Mix ...       1.0\n",
       "1    ΠΩΣ ΕΧΑΣΑ 100€ ΣΤΟ ORTNIT! *15.000 VBUCKS* (or...       1.0\n",
       "2    Ένα τραγούδι η ζωή μας - 70 αγαπημένα τραγούδι...       1.0\n",
       "3    ΚΑΘΕ KILL ΑΛΛΑΖΩ ΠΛΗΚΤΡΟΛΟΓΙΟ CHALLNG! (ortnit...       1.0\n",
       "4                    Ελληνικό Έντεχνο - λαφρολαϊκό mix       1.0\n",
       "..                                                 ...       ...\n",
       "108  ΕΠΙΚΗ ΠΡΩΤΗ ΝΙΚΗ ΣΤΟ ORTNIT ft Alex (LPDudes) ...       1.0\n",
       "109  ΓΙΑ ΚΑΘΕ DATH ΤΡΩΩ ΚΑΥΤΕΡΗ ΚΟΤΟΜΠΟΥΚΙΑ! (ortni...       1.0\n",
       "110      Νατάσα Θεοδωρίδου - Αν πας με άλλη (AI Cover)       1.0\n",
       "111           Ρέμος: Αυτά όχι σε μένα, άσε τις μαγκιές       1.0\n",
       "112                           ΕΠΕΣΤΡΕΨΑ ΣΤΟ OG ORTNIT!       1.0\n",
       "\n",
       "[113 rows x 2 columns]"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# get the std of each link according to all dates\n",
    "toxic_time_var_df = toxic_time_df.groupby([\"link\", \"toxicity\"]).std().reset_index().sort_values(\"toxicity\", ascending=True)\n",
    "uniform_toxic_df = toxic_time_var_df[toxic_time_var_df.toxicity == 1]\n",
    "\n",
    "# get video titles \n",
    "uniform_toxic_df = uniform_toxic_df.merge(\n",
    "                    crawl_df.loc[:, [\"link\", \"title\"]].drop_duplicates(), \n",
    "                    on=\"link\", \n",
    "                    how=\"inner\").loc[:, [\"title\", \"toxicity\"]]\n",
    "\n",
    "# remove emojis to be nice to latex\n",
    "uniform_toxic_df.title = uniform_toxic_df.title.apply(lambda x: remove_emojis(x))\n",
    "caption = \"Videos where comment toxicity stayed uniform over time.\"\n",
    "# only export 30 videos (to fit in latex table)\n",
    "export_to_latex(uniform_toxic_df.head(30), \"toxic_uniform.tex\", caption=caption, col_format=long_col_format)\n",
    "\n",
    "uniform_toxic_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f2998d0f-dc46-413c-8af2-e60a8758f000",
   "metadata": {},
   "source": [
    "#### Finding videos where toxicity increases over time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "id": "c23eef98-1d9f-4f0d-bc19-487f6f69b3f8",
   "metadata": {
    "scrolled": true
   },
   "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>link</th>\n",
       "      <th>toxicity_diff</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>https://www.youtube.com/watch?v=0sTegFKn-nQ&amp;pp...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>https://www.youtube.com/watch?v=2duliv41A1I&amp;pp...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>https://www.youtube.com/watch?v=Iz1U4yxmRT4&amp;pp...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>https://www.youtube.com/watch?v=ZTJPZJ453dY&amp;pp...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>https://www.youtube.com/watch?v=b-GnJoG6VE8&amp;pp...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>https://www.youtube.com/watch?v=tGSLjD2YJCY&amp;pp...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>https://www.youtube.com/watch?v=z4DMFzyCkP0&amp;pp...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>https://www.youtube.com/watch?v=ii2To2gvzkU&amp;pp...</td>\n",
       "      <td>2.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>https://www.youtube.com/watch?v=d-6Y4vE3g8U&amp;pp...</td>\n",
       "      <td>2.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>https://www.youtube.com/watch?v=7wuh7H_PabI&amp;pp...</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>https://www.youtube.com/watch?v=8OGFCAfVHIA&amp;pp...</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>https://www.youtube.com/watch?v=GHzb1liwcsI&amp;pp...</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>https://www.youtube.com/watch?v=ivNQq52XHPc&amp;pp...</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>https://www.youtube.com/watch?v=m7gGlw93Mq0&amp;pp...</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>https://www.youtube.com/watch?v=q7elHOjAyEY&amp;pp...</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 link  toxicity_diff\n",
       "0   https://www.youtube.com/watch?v=0sTegFKn-nQ&pp...       3.000000\n",
       "1   https://www.youtube.com/watch?v=2duliv41A1I&pp...       3.000000\n",
       "5   https://www.youtube.com/watch?v=Iz1U4yxmRT4&pp...       3.000000\n",
       "6   https://www.youtube.com/watch?v=ZTJPZJ453dY&pp...       3.000000\n",
       "7   https://www.youtube.com/watch?v=b-GnJoG6VE8&pp...       3.000000\n",
       "13  https://www.youtube.com/watch?v=tGSLjD2YJCY&pp...       3.000000\n",
       "14  https://www.youtube.com/watch?v=z4DMFzyCkP0&pp...       3.000000\n",
       "9   https://www.youtube.com/watch?v=ii2To2gvzkU&pp...       2.500000\n",
       "8   https://www.youtube.com/watch?v=d-6Y4vE3g8U&pp...       2.333333\n",
       "2   https://www.youtube.com/watch?v=7wuh7H_PabI&pp...       2.000000\n",
       "3   https://www.youtube.com/watch?v=8OGFCAfVHIA&pp...       2.000000\n",
       "4   https://www.youtube.com/watch?v=GHzb1liwcsI&pp...       2.000000\n",
       "10  https://www.youtube.com/watch?v=ivNQq52XHPc&pp...       2.000000\n",
       "11  https://www.youtube.com/watch?v=m7gGlw93Mq0&pp...       2.000000\n",
       "12  https://www.youtube.com/watch?v=q7elHOjAyEY&pp...       2.000000"
      ]
     },
     "execution_count": 90,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "toxic_time_incr_df = crawl_df.loc[:,[\"link\", \"date\", \"toxicity\"]].copy()\n",
    "# sort by date\n",
    "toxic_time_incr_df = toxic_time_incr_df.sort_values(\"date\")\n",
    "# get lag 1 difference between dates\n",
    "toxic_time_incr_df[\"toxicity_diff\"] = toxic_time_incr_df.toxicity.diff()\n",
    "# find where toxicity increases\n",
    "toxic_time_incr_df = toxic_time_incr_df.loc[toxic_time_incr_df.toxicity_diff > 0, :]\n",
    "# get mean growth of toxicity along the dates\n",
    "toxic_time_incr_df = toxic_time_incr_df.loc[:, [\"link\", \"toxicity_diff\"]].groupby(\"link\").mean()\n",
    "# print most toxic videos sorted by toxicity\n",
    "toxic_time_incr_df = toxic_time_incr_df.reset_index().sort_values(\"toxicity_diff\", ascending=False)\n",
    "toxic_time_incr_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "id": "0e61dc76-957a-438b-ba4f-dd1dd6b979d8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataframe exported to results\\toxic_increasing.tex\n"
     ]
    },
    {
     "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>title</th>\n",
       "      <th>toxicity_diff</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Greek Music Mix 2021 - Ελληνικα Τραγουδια Mix ...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Αγαπημένα Ελληνικά Τραγούδια / Greek Music Non...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Ρέμος: Αυτά όχι σε μένα, άσε τις μαγκιές</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ΣΚΟΤΩΣΑ ΤΟΝ MONGRAAL M 2OBOMB !</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ΑΝ ΓΕΛΑΣΕΙΣ ΧΑΝΕΙΣ 500 VBUCKS! (ortnite unny M...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Δοκίμασα 30 PS στο UNRAL RANK...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Ελαφρολαϊκά παλιά - 120 μεγάλες επιτυχίες (by ...</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Ένα τραγούδι η ζωή μας - 70 αγαπημένα τραγούδι...</td>\n",
       "      <td>2.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>Δεν Μιλούν Για Τον Μπρούνο (Από το “Ενκάντο: Έ...</td>\n",
       "      <td>2.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>ΝΙΚΗ ΜΟΝΟ ΜΕ ΜΥΘΙΚΑ ΟΠΛΑ CHALLNG! (ortnite Greek)</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>ΕΒΓΑΛΑΝ ΤΟ *BUILDING* ΣΤΗΝ ΝΕΑ SASON ΤΟΥ ORTNI...</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>Νίκη ΜΟΝΟ με Πιστόλι Challenge (ortnite OG)</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>ΠΩΣ ΕΧΑΣΑ 100€ ΣΤΟ ORTNIT! *15.000 VBUCKS* (or...</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>ΟΤΙ ΒΡΩ ΣΤΟ ORTNIT ΤΟ ΤΡΩΩ CHALLNG! (ortnite G...</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>GRK 2K23 SUMMR MIX | VOL. I | by NIKKOS DINNO ...</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                title  toxicity_diff\n",
       "0   Greek Music Mix 2021 - Ελληνικα Τραγουδια Mix ...       3.000000\n",
       "1   Αγαπημένα Ελληνικά Τραγούδια / Greek Music Non...       3.000000\n",
       "2            Ρέμος: Αυτά όχι σε μένα, άσε τις μαγκιές       3.000000\n",
       "3                     ΣΚΟΤΩΣΑ ΤΟΝ MONGRAAL M 2OBOMB !       3.000000\n",
       "4   ΑΝ ΓΕΛΑΣΕΙΣ ΧΑΝΕΙΣ 500 VBUCKS! (ortnite unny M...       3.000000\n",
       "5                    Δοκίμασα 30 PS στο UNRAL RANK...       3.000000\n",
       "6   Ελαφρολαϊκά παλιά - 120 μεγάλες επιτυχίες (by ...       3.000000\n",
       "7   Ένα τραγούδι η ζωή μας - 70 αγαπημένα τραγούδι...       2.500000\n",
       "8   Δεν Μιλούν Για Τον Μπρούνο (Από το “Ενκάντο: Έ...       2.333333\n",
       "9   ΝΙΚΗ ΜΟΝΟ ΜΕ ΜΥΘΙΚΑ ΟΠΛΑ CHALLNG! (ortnite Greek)       2.000000\n",
       "10  ΕΒΓΑΛΑΝ ΤΟ *BUILDING* ΣΤΗΝ ΝΕΑ SASON ΤΟΥ ORTNI...       2.000000\n",
       "11        Νίκη ΜΟΝΟ με Πιστόλι Challenge (ortnite OG)       2.000000\n",
       "12  ΠΩΣ ΕΧΑΣΑ 100€ ΣΤΟ ORTNIT! *15.000 VBUCKS* (or...       2.000000\n",
       "13  ΟΤΙ ΒΡΩ ΣΤΟ ORTNIT ΤΟ ΤΡΩΩ CHALLNG! (ortnite G...       2.000000\n",
       "14  GRK 2K23 SUMMR MIX | VOL. I | by NIKKOS DINNO ...       2.000000"
      ]
     },
     "execution_count": 91,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "toxic_time_incr_df = toxic_time_incr_df.merge(\n",
    "                    crawl_df.loc[:, [\"link\", \"title\"]].drop_duplicates(), \n",
    "                    on=\"link\", \n",
    "                    how=\"inner\").loc[:, [\"title\", \"toxicity_diff\"]]\n",
    "\n",
    "\n",
    "toxic_time_incr_df.title = toxic_time_incr_df.title.apply(lambda x: remove_emojis(x))\n",
    "caption = \"Videos where comment toxicity stayed increased over time.\"\\\n",
    "            \"The toxicity\\_diff represents the average difference between comment toxicity\"\\\n",
    "            \"with lag 1 across each date.\"\n",
    "export_to_latex(toxic_time_incr_df, \"toxic_increasing.tex\", caption=caption, col_format=long_col_format)\n",
    "\n",
    "toxic_time_incr_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c1fb13a1-d21b-494c-a689-660395a08212",
   "metadata": {},
   "source": [
    "## Exporting the operational dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "id": "bb7f6e4c-9678-4ea3-9c4f-a0121eac910e",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File saved successfully as output\\crawl.csv\n"
     ]
    }
   ],
   "source": [
    "csv_output(crawl_df, \"crawl.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d2f2a56c-ccdc-4173-af79-ac72bd485329",
   "metadata": {},
   "source": [
    "**Thanks for following along!**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "id": "bbd4fa74-04f1-48e0-9954-8d77e9f5685c",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Notebook executed in 51 minutes and 9.0 seconds\n"
     ]
    }
   ],
   "source": [
    "print(f\"Notebook executed in {int((time()-start)// 60)} minutes and {(time()-start) % 60:.1f} seconds\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}