notebooks/eda.ipynb

{
   "cells": [
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "# EDA"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 1,
         "metadata": {},
         "outputs": [
            {
               "name": "stderr",
               "output_type": "stream",
               "text": [
                  "/Users/jonas-mika/Library/Caches/pypoetry/virtualenvs/ml-project-2-mlp-a6NSXBdT-py3.10/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
                  "  from .autonotebook import tqdm as notebook_tqdm\n"
               ]
            }
         ],
         "source": [
            "# ruff: noqa\n",
            "%reload_ext autoreload\n",
            "%autoreload 2\n",
            "\n",
            "# Standard library\n",
            "import os\n",
            "import json\n",
            "\n",
            "# External libraries\n",
            "from sklearn.metrics import accuracy_score, f1_score, precision_score, recall_score, cohen_kappa_score\n",
            "from matplotlib import pyplot as plt\n",
            "import numpy as np\n",
            "import pandas as pd\n",
            "import seaborn as sns\n",
            "import rootutils\n",
            "import hydra\n",
            "\n",
            "# Local imports\n",
            "import ml_project_2_mlp.utils as utils"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 2,
         "metadata": {},
         "outputs": [],
         "source": [
            "# Reinitialize hydra on every run\n",
            "hydra.core.global_hydra.GlobalHydra.instance().clear()\n",
            "h = hydra.initialize(config_path=\"../conf\", job_name=\"eda\", version_base=None)\n",
            "\n",
            "# Setup root environment\n",
            "root_path = rootutils.setup_root(\".\")\n",
            "rootutils.set_root(\n",
            "    path=root_path,\n",
            "    project_root_env_var=True,\n",
            ")"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 3,
         "metadata": {},
         "outputs": [],
         "source": [
            "# Global paths\n",
            "ROOT_DIR = root_path\n",
            "DATA_DIR = os.path.join(ROOT_DIR, \"data\")\n",
            "ARTIFACT_DIR = os.path.join(ROOT_DIR, \"artifacts\")\n",
            "FIGURE_DIR = os.path.join(ROOT_DIR, \"report\", \"figures\")\n",
            "TABLE_DIR = os.path.join(ROOT_DIR, \"report\", \"tables\")\n",
            "\n",
            "os.makedirs(FIGURE_DIR, exist_ok=True)\n",
            "os.makedirs(TABLE_DIR, exist_ok=True)"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 4,
         "metadata": {},
         "outputs": [],
         "source": [
            "# Costs per token based on https://openai.com/pricing\n",
            "GPT4_COST_PER_INP_TOKEN = 0.00001\n",
            "GPT4_COST_PER_OUT_TOKEN = 0.00003\n",
            "GPT3_5_COST_PER_INP_TOKEN = 0.000001\n",
            "GPT3_5_COST_PER_OUT_TOKEN = 0.000002"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 5,
         "metadata": {},
         "outputs": [],
         "source": [
            "rename_dict = {\n",
            "    \"human\": \"Human\",\n",
            "    \"gpt3.5\": \"GPT-3.5\",\n",
            "    \"gpt4\": \"GPT-4\",\n",
            "    \"context1\": \"C1\",\n",
            "    \"context2\": \"C2\",\n",
            "    \"context3\": \"C3\",\n",
            "    \"zeroshot\": \"0-shot\",\n",
            "    \"oneshot\": \"1-shot\",\n",
            "    \"f1\": \"Macro F1\",\n",
            "    \"acc\": \"Acc\",\n",
            "    \"precision\": \"Precision\",\n",
            "    \"recall\": \"Recall\",\n",
            "    \"lpp\": \"Labels Per Page\",\n",
            "    \"model\": \"Model\",\n",
            "    \"context\": \"Context\",\n",
            "    \"shot\": \"Shot\",\n",
            "    \"gpt-3.5-turbo-1106\": \"GPT-3.5\",\n",
            "    \"gpt-4-1106-preview\": \"GPT-4\",\n",
            "    False: \"0-shot\",\n",
            "    True: \"1-shot\",\n",
            "    \"Kids_and_Teens\": \"Kids & Teens\",\n",
            "}\n",
            "\n",
            "def get_labeler_name(name: str):\n",
            "    return \" + \".join([rename_dict.get(n, n) for n in name.split(\"-\")])\n",
            "\n",
            "def get_metric_name(name: str):\n",
            "    if \"/\" in name:\n",
            "        split, metric = name.split(\"/\")\n",
            "        return f\"{rename_dict.get(split, split)} {rename_dict.get(metric, metric)}\"\n",
            "    else:\n",
            "        return rename_dict.get(name, name)"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 6,
         "metadata": {},
         "outputs": [],
         "source": [
            "style = \"whitegrid\"\n",
            "palette = \"gist_stern\"\n",
            "sns.set_style(style)\n",
            "sns.set_palette(palette)"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 7,
         "metadata": {},
         "outputs": [],
         "source": [
            "def get_num_features(feat, data):\n",
            "    return len([w[feat] for w in data.values() if w[feat] is not None and w[feat] != []])"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "## Websites\n",
            "\n",
            "---\n",
            "\n",
            "There are three copora of websites in this dataset:\n",
            "\n",
            "* `crowdsourced`: 761 websites from the crowdsourced dataset from the [Homepage2Vec paper](https://arxiv.org/abs/1905.09786)\n",
            "* `curlie`: A filtered version of the [curlie](https://curlie.org) dataset, containing ~1M websites\n",
            "\n",
            "For each website, the repository contains a CSV file at the path `data/raw/<corpus>.csv` with the two columns - `wid` and `url`. The `wid` is a unique identifier for the website, and the `url` is the URL of the website."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 8,
         "metadata": {},
         "outputs": [],
         "source": [
            "# Initialise Config\n",
            "crowdsourced_cfg = hydra.compose(config_name=\"eda\", overrides=[\"data=crowdsourced\"])\n",
            "curlie_cfg = hydra.compose(config_name=\"eda\", overrides=[\"data=curlie\"])"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 9,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "Classifying in 14 categories: Arts, Business, Computers, Games, Health, Home, Kids_and_Teens, News, Recreation, Reference, Science, Shopping, Society, Sports\n"
               ]
            }
         ],
         "source": [
            "# Load categories\n",
            "path = os.path.join(DATA_DIR, \"meta\", \"categories.json\")\n",
            "with open(path) as f:\n",
            "    categories = json.load(f)\n",
            "\n",
            "print(f\"Classifying in {len(categories)} categories: {', '.join(categories)}\")"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "We will continue by exploring each of our datasets, we will store info about each of the indivudal datasets in a list of dictionaries."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 10,
         "metadata": {},
         "outputs": [],
         "source": [
            "webinfo = []"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "### Crowdsourced Data\n",
            "\n",
            "This is the annotations crowdsourced for 840 websites in the [Homepage2Vec paper](https://arxiv.org/abs/1905.09786)."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 11,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "Total number of samples: 840\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>wid</th>\n",
                     "      <th>url</th>\n",
                     "    </tr>\n",
                     "  </thead>\n",
                     "  <tbody>\n",
                     "    <tr>\n",
                     "      <th>0</th>\n",
                     "      <td>1161124</td>\n",
                     "      <td>www.pointlesssites.com</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>1</th>\n",
                     "      <td>1081241</td>\n",
                     "      <td>www.connecticutplastics.com</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>2</th>\n",
                     "      <td>1162420</td>\n",
                     "      <td>99percentinvisible.org</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>3</th>\n",
                     "      <td>1146040</td>\n",
                     "      <td>www.medicaid.gov</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>4</th>\n",
                     "      <td>1117243</td>\n",
                     "      <td>www.graalonline.com</td>\n",
                     "    </tr>\n",
                     "  </tbody>\n",
                     "</table>\n",
                     "</div>"
                  ],
                  "text/plain": [
                     "       wid                          url\n",
                     "0  1161124       www.pointlesssites.com\n",
                     "1  1081241  www.connecticutplastics.com\n",
                     "2  1162420       99percentinvisible.org\n",
                     "3  1146040             www.medicaid.gov\n",
                     "4  1117243          www.graalonline.com"
                  ]
               },
               "execution_count": 11,
               "metadata": {},
               "output_type": "execute_result"
            }
         ],
         "source": [
            "# Raw data\n",
            "crowdsourced_data= hydra.utils.instantiate(crowdsourced_cfg.data)\n",
            "\n",
            "raw_data = crowdsourced_data.get_raw_data()\n",
            "processed_data = crowdsourced_data.get_processed_data()\n",
            "embedded_data = crowdsourced_data.get_embeddings()\n",
            "\n",
            "print(f\"Total number of samples: {len(raw_data)}\")\n",
            "raw_data.head(5)"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 12,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "Collected data on ['title', 'description', 'keywords', 'links', 'sentences', 'metatags', 'tld', 'domain']\n",
                  "\n",
                  "Title: PointlessSites.com  Fun Things To Do When You're Bored\n",
                  "Description: Are you bored? Want something fun to do? Check out these funny websites, pointless facts and stupid pictures brought to you by Pointless Sites!\n",
                  "Keywords: ['Pointless', 'Sites', 'portal,', 'useless', 'pointlessness']\n",
                  "Tags: ['description', 'viewport', 'author', 'keywords', 'robots', 'copyright']\n",
                  "Domain: pointlesssites\n",
                  "TLD: com\n"
               ]
            }
         ],
         "source": [
            "# Example of processed website\n",
            "wid = list(processed_data.keys())[0]\n",
            "data = processed_data[wid]\n",
            "\n",
            "print(f\"Collected data on {list(data.keys())}\")\n",
            "\n",
            "# Show some examples\n",
            "print(f\"\\nTitle: {data['title']}\")\n",
            "print(f\"Description: {data['description']}\")\n",
            "print(f\"Keywords: {data['keywords']}\")\n",
            "print(f\"Tags: {data['metatags']}\")\n",
            "print(f\"Domain: {data['domain']}\")\n",
            "print(f\"TLD: {data['tld']}\")"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 13,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "ℹ️ Percentage of sites with tld: 100.00%\n",
                  "ℹ️ Percentage of sites with domain: 100.00%\n",
                  "ℹ️ Percentage of sites with tags: 93.69%\n",
                  "ℹ️ Percentage of sites with titles: 98.42%\n",
                  "ℹ️ Percentage of sites with descriptions: 54.93%\n",
                  "ℹ️ Percentage of sites with keywords: 19.58%\n",
                  "ℹ️ Percentage of sites with links: 89.88%\n",
                  "ℹ️ Percentage of sites with sentences: 99.08%\n"
               ]
            }
         ],
         "source": [
            "# Setup info dict for original\n",
            "original_info = dict()\n",
            "\n",
            "# Save these into a dict\n",
            "original_info[\"n\"] = len(processed_data)\n",
            "original_info[\"tld\"] = get_num_features(\"tld\", processed_data) / original_info[\"n\"] * 100\n",
            "original_info[\"domain\"] = get_num_features(\"domain\", processed_data) / original_info[\"n\"] * 100\n",
            "original_info[\"tags\"] = get_num_features(\"metatags\", processed_data) / original_info[\"n\"] * 100\n",
            "original_info[\"titles\"] = get_num_features(\"title\", processed_data) / original_info[\"n\"] * 100\n",
            "original_info[\"descriptions\"] = get_num_features(\"description\", processed_data) / original_info[\"n\"] * 100\n",
            "original_info[\"keywords\"] = get_num_features(\"keywords\", processed_data) / original_info[\"n\"] * 100\n",
            "original_info[\"links\"] = get_num_features(\"links\", processed_data) / original_info[\"n\"] * 100\n",
            "original_info[\"sentences\"] = get_num_features(\"sentences\", processed_data) / original_info[\"n\"] * 100\n",
            "\n",
            "for k, v in original_info.items():\n",
            "    if k != \"n\":\n",
            "        print(f\"ℹ️ Percentage of sites with {k}: {v:.2f}%\")\n"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "### Curlie\n",
            "\n",
            "This is the random subset of 10,000 websites from the Curlie website index that we wish to re-annotate using GPT labelers."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 14,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "Total number of samples: 10000\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>wid</th>\n",
                     "      <th>url</th>\n",
                     "    </tr>\n",
                     "  </thead>\n",
                     "  <tbody>\n",
                     "    <tr>\n",
                     "      <th>0</th>\n",
                     "      <td>917678</td>\n",
                     "      <td>www.winandsoft.fr</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>1</th>\n",
                     "      <td>443072</td>\n",
                     "      <td>gaude-ag.de</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>2</th>\n",
                     "      <td>728091</td>\n",
                     "      <td>www.housing.ucsb.edu</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>3</th>\n",
                     "      <td>132596</td>\n",
                     "      <td>www.daccad.nl</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>4</th>\n",
                     "      <td>464355</td>\n",
                     "      <td>www.rockhall.com</td>\n",
                     "    </tr>\n",
                     "  </tbody>\n",
                     "</table>\n",
                     "</div>"
                  ],
                  "text/plain": [
                     "      wid                   url\n",
                     "0  917678     www.winandsoft.fr\n",
                     "1  443072           gaude-ag.de\n",
                     "2  728091  www.housing.ucsb.edu\n",
                     "3  132596         www.daccad.nl\n",
                     "4  464355      www.rockhall.com"
                  ]
               },
               "execution_count": 14,
               "metadata": {},
               "output_type": "execute_result"
            }
         ],
         "source": [
            "# Raw data\n",
            "curlie_data = hydra.utils.instantiate(curlie_cfg.data)\n",
            "\n",
            "raw_data = curlie_data.get_raw_data()\n",
            "processed_data = curlie_data.get_processed_data()\n",
            "embedded_jdata = curlie_data.get_embeddings()\n",
            "\n",
            "print(f\"Total number of samples: {len(raw_data)}\")\n",
            "raw_data.head(5)"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 15,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "Collected data on ['title', 'description', 'keywords', 'links', 'sentences', 'metatags', 'tld', 'domain']\n",
                  "\n",
                  "Title: Logiciel bibliothèque, médiathèque, vidéothèque.\n",
                  "Description: Logiciel bibliothèque  logiciel permettant le catalogage de tous types de documents ou supports pour les besoins d'une bibliothèque. Le logiciel bibliothèque vous garantit la gestion documentaire la plus efficace (livres, cartes, manuscrits, incunables, etc.).\n",
                  "Keywords: ['logiciel', 'bibliotheque,logiciel', 'mediatheque,logiciel', 'gestion', 'bibliotheque,logiciel', 'gestion', 'mediatheque,logiciel,bibliotheque,mediatheque']\n",
                  "Tags: ['description', 'copyright', 'robots', 'keywords', 'author']\n",
                  "Domain: winandsoft\n",
                  "TLD: fr\n"
               ]
            }
         ],
         "source": [
            "# Example of processed website\n",
            "wid = list(processed_data.keys())[0]\n",
            "data = processed_data[wid]\n",
            "\n",
            "print(f\"Collected data on {list(data.keys())}\")\n",
            "\n",
            "# Show some examples\n",
            "print(f\"\\nTitle: {data['title']}\")\n",
            "print(f\"Description: {data['description']}\")\n",
            "print(f\"Keywords: {data['keywords']}\")\n",
            "print(f\"Tags: {data['metatags']}\")\n",
            "print(f\"Domain: {data['domain']}\")\n",
            "print(f\"TLD: {data['tld']}\")"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 16,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "ℹ️ Percentage of sites with tld: 100.00%\n",
                  "ℹ️ Percentage of sites with domain: 100.00%\n",
                  "ℹ️ Percentage of sites with tags: 95.47%\n",
                  "ℹ️ Percentage of sites with titles: 98.28%\n",
                  "ℹ️ Percentage of sites with descriptions: 62.95%\n",
                  "ℹ️ Percentage of sites with keywords: 27.29%\n",
                  "ℹ️ Percentage of sites with links: 91.62%\n",
                  "ℹ️ Percentage of sites with sentences: 99.03%\n"
               ]
            }
         ],
         "source": [
            "curlie_info = dict()\n",
            "\n",
            "# Save these into a dict\n",
            "curlie_info[\"n\"] = len(processed_data)\n",
            "curlie_info[\"tld\"] = get_num_features(\"tld\", processed_data) / curlie_info[\"n\"] * 100\n",
            "curlie_info[\"domain\"] = get_num_features(\"domain\", processed_data) / curlie_info[\"n\"] * 100\n",
            "curlie_info[\"tags\"] = get_num_features(\"metatags\", processed_data) / curlie_info[\"n\"] * 100\n",
            "curlie_info[\"titles\"] = get_num_features(\"title\", processed_data) / curlie_info[\"n\"] * 100\n",
            "curlie_info[\"descriptions\"] = get_num_features(\"description\", processed_data) / curlie_info[\"n\"] * 100\n",
            "curlie_info[\"keywords\"] = get_num_features(\"keywords\", processed_data) / curlie_info[\"n\"] * 100\n",
            "curlie_info[\"links\"] = get_num_features(\"links\", processed_data) / curlie_info[\"n\"] * 100\n",
            "curlie_info[\"sentences\"] = get_num_features(\"sentences\", processed_data) / curlie_info[\"n\"] * 100\n",
            "\n",
            "for k, v in curlie_info.items():\n",
            "    if k != \"n\":\n",
            "        print(f\"ℹ️ Percentage of sites with {k}: {v:.2f}%\")"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "### LaTeX Table"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 17,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "✅ Saved table to /Users/jonas-mika/epfl/coursework/projects/ml-project-2-mlp/report/tables/feature-info.tex\n"
               ]
            }
         ],
         "source": [
            "# Put the data info into a dataframe\n",
            "df = pd.DataFrame([original_info, curlie_info], index=[\"Crowdsourced\", \"Curlie-10k\"]).round(2)\n",
            "\n",
            "# Change all columns names to include (%)\n",
            "df.columns = [f\"{c} (%)\" if c != \"n\" else c for c in df.columns]\n",
            "\n",
            "df = df.T\n",
            "\n",
            "# Save the dataframe to a latex table\n",
            "position = \"!ht\"\n",
            "latex = df.to_latex(\n",
            "    caption=\"Percentage of websites with each feature accross our datasets.\", \n",
            "    label=\"tab:feature-info\",\n",
            "    escape=True,\n",
            "    float_format=\"%.2f\",\n",
            "    position=position\n",
            ")\n",
            "\n",
            "# Add \\centering right after \\begin{table}\n",
            "latex = latex.replace(\"\\\\begin{table}\" + f\"[{position}]\", \"\\\\begin{table}\" + f\"[{position}]\" + \"\\n\\\\centering\")\n",
            "\n",
            "# Save the latex table\n",
            "save_path = os.path.join(root_path, \"report\", \"tables\", \"feature-info.tex\")\n",
            "with open(save_path, \"w\") as f:\n",
            "    f.write(latex)\n",
            "print(f\"✅ Saved table to {save_path}\")"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "## Labelers\n",
            "\n",
            "---\n",
            "\n",
            "There are multiple GPT labeler instances that can be used to label the data. The labelers are defined in the labelers module and are identified by a `context`, `model` and `fewshow` parameter. The `context` parameter defines the context that is used to prompt the model. The `model` parameter defines the model that is used to generate the labels. The `fewshot` parameter defines whether the model is trained in a few-shot manner or not.\n",
            "\n",
            "| Parameter | Variants | Description |\n",
            "| --- | --- | ---  |\n",
            "| `context` | `context1` | Uses the `tld`, `domain` and `metatags` as context |\n",
            "| | `context2` | Uses the `tld`, `domain`, `metatags`, `links` and `text` as context |\n",
            "| | `context3` | Uses the `tld`, `domain`, `metatags`, `links`, `text` and `images` as context |\n",
            "| `model` | `gpt3.5` | Uses GPT-3.5 (`gpt-3.5-turbo-1106`) |\n",
            "| | `gpt4` | Uses GPT-4 (`gpt-4-1106-preview`) |\n",
            "| `fewshot` | `fewshot` | Injects an example website and label into the system prompt |\n",
            "| | `zeroshot` | Does not inject any example website or label into the system prompt |\n",
            "\n",
            "We are considering all contexts and few-shot settings for the GPT-3.5 model, but only the `context2` for the GPT-4 model. Additionally, we have the `human` labeler which simply loads the annotatations for the `original` dataset from the [Homepage2Vec paper](https://arxiv.org/abs/1905.09786)."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 18,
         "metadata": {},
         "outputs": [],
         "source": [
            "# Initialise configuration for all labelers\n",
            "gpt_labeler_names = [\n",
            "    \"gpt3.5-zeroshot-context1\", \n",
            "    \"gpt3.5-oneshot-context1\",\n",
            "    \"gpt3.5-zeroshot-context2\",\n",
            "    \"gpt3.5-oneshot-context2\",\n",
            "    \"gpt3.5-zeroshot-context3\",\n",
            "    \"gpt3.5-oneshot-context3\",\n",
            "    \"gpt4-zeroshot-context1\",\n",
            "    \"gpt4-oneshot-context1\",\n",
            "    \"gpt4-zeroshot-context2\",\n",
            "    \"gpt4-oneshot-context2\",\n",
            "    \"gpt4-zeroshot-context3\",\n",
            "    \"gpt4-oneshot-context3\"\n",
            "] \n",
            "\n",
            "gpt_labelers_cfg = {labeler: hydra.compose(config_name=\"eda\", overrides=[f\"labeler={labeler}\"]) for labeler in gpt_labeler_names}"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 19,
         "metadata": {},
         "outputs": [],
         "source": [
            "# Instantiate data\n",
            "crowdsourced_data_cfg = hydra.compose(config_name=\"eda\", overrides=[\"data=crowdsourced\"])\n",
            "crowdsourced_data = hydra.utils.instantiate(crowdsourced_data_cfg.data)"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 20,
         "metadata": {},
         "outputs": [],
         "source": [
            "# Instantiate labelers\n",
            "gpt_labelers = {labeler: hydra.utils.instantiate(cfg.labeler, data=crowdsourced_data) for labeler, cfg in gpt_labelers_cfg.items()}"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 21,
         "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>model</th>\n",
                     "      <th>context</th>\n",
                     "      <th>shot</th>\n",
                     "    </tr>\n",
                     "  </thead>\n",
                     "  <tbody>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 0-shot + C1</th>\n",
                     "      <td>gpt-3.5-turbo-1106</td>\n",
                     "      <td>context1</td>\n",
                     "      <td>False</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C1</th>\n",
                     "      <td>gpt-3.5-turbo-1106</td>\n",
                     "      <td>context1</td>\n",
                     "      <td>True</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 0-shot + C2</th>\n",
                     "      <td>gpt-3.5-turbo-1106</td>\n",
                     "      <td>context2</td>\n",
                     "      <td>False</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C2</th>\n",
                     "      <td>gpt-3.5-turbo-1106</td>\n",
                     "      <td>context2</td>\n",
                     "      <td>True</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 0-shot + C3</th>\n",
                     "      <td>gpt-3.5-turbo-1106</td>\n",
                     "      <td>context3</td>\n",
                     "      <td>False</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C3</th>\n",
                     "      <td>gpt-3.5-turbo-1106</td>\n",
                     "      <td>context3</td>\n",
                     "      <td>True</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C1</th>\n",
                     "      <td>gpt-4-1106-preview</td>\n",
                     "      <td>context1</td>\n",
                     "      <td>False</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C1</th>\n",
                     "      <td>gpt-4-1106-preview</td>\n",
                     "      <td>context1</td>\n",
                     "      <td>True</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C2</th>\n",
                     "      <td>gpt-4-1106-preview</td>\n",
                     "      <td>context2</td>\n",
                     "      <td>False</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C2</th>\n",
                     "      <td>gpt-4-1106-preview</td>\n",
                     "      <td>context2</td>\n",
                     "      <td>True</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C3</th>\n",
                     "      <td>gpt-4-1106-preview</td>\n",
                     "      <td>context3</td>\n",
                     "      <td>False</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C3</th>\n",
                     "      <td>gpt-4-1106-preview</td>\n",
                     "      <td>context3</td>\n",
                     "      <td>True</td>\n",
                     "    </tr>\n",
                     "  </tbody>\n",
                     "</table>\n",
                     "</div>"
                  ],
                  "text/plain": [
                     "                                    model   context   shot\n",
                     "GPT-3.5 + 0-shot + C1  gpt-3.5-turbo-1106  context1  False\n",
                     "GPT-3.5 + 1-shot + C1  gpt-3.5-turbo-1106  context1   True\n",
                     "GPT-3.5 + 0-shot + C2  gpt-3.5-turbo-1106  context2  False\n",
                     "GPT-3.5 + 1-shot + C2  gpt-3.5-turbo-1106  context2   True\n",
                     "GPT-3.5 + 0-shot + C3  gpt-3.5-turbo-1106  context3  False\n",
                     "GPT-3.5 + 1-shot + C3  gpt-3.5-turbo-1106  context3   True\n",
                     "GPT-4 + 0-shot + C1    gpt-4-1106-preview  context1  False\n",
                     "GPT-4 + 1-shot + C1    gpt-4-1106-preview  context1   True\n",
                     "GPT-4 + 0-shot + C2    gpt-4-1106-preview  context2  False\n",
                     "GPT-4 + 1-shot + C2    gpt-4-1106-preview  context2   True\n",
                     "GPT-4 + 0-shot + C3    gpt-4-1106-preview  context3  False\n",
                     "GPT-4 + 1-shot + C3    gpt-4-1106-preview  context3   True"
                  ]
               },
               "execution_count": 21,
               "metadata": {},
               "output_type": "execute_result"
            }
         ],
         "source": [
            "# Instatntiate labelers dataframe\n",
            "def get_info(labeler):\n",
            "    return {\n",
            "        \"model\": labeler.model,\n",
            "        \"context\": labeler.name.split(\"-\")[-1],\n",
            "        \"shot\": labeler.fewshot,\n",
            "    }\n",
            "\n",
            "labeler_info = pd.DataFrame(\n",
            "    [get_info(labeler) for labeler in gpt_labelers.values()],\n",
            "    index=[get_labeler_name(name) for name in gpt_labelers.keys()])\n",
            "labeler_info"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "Let's verify that the labelers are working as expected by checking the number of labeled webpages."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 22,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "ℹ️ Number of processed websites: 761\n",
                  "ℹ️ Number of gpt3.5-zeroshot-context1 labels: 761\n",
                  "ℹ️ Number of gpt3.5-oneshot-context1 labels: 761\n",
                  "ℹ️ Number of gpt3.5-zeroshot-context2 labels: 761\n",
                  "ℹ️ Number of gpt3.5-oneshot-context2 labels: 761\n",
                  "ℹ️ Number of gpt3.5-zeroshot-context3 labels: 761\n",
                  "ℹ️ Number of gpt3.5-oneshot-context3 labels: 761\n",
                  "ℹ️ Number of gpt4-zeroshot-context1 labels: 761\n",
                  "ℹ️ Number of gpt4-oneshot-context1 labels: 761\n",
                  "ℹ️ Number of gpt4-zeroshot-context2 labels: 761\n",
                  "ℹ️ Number of gpt4-oneshot-context2 labels: 761\n",
                  "ℹ️ Number of gpt4-zeroshot-context3 labels: 761\n",
                  "ℹ️ Number of gpt4-oneshot-context3 labels: 761\n"
               ]
            }
         ],
         "source": [
            "num_processed_websites = len(crowdsourced_data.get_processed_data())\n",
            "\n",
            "print(f\"ℹ️ Number of processed websites: {num_processed_websites}\")\n",
            "for name, labeler in gpt_labelers.items():\n",
            "    num_labels = len(labeler.get_labels())\n",
            "    print(f\"ℹ️ Number of {name} labels: {num_labels}\")"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "### Labeling statistics\n",
            "\n",
            "Let's investigate some statistics about the labelers. We will compute:\n",
            "\n",
            "* The average number of labels per website\n",
            "* The number of valid labels\n",
            "* The number of invalid labels\n",
            "* The average time taken to label a website\n",
            "* The average number of prompt and completion tokens used to label a website\n",
            "* The estimated cost of labeling the entire dataset"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 23,
         "metadata": {},
         "outputs": [],
         "source": [
            "def get_statistics(labeler):\n",
            "    statistics = {\"valid\": 0, \"invalid\": 0, \"lpp\": [], \"durations\": [], \"prompt_tokens\": [], \"completion_tokens\": []}\n",
            "    for website in labeler.get_labels().values():\n",
            "        if not website[\"is_valid\"]:\n",
            "            statistics[\"invalid\"] += 1\n",
            "            continue\n",
            "\n",
            "        statistics[\"valid\"] += 1\n",
            "        statistics[\"lpp\"].append(sum(website[\"labels\"]))\n",
            "        statistics[\"durations\"].append(website[\"duration\"])\n",
            "        statistics[\"prompt_tokens\"].append(website[\"prompt_tokens\"])\n",
            "        statistics[\"completion_tokens\"].append(website[\"completion_tokens\"])\n",
            "\n",
            "    lpps = np.array(statistics[\"lpp\"])\n",
            "    durations = np.array(statistics[\"durations\"])\n",
            "    prompt_tokens = np.array(statistics[\"prompt_tokens\"])\n",
            "    completion_tokens = np.array(statistics[\"completion_tokens\"])\n",
            "\n",
            "    statistics[\"lpp\"] = f\"{lpps.mean():.2f} ± {lpps.std():.2f}\"\n",
            "    statistics[\"durations\"] = f\"{durations.mean():.2f} ± {durations.std():.2f}\"\n",
            "    statistics[\"prompt_tokens\"] = f\"{prompt_tokens.mean():.2f} ± {prompt_tokens.std():.2f}\"\n",
            "    statistics[\"completion_tokens\"] = f\"{completion_tokens.mean():.2f} ± {completion_tokens.std():.2f}\"\n",
            "    \n",
            "    # Compute estimated cost\n",
            "    model = labeler.model\n",
            "    COST_PER_INP_TOKEN = GPT4_COST_PER_INP_TOKEN if \"gpt-4\" in model else GPT3_5_COST_PER_INP_TOKEN\n",
            "    COST_PER_OUT_TOKEN = GPT4_COST_PER_OUT_TOKEN if \"gpt-4\" in model else GPT3_5_COST_PER_OUT_TOKEN\n",
            "\n",
            "    statistics[\"estimated_cost\"] = (prompt_tokens.sum() * COST_PER_INP_TOKEN + completion_tokens.sum() * COST_PER_OUT_TOKEN)\n",
            "    statistics[\"cost_per_1k_page\"] = 1000 * statistics[\"estimated_cost\"] / statistics[\"valid\"]\n",
            "\n",
            "    return statistics"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 24,
         "metadata": {},
         "outputs": [
            {
               "data": {
                  "text/html": [
                     "<div>\n",
                     "<style scoped>\n",
                     "    .dataframe tbody tr th:only-of-type {\n",
                     "        vertical-align: middle;\n",
                     "    }\n",
                     "\n",
                     "    .dataframe tbody tr th {\n",
                     "        vertical-align: top;\n",
                     "    }\n",
                     "\n",
                     "    .dataframe thead th {\n",
                     "        text-align: right;\n",
                     "    }\n",
                     "</style>\n",
                     "<table border=\"1\" class=\"dataframe\">\n",
                     "  <thead>\n",
                     "    <tr style=\"text-align: right;\">\n",
                     "      <th></th>\n",
                     "      <th>valid</th>\n",
                     "      <th>invalid</th>\n",
                     "      <th>lpp</th>\n",
                     "      <th>durations</th>\n",
                     "      <th>prompt_tokens</th>\n",
                     "      <th>completion_tokens</th>\n",
                     "      <th>estimated_cost</th>\n",
                     "      <th>cost_per_1k_page</th>\n",
                     "    </tr>\n",
                     "  </thead>\n",
                     "  <tbody>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 0-shot + C1</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>0.39 ± 0.61</td>\n",
                     "      <td>3.51 ± 5.87</td>\n",
                     "      <td>148.80 ± 18.14</td>\n",
                     "      <td>107.54 ± 2.75</td>\n",
                     "      <td>0.276907</td>\n",
                     "      <td>0.363873</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C1</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>0.91 ± 0.95</td>\n",
                     "      <td>2.99 ± 4.61</td>\n",
                     "      <td>292.80 ± 18.14</td>\n",
                     "      <td>92.71 ± 3.31</td>\n",
                     "      <td>0.363929</td>\n",
                     "      <td>0.478225</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 0-shot + C2</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>1.39 ± 0.98</td>\n",
                     "      <td>3.63 ± 6.79</td>\n",
                     "      <td>206.30 ± 60.33</td>\n",
                     "      <td>107.90 ± 1.48</td>\n",
                     "      <td>0.321219</td>\n",
                     "      <td>0.422101</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C2</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>1.68 ± 1.15</td>\n",
                     "      <td>2.85 ± 0.42</td>\n",
                     "      <td>448.30 ± 60.33</td>\n",
                     "      <td>93.11 ± 4.06</td>\n",
                     "      <td>0.482875</td>\n",
                     "      <td>0.634527</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 0-shot + C3</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>1.57 ± 1.08</td>\n",
                     "      <td>3.27 ± 0.69</td>\n",
                     "      <td>350.13 ± 124.30</td>\n",
                     "      <td>108.00 ± 0.06</td>\n",
                     "      <td>0.430822</td>\n",
                     "      <td>0.566126</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C3</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>1.85 ± 1.24</td>\n",
                     "      <td>3.07 ± 4.31</td>\n",
                     "      <td>615.13 ± 124.30</td>\n",
                     "      <td>94.92 ± 6.18</td>\n",
                     "      <td>0.612587</td>\n",
                     "      <td>0.804976</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C1</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>1.50 ± 0.93</td>\n",
                     "      <td>9.15 ± 3.44</td>\n",
                     "      <td>148.80 ± 18.14</td>\n",
                     "      <td>106.29 ± 8.45</td>\n",
                     "      <td>3.558880</td>\n",
                     "      <td>4.676583</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C1</th>\n",
                     "      <td>760</td>\n",
                     "      <td>1</td>\n",
                     "      <td>1.83 ± 1.36</td>\n",
                     "      <td>7.92 ± 2.73</td>\n",
                     "      <td>292.60 ± 17.34</td>\n",
                     "      <td>94.01 ± 10.96</td>\n",
                     "      <td>4.367230</td>\n",
                     "      <td>5.746355</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C2</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>2.16 ± 1.03</td>\n",
                     "      <td>5.73 ± 1.98</td>\n",
                     "      <td>206.30 ± 60.33</td>\n",
                     "      <td>106.61 ± 7.14</td>\n",
                     "      <td>4.003900</td>\n",
                     "      <td>5.261367</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C2</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>2.49 ± 1.28</td>\n",
                     "      <td>5.16 ± 2.35</td>\n",
                     "      <td>448.30 ± 60.33</td>\n",
                     "      <td>92.29 ± 2.52</td>\n",
                     "      <td>5.518490</td>\n",
                     "      <td>7.251629</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C3</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>2.30 ± 1.11</td>\n",
                     "      <td>9.44 ± 3.09</td>\n",
                     "      <td>350.13 ± 124.30</td>\n",
                     "      <td>105.80 ± 8.67</td>\n",
                     "      <td>5.079990</td>\n",
                     "      <td>6.675414</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C3</th>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>2.80 ± 1.30</td>\n",
                     "      <td>7.93 ± 2.70</td>\n",
                     "      <td>615.13 ± 124.30</td>\n",
                     "      <td>94.68 ± 6.23</td>\n",
                     "      <td>6.842570</td>\n",
                     "      <td>8.991551</td>\n",
                     "    </tr>\n",
                     "  </tbody>\n",
                     "</table>\n",
                     "</div>"
                  ],
                  "text/plain": [
                     "                       valid  invalid          lpp    durations  \\\n",
                     "GPT-3.5 + 0-shot + C1    761        0  0.39 ± 0.61  3.51 ± 5.87   \n",
                     "GPT-3.5 + 1-shot + C1    761        0  0.91 ± 0.95  2.99 ± 4.61   \n",
                     "GPT-3.5 + 0-shot + C2    761        0  1.39 ± 0.98  3.63 ± 6.79   \n",
                     "GPT-3.5 + 1-shot + C2    761        0  1.68 ± 1.15  2.85 ± 0.42   \n",
                     "GPT-3.5 + 0-shot + C3    761        0  1.57 ± 1.08  3.27 ± 0.69   \n",
                     "GPT-3.5 + 1-shot + C3    761        0  1.85 ± 1.24  3.07 ± 4.31   \n",
                     "GPT-4 + 0-shot + C1      761        0  1.50 ± 0.93  9.15 ± 3.44   \n",
                     "GPT-4 + 1-shot + C1      760        1  1.83 ± 1.36  7.92 ± 2.73   \n",
                     "GPT-4 + 0-shot + C2      761        0  2.16 ± 1.03  5.73 ± 1.98   \n",
                     "GPT-4 + 1-shot + C2      761        0  2.49 ± 1.28  5.16 ± 2.35   \n",
                     "GPT-4 + 0-shot + C3      761        0  2.30 ± 1.11  9.44 ± 3.09   \n",
                     "GPT-4 + 1-shot + C3      761        0  2.80 ± 1.30  7.93 ± 2.70   \n",
                     "\n",
                     "                         prompt_tokens completion_tokens  estimated_cost  \\\n",
                     "GPT-3.5 + 0-shot + C1   148.80 ± 18.14     107.54 ± 2.75        0.276907   \n",
                     "GPT-3.5 + 1-shot + C1   292.80 ± 18.14      92.71 ± 3.31        0.363929   \n",
                     "GPT-3.5 + 0-shot + C2   206.30 ± 60.33     107.90 ± 1.48        0.321219   \n",
                     "GPT-3.5 + 1-shot + C2   448.30 ± 60.33      93.11 ± 4.06        0.482875   \n",
                     "GPT-3.5 + 0-shot + C3  350.13 ± 124.30     108.00 ± 0.06        0.430822   \n",
                     "GPT-3.5 + 1-shot + C3  615.13 ± 124.30      94.92 ± 6.18        0.612587   \n",
                     "GPT-4 + 0-shot + C1     148.80 ± 18.14     106.29 ± 8.45        3.558880   \n",
                     "GPT-4 + 1-shot + C1     292.60 ± 17.34     94.01 ± 10.96        4.367230   \n",
                     "GPT-4 + 0-shot + C2     206.30 ± 60.33     106.61 ± 7.14        4.003900   \n",
                     "GPT-4 + 1-shot + C2     448.30 ± 60.33      92.29 ± 2.52        5.518490   \n",
                     "GPT-4 + 0-shot + C3    350.13 ± 124.30     105.80 ± 8.67        5.079990   \n",
                     "GPT-4 + 1-shot + C3    615.13 ± 124.30      94.68 ± 6.23        6.842570   \n",
                     "\n",
                     "                       cost_per_1k_page  \n",
                     "GPT-3.5 + 0-shot + C1          0.363873  \n",
                     "GPT-3.5 + 1-shot + C1          0.478225  \n",
                     "GPT-3.5 + 0-shot + C2          0.422101  \n",
                     "GPT-3.5 + 1-shot + C2          0.634527  \n",
                     "GPT-3.5 + 0-shot + C3          0.566126  \n",
                     "GPT-3.5 + 1-shot + C3          0.804976  \n",
                     "GPT-4 + 0-shot + C1            4.676583  \n",
                     "GPT-4 + 1-shot + C1            5.746355  \n",
                     "GPT-4 + 0-shot + C2            5.261367  \n",
                     "GPT-4 + 1-shot + C2            7.251629  \n",
                     "GPT-4 + 0-shot + C3            6.675414  \n",
                     "GPT-4 + 1-shot + C3            8.991551  "
                  ]
               },
               "execution_count": 24,
               "metadata": {},
               "output_type": "execute_result"
            }
         ],
         "source": [
            "labeler_statistics = pd.DataFrame(\n",
            "    [get_statistics(labeler) for labeler in gpt_labelers.values()], \n",
            "    index=[get_labeler_name(name) for name in gpt_labelers.keys()]) \n",
            "labeler_statistics"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "### Class Distribution"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 25,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "✅ Plotted figure to /Users/jonas-mika/epfl/coursework/projects/ml-project-2-mlp/data/meta/categories.json\n"
               ]
            },
            {
               "data": {
                  "image/png": "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",
                  "text/plain": [
                     "<Figure size 1500x600 with 1 Axes>"
                  ]
               },
               "metadata": {},
               "output_type": "display_data"
            }
         ],
         "source": [
            "def get_label_dist(labelers):\n",
            "    rows = []\n",
            "    for name, class_dist in {name: labeler.get_class_dist(normalise=True) for name, labeler in labelers.items()}.items():\n",
            "        for category, freq in class_dist.items():\n",
            "            rows.append({\"labeler\": get_labeler_name(name), \"category\": category, \"freq\": 100 * freq})\n",
            "    return pd.DataFrame(rows)\n",
            "\n",
            "label_dist = get_label_dist(gpt_labelers)\n",
            "fig, ax = plt.subplots(figsize=(15, 6))\n",
            "fig.tight_layout(pad=3.0)\n",
            "sns.barplot(\n",
            "    label_dist,\n",
            "    x=\"category\",\n",
            "    y=\"freq\",\n",
            "    hue=\"labeler\",\n",
            "    palette=palette,\n",
            "    order=label_dist.groupby(\"category\").sum().sort_values(\"freq\", ascending=False).index,\n",
            "    ax=ax\n",
            ")\n",
            "ax.get_legend().set_title(\"Labeler\")\n",
            "ax.set_xlabel(\"Category\", fontsize=14)\n",
            "ax.set_ylabel(\"Frequency (%)\", fontsize=14)\n",
            "ax.set_title(\"Original Label Distribution\", fontsize=14)\n",
            "\n",
            "print(f\"✅ Plotted figure to {path}\")"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "### Labeling Quality: Macro F1, Precision, Recall, etc.\n",
            "\n",
            "The goal of all GPT labelers is to replicate the ground truth labels provide by the human annotators as closely as possible. As we only have human annotations for the crowdsourcede dataset, we can only evaluate the labelers on this dataset."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 26,
         "metadata": {},
         "outputs": [],
         "source": [
            "# Initialise crowdsourced config and labeler\n",
            "crowdsourced_data_cfg = hydra.compose(config_name=\"eda\", overrides=[\"data=crowdsourced\"])\n",
            "human_labeler_cfg = hydra.compose(config_name=\"eda\", overrides=[\"labeler=human\", \"data=crowdsourced\"])\n",
            "\n",
            "# Instantitate\n",
            "crowdsourced_data = hydra.utils.instantiate(crowdsourced_data_cfg.data)\n",
            "human_labeler = hydra.utils.instantiate(human_labeler_cfg.labeler, data=crowdsourced_data)"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 27,
         "metadata": {},
         "outputs": [],
         "source": [
            "def match_labels(labeler1, labeler2, subset = None):\n",
            "    labels1 = labeler1.get_labels()\n",
            "    labels2 = labeler2.get_labels()\n",
            "    wid1 = set(labels1.keys())\n",
            "    wid2 = set(labels2.keys())\n",
            "    matched_wid = wid1 & wid2\n",
            "    if subset:\n",
            "        matched_wid = matched_wid & subset\n",
            "\n",
            "    labels1 = np.array([labels1[wid][\"labels\"] for wid in matched_wid])\n",
            "    labels2 = np.array([labels2[wid][\"labels\"] for wid in matched_wid])\n",
            "\n",
            "    return labels1, labels2"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 28,
         "metadata": {},
         "outputs": [],
         "source": [
            "def get_labeler_perf(labeler):\n",
            "    labels1, labels2 = match_labels(human_labeler, labeler)\n",
            "\n",
            "    acc = accuracy_score(labels1.flatten(), labels2.flatten())\n",
            "    subset_acc = accuracy_score(labels1, labels2)\n",
            "    macro_f1 = f1_score(labels1, labels2, average=\"macro\")\n",
            "    micro_f1 = f1_score(labels1, labels2, average=\"micro\")\n",
            "    weighted_f1 = f1_score(labels1, labels2, average=\"weighted\")\n",
            "    macro_precision = precision_score(labels1, labels2, average=\"macro\", zero_division=0)\n",
            "    micro_precision = precision_score(labels1, labels2, average=\"micro\", zero_division=0)\n",
            "    macro_recall = recall_score(labels1, labels2, average=\"macro\", zero_division=0)\n",
            "    micro_recall = recall_score(labels1, labels2, average=\"micro\", zero_division=0)\n",
            "\n",
            "    return {\n",
            "        \"acc\": acc,\n",
            "        \"subset_acc\": subset_acc,\n",
            "        \"macro_f1\": macro_f1,\n",
            "        \"micro_f1\": micro_f1,\n",
            "        \"weighted_f1\": weighted_f1,\n",
            "        \"macro_precision\": macro_precision,\n",
            "        \"micro_precision\": micro_precision,\n",
            "        \"macro_recall\": macro_recall,\n",
            "        \"micro_recall\": micro_recall,\n",
            "    }"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 29,
         "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>acc</th>\n",
                     "      <th>subset_acc</th>\n",
                     "      <th>macro_f1</th>\n",
                     "      <th>micro_f1</th>\n",
                     "      <th>weighted_f1</th>\n",
                     "      <th>macro_precision</th>\n",
                     "      <th>micro_precision</th>\n",
                     "      <th>macro_recall</th>\n",
                     "      <th>micro_recall</th>\n",
                     "    </tr>\n",
                     "  </thead>\n",
                     "  <tbody>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 0-shot + C1</th>\n",
                     "      <td>0.827013</td>\n",
                     "      <td>0.085414</td>\n",
                     "      <td>0.159570</td>\n",
                     "      <td>0.162653</td>\n",
                     "      <td>0.155489</td>\n",
                     "      <td>0.635233</td>\n",
                     "      <td>0.606780</td>\n",
                     "      <td>0.097283</td>\n",
                     "      <td>0.093914</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C1</th>\n",
                     "      <td>0.826826</td>\n",
                     "      <td>0.091984</td>\n",
                     "      <td>0.232630</td>\n",
                     "      <td>0.289017</td>\n",
                     "      <td>0.277751</td>\n",
                     "      <td>0.565566</td>\n",
                     "      <td>0.544267</td>\n",
                     "      <td>0.161507</td>\n",
                     "      <td>0.196747</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 0-shot + C2</th>\n",
                     "      <td>0.836306</td>\n",
                     "      <td>0.139290</td>\n",
                     "      <td>0.375855</td>\n",
                     "      <td>0.412003</td>\n",
                     "      <td>0.411554</td>\n",
                     "      <td>0.562262</td>\n",
                     "      <td>0.576415</td>\n",
                     "      <td>0.297584</td>\n",
                     "      <td>0.320567</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C2</th>\n",
                     "      <td>0.832551</td>\n",
                     "      <td>0.127464</td>\n",
                     "      <td>0.386863</td>\n",
                     "      <td>0.440402</td>\n",
                     "      <td>0.440510</td>\n",
                     "      <td>0.526652</td>\n",
                     "      <td>0.547582</td>\n",
                     "      <td>0.320552</td>\n",
                     "      <td>0.368311</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 0-shot + C3</th>\n",
                     "      <td>0.828046</td>\n",
                     "      <td>0.128778</td>\n",
                     "      <td>0.372394</td>\n",
                     "      <td>0.409794</td>\n",
                     "      <td>0.410367</td>\n",
                     "      <td>0.499444</td>\n",
                     "      <td>0.530885</td>\n",
                     "      <td>0.314836</td>\n",
                     "      <td>0.333683</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C3</th>\n",
                     "      <td>0.821476</td>\n",
                     "      <td>0.115637</td>\n",
                     "      <td>0.377024</td>\n",
                     "      <td>0.426071</td>\n",
                     "      <td>0.428640</td>\n",
                     "      <td>0.464730</td>\n",
                     "      <td>0.501420</td>\n",
                     "      <td>0.328673</td>\n",
                     "      <td>0.370409</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C1</th>\n",
                     "      <td>0.820537</td>\n",
                     "      <td>0.095926</td>\n",
                     "      <td>0.355475</td>\n",
                     "      <td>0.373115</td>\n",
                     "      <td>0.374361</td>\n",
                     "      <td>0.474167</td>\n",
                     "      <td>0.497378</td>\n",
                     "      <td>0.295117</td>\n",
                     "      <td>0.298531</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C1</th>\n",
                     "      <td>0.808804</td>\n",
                     "      <td>0.069645</td>\n",
                     "      <td>0.361039</td>\n",
                     "      <td>0.381791</td>\n",
                     "      <td>0.385413</td>\n",
                     "      <td>0.452552</td>\n",
                     "      <td>0.452844</td>\n",
                     "      <td>0.320988</td>\n",
                     "      <td>0.330010</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C2</th>\n",
                     "      <td>0.828515</td>\n",
                     "      <td>0.115637</td>\n",
                     "      <td>0.453939</td>\n",
                     "      <td>0.485497</td>\n",
                     "      <td>0.490625</td>\n",
                     "      <td>0.499528</td>\n",
                     "      <td>0.524012</td>\n",
                     "      <td>0.429976</td>\n",
                     "      <td>0.452256</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C2</th>\n",
                     "      <td>0.820255</td>\n",
                     "      <td>0.086728</td>\n",
                     "      <td>0.459306</td>\n",
                     "      <td>0.496450</td>\n",
                     "      <td>0.505203</td>\n",
                     "      <td>0.484074</td>\n",
                     "      <td>0.497628</td>\n",
                     "      <td>0.461682</td>\n",
                     "      <td>0.495278</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C3</th>\n",
                     "      <td>0.820631</td>\n",
                     "      <td>0.099869</td>\n",
                     "      <td>0.440991</td>\n",
                     "      <td>0.477441</td>\n",
                     "      <td>0.484140</td>\n",
                     "      <td>0.461169</td>\n",
                     "      <td>0.498572</td>\n",
                     "      <td>0.433199</td>\n",
                     "      <td>0.458027</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C3</th>\n",
                     "      <td>0.809461</td>\n",
                     "      <td>0.077530</td>\n",
                     "      <td>0.461246</td>\n",
                     "      <td>0.496777</td>\n",
                     "      <td>0.509009</td>\n",
                     "      <td>0.457310</td>\n",
                     "      <td>0.470865</td>\n",
                     "      <td>0.492367</td>\n",
                     "      <td>0.525708</td>\n",
                     "    </tr>\n",
                     "  </tbody>\n",
                     "</table>\n",
                     "</div>"
                  ],
                  "text/plain": [
                     "                            acc  subset_acc  macro_f1  micro_f1  weighted_f1  \\\n",
                     "GPT-3.5 + 0-shot + C1  0.827013    0.085414  0.159570  0.162653     0.155489   \n",
                     "GPT-3.5 + 1-shot + C1  0.826826    0.091984  0.232630  0.289017     0.277751   \n",
                     "GPT-3.5 + 0-shot + C2  0.836306    0.139290  0.375855  0.412003     0.411554   \n",
                     "GPT-3.5 + 1-shot + C2  0.832551    0.127464  0.386863  0.440402     0.440510   \n",
                     "GPT-3.5 + 0-shot + C3  0.828046    0.128778  0.372394  0.409794     0.410367   \n",
                     "GPT-3.5 + 1-shot + C3  0.821476    0.115637  0.377024  0.426071     0.428640   \n",
                     "GPT-4 + 0-shot + C1    0.820537    0.095926  0.355475  0.373115     0.374361   \n",
                     "GPT-4 + 1-shot + C1    0.808804    0.069645  0.361039  0.381791     0.385413   \n",
                     "GPT-4 + 0-shot + C2    0.828515    0.115637  0.453939  0.485497     0.490625   \n",
                     "GPT-4 + 1-shot + C2    0.820255    0.086728  0.459306  0.496450     0.505203   \n",
                     "GPT-4 + 0-shot + C3    0.820631    0.099869  0.440991  0.477441     0.484140   \n",
                     "GPT-4 + 1-shot + C3    0.809461    0.077530  0.461246  0.496777     0.509009   \n",
                     "\n",
                     "                       macro_precision  micro_precision  macro_recall  \\\n",
                     "GPT-3.5 + 0-shot + C1         0.635233         0.606780      0.097283   \n",
                     "GPT-3.5 + 1-shot + C1         0.565566         0.544267      0.161507   \n",
                     "GPT-3.5 + 0-shot + C2         0.562262         0.576415      0.297584   \n",
                     "GPT-3.5 + 1-shot + C2         0.526652         0.547582      0.320552   \n",
                     "GPT-3.5 + 0-shot + C3         0.499444         0.530885      0.314836   \n",
                     "GPT-3.5 + 1-shot + C3         0.464730         0.501420      0.328673   \n",
                     "GPT-4 + 0-shot + C1           0.474167         0.497378      0.295117   \n",
                     "GPT-4 + 1-shot + C1           0.452552         0.452844      0.320988   \n",
                     "GPT-4 + 0-shot + C2           0.499528         0.524012      0.429976   \n",
                     "GPT-4 + 1-shot + C2           0.484074         0.497628      0.461682   \n",
                     "GPT-4 + 0-shot + C3           0.461169         0.498572      0.433199   \n",
                     "GPT-4 + 1-shot + C3           0.457310         0.470865      0.492367   \n",
                     "\n",
                     "                       micro_recall  \n",
                     "GPT-3.5 + 0-shot + C1      0.093914  \n",
                     "GPT-3.5 + 1-shot + C1      0.196747  \n",
                     "GPT-3.5 + 0-shot + C2      0.320567  \n",
                     "GPT-3.5 + 1-shot + C2      0.368311  \n",
                     "GPT-3.5 + 0-shot + C3      0.333683  \n",
                     "GPT-3.5 + 1-shot + C3      0.370409  \n",
                     "GPT-4 + 0-shot + C1        0.298531  \n",
                     "GPT-4 + 1-shot + C1        0.330010  \n",
                     "GPT-4 + 0-shot + C2        0.452256  \n",
                     "GPT-4 + 1-shot + C2        0.495278  \n",
                     "GPT-4 + 0-shot + C3        0.458027  \n",
                     "GPT-4 + 1-shot + C3        0.525708  "
                  ]
               },
               "execution_count": 29,
               "metadata": {},
               "output_type": "execute_result"
            }
         ],
         "source": [
            "labeler_perf = pd.DataFrame([get_labeler_perf(labeler) for labeler in gpt_labelers.values()], \n",
            "                            index=[get_labeler_name(name) for name in gpt_labelers.keys()])\n",
            "labeler_perf"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "### Analysis"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 30,
         "metadata": {},
         "outputs": [],
         "source": [
            "# Join with labeling statistics\n",
            "labelers_df = pd.concat([labeler_info, labeler_statistics, labeler_perf], axis=1)"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 31,
         "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>model</th>\n",
                     "      <th>context</th>\n",
                     "      <th>shot</th>\n",
                     "      <th>valid</th>\n",
                     "      <th>invalid</th>\n",
                     "      <th>lpp</th>\n",
                     "      <th>durations</th>\n",
                     "      <th>prompt_tokens</th>\n",
                     "      <th>completion_tokens</th>\n",
                     "      <th>estimated_cost</th>\n",
                     "      <th>cost_per_1k_page</th>\n",
                     "      <th>acc</th>\n",
                     "      <th>subset_acc</th>\n",
                     "      <th>macro_f1</th>\n",
                     "      <th>micro_f1</th>\n",
                     "      <th>weighted_f1</th>\n",
                     "      <th>macro_precision</th>\n",
                     "      <th>micro_precision</th>\n",
                     "      <th>macro_recall</th>\n",
                     "      <th>micro_recall</th>\n",
                     "    </tr>\n",
                     "  </thead>\n",
                     "  <tbody>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C3</th>\n",
                     "      <td>gpt-4-1106-preview</td>\n",
                     "      <td>context3</td>\n",
                     "      <td>True</td>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>2.80 ± 1.30</td>\n",
                     "      <td>7.93 ± 2.70</td>\n",
                     "      <td>615.13 ± 124.30</td>\n",
                     "      <td>94.68 ± 6.23</td>\n",
                     "      <td>6.842570</td>\n",
                     "      <td>8.991551</td>\n",
                     "      <td>0.809461</td>\n",
                     "      <td>0.077530</td>\n",
                     "      <td>0.461246</td>\n",
                     "      <td>0.496777</td>\n",
                     "      <td>0.509009</td>\n",
                     "      <td>0.457310</td>\n",
                     "      <td>0.470865</td>\n",
                     "      <td>0.492367</td>\n",
                     "      <td>0.525708</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 1-shot + C2</th>\n",
                     "      <td>gpt-4-1106-preview</td>\n",
                     "      <td>context2</td>\n",
                     "      <td>True</td>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>2.49 ± 1.28</td>\n",
                     "      <td>5.16 ± 2.35</td>\n",
                     "      <td>448.30 ± 60.33</td>\n",
                     "      <td>92.29 ± 2.52</td>\n",
                     "      <td>5.518490</td>\n",
                     "      <td>7.251629</td>\n",
                     "      <td>0.820255</td>\n",
                     "      <td>0.086728</td>\n",
                     "      <td>0.459306</td>\n",
                     "      <td>0.496450</td>\n",
                     "      <td>0.505203</td>\n",
                     "      <td>0.484074</td>\n",
                     "      <td>0.497628</td>\n",
                     "      <td>0.461682</td>\n",
                     "      <td>0.495278</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C2</th>\n",
                     "      <td>gpt-4-1106-preview</td>\n",
                     "      <td>context2</td>\n",
                     "      <td>False</td>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>2.16 ± 1.03</td>\n",
                     "      <td>5.73 ± 1.98</td>\n",
                     "      <td>206.30 ± 60.33</td>\n",
                     "      <td>106.61 ± 7.14</td>\n",
                     "      <td>4.003900</td>\n",
                     "      <td>5.261367</td>\n",
                     "      <td>0.828515</td>\n",
                     "      <td>0.115637</td>\n",
                     "      <td>0.453939</td>\n",
                     "      <td>0.485497</td>\n",
                     "      <td>0.490625</td>\n",
                     "      <td>0.499528</td>\n",
                     "      <td>0.524012</td>\n",
                     "      <td>0.429976</td>\n",
                     "      <td>0.452256</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C3</th>\n",
                     "      <td>gpt-4-1106-preview</td>\n",
                     "      <td>context3</td>\n",
                     "      <td>False</td>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>2.30 ± 1.11</td>\n",
                     "      <td>9.44 ± 3.09</td>\n",
                     "      <td>350.13 ± 124.30</td>\n",
                     "      <td>105.80 ± 8.67</td>\n",
                     "      <td>5.079990</td>\n",
                     "      <td>6.675414</td>\n",
                     "      <td>0.820631</td>\n",
                     "      <td>0.099869</td>\n",
                     "      <td>0.440991</td>\n",
                     "      <td>0.477441</td>\n",
                     "      <td>0.484140</td>\n",
                     "      <td>0.461169</td>\n",
                     "      <td>0.498572</td>\n",
                     "      <td>0.433199</td>\n",
                     "      <td>0.458027</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C2</th>\n",
                     "      <td>gpt-3.5-turbo-1106</td>\n",
                     "      <td>context2</td>\n",
                     "      <td>True</td>\n",
                     "      <td>761</td>\n",
                     "      <td>0</td>\n",
                     "      <td>1.68 ± 1.15</td>\n",
                     "      <td>2.85 ± 0.42</td>\n",
                     "      <td>448.30 ± 60.33</td>\n",
                     "      <td>93.11 ± 4.06</td>\n",
                     "      <td>0.482875</td>\n",
                     "      <td>0.634527</td>\n",
                     "      <td>0.832551</td>\n",
                     "      <td>0.127464</td>\n",
                     "      <td>0.386863</td>\n",
                     "      <td>0.440402</td>\n",
                     "      <td>0.440510</td>\n",
                     "      <td>0.526652</td>\n",
                     "      <td>0.547582</td>\n",
                     "      <td>0.320552</td>\n",
                     "      <td>0.368311</td>\n",
                     "    </tr>\n",
                     "  </tbody>\n",
                     "</table>\n",
                     "</div>"
                  ],
                  "text/plain": [
                     "                                    model   context   shot  valid  invalid  \\\n",
                     "GPT-4 + 1-shot + C3    gpt-4-1106-preview  context3   True    761        0   \n",
                     "GPT-4 + 1-shot + C2    gpt-4-1106-preview  context2   True    761        0   \n",
                     "GPT-4 + 0-shot + C2    gpt-4-1106-preview  context2  False    761        0   \n",
                     "GPT-4 + 0-shot + C3    gpt-4-1106-preview  context3  False    761        0   \n",
                     "GPT-3.5 + 1-shot + C2  gpt-3.5-turbo-1106  context2   True    761        0   \n",
                     "\n",
                     "                               lpp    durations    prompt_tokens  \\\n",
                     "GPT-4 + 1-shot + C3    2.80 ± 1.30  7.93 ± 2.70  615.13 ± 124.30   \n",
                     "GPT-4 + 1-shot + C2    2.49 ± 1.28  5.16 ± 2.35   448.30 ± 60.33   \n",
                     "GPT-4 + 0-shot + C2    2.16 ± 1.03  5.73 ± 1.98   206.30 ± 60.33   \n",
                     "GPT-4 + 0-shot + C3    2.30 ± 1.11  9.44 ± 3.09  350.13 ± 124.30   \n",
                     "GPT-3.5 + 1-shot + C2  1.68 ± 1.15  2.85 ± 0.42   448.30 ± 60.33   \n",
                     "\n",
                     "                      completion_tokens  estimated_cost  cost_per_1k_page  \\\n",
                     "GPT-4 + 1-shot + C3        94.68 ± 6.23        6.842570          8.991551   \n",
                     "GPT-4 + 1-shot + C2        92.29 ± 2.52        5.518490          7.251629   \n",
                     "GPT-4 + 0-shot + C2       106.61 ± 7.14        4.003900          5.261367   \n",
                     "GPT-4 + 0-shot + C3       105.80 ± 8.67        5.079990          6.675414   \n",
                     "GPT-3.5 + 1-shot + C2      93.11 ± 4.06        0.482875          0.634527   \n",
                     "\n",
                     "                            acc  subset_acc  macro_f1  micro_f1  weighted_f1  \\\n",
                     "GPT-4 + 1-shot + C3    0.809461    0.077530  0.461246  0.496777     0.509009   \n",
                     "GPT-4 + 1-shot + C2    0.820255    0.086728  0.459306  0.496450     0.505203   \n",
                     "GPT-4 + 0-shot + C2    0.828515    0.115637  0.453939  0.485497     0.490625   \n",
                     "GPT-4 + 0-shot + C3    0.820631    0.099869  0.440991  0.477441     0.484140   \n",
                     "GPT-3.5 + 1-shot + C2  0.832551    0.127464  0.386863  0.440402     0.440510   \n",
                     "\n",
                     "                       macro_precision  micro_precision  macro_recall  \\\n",
                     "GPT-4 + 1-shot + C3           0.457310         0.470865      0.492367   \n",
                     "GPT-4 + 1-shot + C2           0.484074         0.497628      0.461682   \n",
                     "GPT-4 + 0-shot + C2           0.499528         0.524012      0.429976   \n",
                     "GPT-4 + 0-shot + C3           0.461169         0.498572      0.433199   \n",
                     "GPT-3.5 + 1-shot + C2         0.526652         0.547582      0.320552   \n",
                     "\n",
                     "                       micro_recall  \n",
                     "GPT-4 + 1-shot + C3        0.525708  \n",
                     "GPT-4 + 1-shot + C2        0.495278  \n",
                     "GPT-4 + 0-shot + C2        0.452256  \n",
                     "GPT-4 + 0-shot + C3        0.458027  \n",
                     "GPT-3.5 + 1-shot + C2      0.368311  "
                  ]
               },
               "execution_count": 31,
               "metadata": {},
               "output_type": "execute_result"
            }
         ],
         "source": [
            "# Top-k performing labeler\n",
            "top_k = 5\n",
            "labelers_df.sort_values(\"macro_f1\", ascending=False).head(top_k)"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 32,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "❌ Saving disabled to prevent accidental overwrite - only turn on if numbers changed\n"
               ]
            }
         ],
         "source": [
            "# Export table to latex\n",
            "df = labelers_df.copy()\n",
            "\n",
            "df[\"model\"] = df[\"model\"].apply(lambda x: rename_dict[x])\n",
            "df[\"context\"] = df[\"context\"].apply(lambda x: rename_dict[x])\n",
            "df[\"shot\"] = df[\"shot\"].apply(lambda x: rename_dict[x])\n",
            "df[\"macro_f1\"] = 100 * df[\"macro_f1\"]\n",
            "\n",
            "grouped = df.groupby(by=[\"model\", \"context\", \"shot\"]).size()\n",
            "df.index = pd.MultiIndex.from_tuples(grouped.index)\n",
            "\n",
            "cols = {\"lpp\": \"LPP\", \"cost_per_1k_page\": \"Cost ($)\", \"macro_f1\": \"M.-F1 (%)\"}\n",
            "df = df[cols.keys()].rename(columns=cols)\n",
            "\n",
            "# Save the dataframe to a latex table\n",
            "position = \"!ht\"\n",
            "latex = df.to_latex(\n",
            "    caption=\"\\\\textbf{Labeler Statistics.}\", \n",
            "    label=\"tab:labeler-results\",\n",
            "    escape=True,\n",
            "    float_format=\"%.2f\",\n",
            "    position=position,\n",
            "    multirow=True,\n",
            "    multicolumn=True,\n",
            "    multicolumn_format=\"c\",\n",
            ")\n",
            "\n",
            "# Add \\centering right after \\begin{table}\n",
            "latex = latex.replace(\"\\\\begin{table}\" + f\"[{position}]\", \"\\\\begin{table}\" + f\"[{position}]\" + \"\\n\\\\centering\")\n",
            "latex = latex.replace(\"[t]\", \"[c]\")\n",
            "print(\"❌ Saving disabled to prevent accidental overwrite - only turn on if numbers changed\")\n",
            "# save_path = os.path.join(TABLE_DIR, \"labeler-results.tex\")\n",
            "# with open(save_path, \"w\") as f:\n",
            "#     f.write(latex)\n",
            "# print(f\"✅ Saved table to {save_path}\")"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 33,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "✅ Plotted figure to /Users/jonas-mika/epfl/coursework/projects/ml-project-2-mlp/data/meta/categories.json\n"
               ]
            },
            {
               "data": {
                  "image/png": "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",
                  "text/plain": [
                     "<Figure size 500x500 with 1 Axes>"
                  ]
               },
               "metadata": {},
               "output_type": "display_data"
            }
         ],
         "source": [
            "# Time-cost tradeoff\n",
            "df = labelers_df.copy()\n",
            "df[\"time\"] = df[\"durations\"].apply(lambda x: x.split(\" ± \")[0]).astype(float)\n",
            "df[\"cost\"] = df[\"estimated_cost\"].astype(float)\n",
            "\n",
            "fig, ax = plt.subplots(figsize=(5, 5))\n",
            "sns.scatterplot(\n",
            "    data=df,\n",
            "    x=\"cost\",\n",
            "    y=\"macro_f1\",\n",
            "    style=\"model\",\n",
            "    hue=\"context\",\n",
            "    palette=palette,\n",
            "    s=200,\n",
            "    alpha=0.8,\n",
            ")\n",
            "ax.set_ylim(0, 0.5)\n",
            "ax.set_xlabel(\"Cost ($)\")\n",
            "ax.set_ylabel(\"Label Quality (M.-F1)\")\n",
            "\n",
            "# Rename legend labels\n",
            "handles, labels = ax.get_legend_handles_labels()\n",
            "labels = [f\"{rename_dict.get(label, label)}\" for label in labels]\n",
            "ax.legend(handles=handles, labels=labels)\n",
            "\n",
            "print(f\"✅ Plotted figure to {path}\")"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 34,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "✅ Plotted labeler quality as function of model, context and shot\n"
               ]
            },
            {
               "data": {
                  "image/png": "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",
                  "text/plain": [
                     "<Figure size 2000x500 with 3 Axes>"
                  ]
               },
               "metadata": {},
               "output_type": "display_data"
            }
         ],
         "source": [
            "# Univariate performance\n",
            "fig, axs = plt.subplots(ncols=3, figsize=(20, 5))\n",
            "xs = [\"model\", \"context\", \"shot\"]\n",
            "\n",
            "for ax, x in zip(axs, xs):\n",
            "    sns.barplot(\n",
            "        data=labelers_df,\n",
            "        x=x,\n",
            "        y=\"macro_f1\",\n",
            "        ax=ax\n",
            "    )\n",
            "    ax.set_xticks(ax.get_xticks())\n",
            "    ax.set_xticklabels([rename_dict.get(t.get_text(), t.get_text()) for t in ax.get_xticklabels()])\n",
            "    ax.set_xlabel(x.capitalize(), fontsize=14)\n",
            "    ax.set_ylabel(\"\")\n",
            "\n",
            "axs[0].set_ylabel(\"Macro F1\", fontsize=14)\n",
            "\n",
            "print(f\"✅ Plotted labeler quality as function of model, context and shot\")"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 35,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "✅ Saved figure to /Users/jonas-mika/epfl/coursework/projects/ml-project-2-mlp/report/figures/labeler-grid.pdf\n"
               ]
            },
            {
               "data": {
                  "image/png": "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",
                  "text/plain": [
                     "<Figure size 800x800 with 5 Axes>"
                  ]
               },
               "metadata": {},
               "output_type": "display_data"
            }
         ],
         "source": [
            "# Labeler parameters\n",
            "params = [\"model\", \"context\", \"shot\"]\n",
            "fig = utils.grid(labelers_df, params, metric=\"macro_f1\", agg=\"mean\", cmap=\"Blues\", figsize=(8, 8), rename_dict=rename_dict) \n",
            "\n",
            "path = os.path.join(FIGURE_DIR, \"labeler-grid.pdf\")\n",
            "fig.savefig(path, bbox_inches=\"tight\")\n",
            "print(f\"✅ Saved figure to {path}\")"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "### Labeling Quality: Annotator agreeement between GPT and Human\n",
            "\n",
            "We have seen in the [preprocesssing notebook](preprocessing.ipynb) that the inter annotator agreement per website on average between humans measured via Cohen's Kappa was already relatively low ($0.2$). In this section, we therefore focus on measuring the agreement level between the GPT labelers and the human annotators."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 36,
         "metadata": {},
         "outputs": [],
         "source": [
            "gpt_human_agg = dict()\n",
            "human_labels = np.array([v[\"labels\"] for v in human_labeler.get_labels().values()])\n",
            "for name, labeler in gpt_labelers.items():\n",
            "\n",
            "    # Obtain 2d array with labels for each website (n_websites, n_categories)\n",
            "    gpt_labels = np.array([v[\"labels\"] for v in labeler.get_labels().values()])\n",
            "    k = gpt_labels.shape[1]\n",
            "\n",
            "    # Ensure matching between human and gpt labels\n",
            "    human_labs_matched, gpt_labs_matched = match_labels(human_labeler, labeler)\n",
            "\n",
            "    kappa_scores = []\n",
            "    for j in range(k):\n",
            "\n",
            "        # Obtain the labels for kth class by human and gpt\n",
            "        human_kthcls_labels = human_labs_matched[:, j]\n",
            "        gpt_kthcls_labels = gpt_labs_matched[:, j]\n",
            "\n",
            "        # Compute the cohens kappa score\n",
            "        kappa = cohen_kappa_score(human_kthcls_labels, gpt_kthcls_labels)\n",
            "\n",
            "        # Append to list\n",
            "        kappa_scores.append(kappa)\n",
            "\n",
            "\n",
            "    # Save the results\n",
            "    gpt_human_agg[name] = kappa_scores"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "Let's now evalute the results, first we look at the overall picture using dataframe:"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 37,
         "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>gpt3.5-zeroshot-context1</th>\n",
                     "      <th>gpt3.5-oneshot-context1</th>\n",
                     "      <th>gpt3.5-zeroshot-context2</th>\n",
                     "      <th>gpt3.5-oneshot-context2</th>\n",
                     "      <th>gpt3.5-zeroshot-context3</th>\n",
                     "      <th>gpt3.5-oneshot-context3</th>\n",
                     "      <th>gpt4-zeroshot-context1</th>\n",
                     "      <th>gpt4-oneshot-context1</th>\n",
                     "      <th>gpt4-zeroshot-context2</th>\n",
                     "      <th>gpt4-oneshot-context2</th>\n",
                     "      <th>gpt4-zeroshot-context3</th>\n",
                     "      <th>gpt4-oneshot-context3</th>\n",
                     "    </tr>\n",
                     "  </thead>\n",
                     "  <tbody>\n",
                     "    <tr>\n",
                     "      <th>Arts</th>\n",
                     "      <td>0.16</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.47</td>\n",
                     "      <td>0.47</td>\n",
                     "      <td>0.46</td>\n",
                     "      <td>0.42</td>\n",
                     "      <td>0.33</td>\n",
                     "      <td>0.30</td>\n",
                     "      <td>0.48</td>\n",
                     "      <td>0.49</td>\n",
                     "      <td>0.47</td>\n",
                     "      <td>0.51</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Business</th>\n",
                     "      <td>0.03</td>\n",
                     "      <td>0.14</td>\n",
                     "      <td>0.23</td>\n",
                     "      <td>0.28</td>\n",
                     "      <td>0.22</td>\n",
                     "      <td>0.27</td>\n",
                     "      <td>0.16</td>\n",
                     "      <td>0.15</td>\n",
                     "      <td>0.26</td>\n",
                     "      <td>0.34</td>\n",
                     "      <td>0.26</td>\n",
                     "      <td>0.34</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Computers</th>\n",
                     "      <td>0.15</td>\n",
                     "      <td>0.21</td>\n",
                     "      <td>0.35</td>\n",
                     "      <td>0.36</td>\n",
                     "      <td>0.35</td>\n",
                     "      <td>0.34</td>\n",
                     "      <td>0.23</td>\n",
                     "      <td>0.16</td>\n",
                     "      <td>0.40</td>\n",
                     "      <td>0.38</td>\n",
                     "      <td>0.35</td>\n",
                     "      <td>0.35</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Games</th>\n",
                     "      <td>0.21</td>\n",
                     "      <td>0.30</td>\n",
                     "      <td>0.38</td>\n",
                     "      <td>0.40</td>\n",
                     "      <td>0.42</td>\n",
                     "      <td>0.40</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.28</td>\n",
                     "      <td>0.38</td>\n",
                     "      <td>0.39</td>\n",
                     "      <td>0.34</td>\n",
                     "      <td>0.36</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Health</th>\n",
                     "      <td>0.22</td>\n",
                     "      <td>0.18</td>\n",
                     "      <td>0.42</td>\n",
                     "      <td>0.40</td>\n",
                     "      <td>0.39</td>\n",
                     "      <td>0.38</td>\n",
                     "      <td>0.36</td>\n",
                     "      <td>0.34</td>\n",
                     "      <td>0.45</td>\n",
                     "      <td>0.43</td>\n",
                     "      <td>0.46</td>\n",
                     "      <td>0.43</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Home</th>\n",
                     "      <td>0.03</td>\n",
                     "      <td>0.01</td>\n",
                     "      <td>0.11</td>\n",
                     "      <td>0.09</td>\n",
                     "      <td>0.06</td>\n",
                     "      <td>0.08</td>\n",
                     "      <td>0.13</td>\n",
                     "      <td>0.16</td>\n",
                     "      <td>0.15</td>\n",
                     "      <td>0.16</td>\n",
                     "      <td>0.13</td>\n",
                     "      <td>0.21</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Kids_and_Teens</th>\n",
                     "      <td>0.10</td>\n",
                     "      <td>0.18</td>\n",
                     "      <td>0.15</td>\n",
                     "      <td>0.17</td>\n",
                     "      <td>0.08</td>\n",
                     "      <td>0.15</td>\n",
                     "      <td>0.27</td>\n",
                     "      <td>0.21</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.34</td>\n",
                     "      <td>0.23</td>\n",
                     "      <td>0.31</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>News</th>\n",
                     "      <td>0.14</td>\n",
                     "      <td>0.13</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.27</td>\n",
                     "      <td>0.24</td>\n",
                     "      <td>0.21</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.23</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.30</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.32</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Recreation</th>\n",
                     "      <td>0.01</td>\n",
                     "      <td>0.12</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.27</td>\n",
                     "      <td>0.18</td>\n",
                     "      <td>0.24</td>\n",
                     "      <td>0.21</td>\n",
                     "      <td>0.32</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.31</td>\n",
                     "      <td>0.21</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Reference</th>\n",
                     "      <td>0.04</td>\n",
                     "      <td>0.06</td>\n",
                     "      <td>0.18</td>\n",
                     "      <td>0.19</td>\n",
                     "      <td>0.12</td>\n",
                     "      <td>0.04</td>\n",
                     "      <td>0.20</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.27</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.24</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Science</th>\n",
                     "      <td>0.12</td>\n",
                     "      <td>0.09</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.31</td>\n",
                     "      <td>0.21</td>\n",
                     "      <td>0.31</td>\n",
                     "      <td>0.28</td>\n",
                     "      <td>0.31</td>\n",
                     "      <td>0.42</td>\n",
                     "      <td>0.33</td>\n",
                     "      <td>0.37</td>\n",
                     "      <td>0.35</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Shopping</th>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.27</td>\n",
                     "      <td>0.36</td>\n",
                     "      <td>0.32</td>\n",
                     "      <td>0.39</td>\n",
                     "      <td>0.35</td>\n",
                     "      <td>0.31</td>\n",
                     "      <td>0.33</td>\n",
                     "      <td>0.37</td>\n",
                     "      <td>0.40</td>\n",
                     "      <td>0.38</td>\n",
                     "      <td>0.35</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Society</th>\n",
                     "      <td>0.05</td>\n",
                     "      <td>0.09</td>\n",
                     "      <td>0.19</td>\n",
                     "      <td>0.27</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.23</td>\n",
                     "      <td>0.22</td>\n",
                     "      <td>0.35</td>\n",
                     "      <td>0.37</td>\n",
                     "      <td>0.34</td>\n",
                     "      <td>0.36</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>Sports</th>\n",
                     "      <td>0.16</td>\n",
                     "      <td>0.13</td>\n",
                     "      <td>0.38</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.42</td>\n",
                     "      <td>0.36</td>\n",
                     "      <td>0.36</td>\n",
                     "      <td>0.31</td>\n",
                     "      <td>0.45</td>\n",
                     "      <td>0.41</td>\n",
                     "      <td>0.46</td>\n",
                     "      <td>0.44</td>\n",
                     "    </tr>\n",
                     "  </tbody>\n",
                     "</table>\n",
                     "</div>"
                  ],
                  "text/plain": [
                     "                gpt3.5-zeroshot-context1  gpt3.5-oneshot-context1  \\\n",
                     "Arts                                0.16                     0.29   \n",
                     "Business                            0.03                     0.14   \n",
                     "Computers                           0.15                     0.21   \n",
                     "Games                               0.21                     0.30   \n",
                     "Health                              0.22                     0.18   \n",
                     "Home                                0.03                     0.01   \n",
                     "Kids_and_Teens                      0.10                     0.18   \n",
                     "News                                0.14                     0.13   \n",
                     "Recreation                          0.01                     0.12   \n",
                     "Reference                           0.04                     0.06   \n",
                     "Science                             0.12                     0.09   \n",
                     "Shopping                            0.25                     0.27   \n",
                     "Society                             0.05                     0.09   \n",
                     "Sports                              0.16                     0.13   \n",
                     "\n",
                     "                gpt3.5-zeroshot-context2  gpt3.5-oneshot-context2  \\\n",
                     "Arts                                0.47                     0.47   \n",
                     "Business                            0.23                     0.28   \n",
                     "Computers                           0.35                     0.36   \n",
                     "Games                               0.38                     0.40   \n",
                     "Health                              0.42                     0.40   \n",
                     "Home                                0.11                     0.09   \n",
                     "Kids_and_Teens                      0.15                     0.17   \n",
                     "News                                0.29                     0.27   \n",
                     "Recreation                          0.29                     0.25   \n",
                     "Reference                           0.18                     0.19   \n",
                     "Science                             0.25                     0.31   \n",
                     "Shopping                            0.36                     0.32   \n",
                     "Society                             0.19                     0.27   \n",
                     "Sports                              0.38                     0.29   \n",
                     "\n",
                     "                gpt3.5-zeroshot-context3  gpt3.5-oneshot-context3  \\\n",
                     "Arts                                0.46                     0.42   \n",
                     "Business                            0.22                     0.27   \n",
                     "Computers                           0.35                     0.34   \n",
                     "Games                               0.42                     0.40   \n",
                     "Health                              0.39                     0.38   \n",
                     "Home                                0.06                     0.08   \n",
                     "Kids_and_Teens                      0.08                     0.15   \n",
                     "News                                0.24                     0.21   \n",
                     "Recreation                          0.27                     0.18   \n",
                     "Reference                           0.12                     0.04   \n",
                     "Science                             0.21                     0.31   \n",
                     "Shopping                            0.39                     0.35   \n",
                     "Society                             0.25                     0.29   \n",
                     "Sports                              0.42                     0.36   \n",
                     "\n",
                     "                gpt4-zeroshot-context1  gpt4-oneshot-context1  \\\n",
                     "Arts                              0.33                   0.30   \n",
                     "Business                          0.16                   0.15   \n",
                     "Computers                         0.23                   0.16   \n",
                     "Games                             0.25                   0.28   \n",
                     "Health                            0.36                   0.34   \n",
                     "Home                              0.13                   0.16   \n",
                     "Kids_and_Teens                    0.27                   0.21   \n",
                     "News                              0.25                   0.23   \n",
                     "Recreation                        0.24                   0.21   \n",
                     "Reference                         0.20                   0.29   \n",
                     "Science                           0.28                   0.31   \n",
                     "Shopping                          0.31                   0.33   \n",
                     "Society                           0.23                   0.22   \n",
                     "Sports                            0.36                   0.31   \n",
                     "\n",
                     "                gpt4-zeroshot-context2  gpt4-oneshot-context2  \\\n",
                     "Arts                              0.48                   0.49   \n",
                     "Business                          0.26                   0.34   \n",
                     "Computers                         0.40                   0.38   \n",
                     "Games                             0.38                   0.39   \n",
                     "Health                            0.45                   0.43   \n",
                     "Home                              0.15                   0.16   \n",
                     "Kids_and_Teens                    0.25                   0.34   \n",
                     "News                              0.29                   0.30   \n",
                     "Recreation                        0.32                   0.25   \n",
                     "Reference                         0.29                   0.27   \n",
                     "Science                           0.42                   0.33   \n",
                     "Shopping                          0.37                   0.40   \n",
                     "Society                           0.35                   0.37   \n",
                     "Sports                            0.45                   0.41   \n",
                     "\n",
                     "                gpt4-zeroshot-context3  gpt4-oneshot-context3  \n",
                     "Arts                              0.47                   0.51  \n",
                     "Business                          0.26                   0.34  \n",
                     "Computers                         0.35                   0.35  \n",
                     "Games                             0.34                   0.36  \n",
                     "Health                            0.46                   0.43  \n",
                     "Home                              0.13                   0.21  \n",
                     "Kids_and_Teens                    0.23                   0.31  \n",
                     "News                              0.25                   0.32  \n",
                     "Recreation                        0.31                   0.21  \n",
                     "Reference                         0.25                   0.24  \n",
                     "Science                           0.37                   0.35  \n",
                     "Shopping                          0.38                   0.35  \n",
                     "Society                           0.34                   0.36  \n",
                     "Sports                            0.46                   0.44  "
                  ]
               },
               "execution_count": 37,
               "metadata": {},
               "output_type": "execute_result"
            }
         ],
         "source": [
            "aggreement_df = pd.DataFrame(gpt_human_agg, index=categories).round(2)\n",
            "aggreement_df"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "Let's now get some insights about the statistics of the agreement level between the GPT labelers and the human annotators:"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 38,
         "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>gpt3.5-zeroshot-context1</th>\n",
                     "      <th>gpt3.5-oneshot-context1</th>\n",
                     "      <th>gpt3.5-zeroshot-context2</th>\n",
                     "      <th>gpt3.5-oneshot-context2</th>\n",
                     "      <th>gpt3.5-zeroshot-context3</th>\n",
                     "      <th>gpt3.5-oneshot-context3</th>\n",
                     "      <th>gpt4-zeroshot-context1</th>\n",
                     "      <th>gpt4-oneshot-context1</th>\n",
                     "      <th>gpt4-zeroshot-context2</th>\n",
                     "      <th>gpt4-oneshot-context2</th>\n",
                     "      <th>gpt4-zeroshot-context3</th>\n",
                     "      <th>gpt4-oneshot-context3</th>\n",
                     "    </tr>\n",
                     "  </thead>\n",
                     "  <tbody>\n",
                     "    <tr>\n",
                     "      <th>count</th>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "      <td>14.00</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>mean</th>\n",
                     "      <td>0.12</td>\n",
                     "      <td>0.16</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.28</td>\n",
                     "      <td>0.27</td>\n",
                     "      <td>0.26</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.35</td>\n",
                     "      <td>0.35</td>\n",
                     "      <td>0.33</td>\n",
                     "      <td>0.34</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>std</th>\n",
                     "      <td>0.08</td>\n",
                     "      <td>0.09</td>\n",
                     "      <td>0.11</td>\n",
                     "      <td>0.10</td>\n",
                     "      <td>0.13</td>\n",
                     "      <td>0.12</td>\n",
                     "      <td>0.07</td>\n",
                     "      <td>0.07</td>\n",
                     "      <td>0.09</td>\n",
                     "      <td>0.08</td>\n",
                     "      <td>0.10</td>\n",
                     "      <td>0.08</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>min</th>\n",
                     "      <td>0.01</td>\n",
                     "      <td>0.01</td>\n",
                     "      <td>0.11</td>\n",
                     "      <td>0.09</td>\n",
                     "      <td>0.06</td>\n",
                     "      <td>0.04</td>\n",
                     "      <td>0.13</td>\n",
                     "      <td>0.15</td>\n",
                     "      <td>0.15</td>\n",
                     "      <td>0.16</td>\n",
                     "      <td>0.13</td>\n",
                     "      <td>0.21</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>25%</th>\n",
                     "      <td>0.04</td>\n",
                     "      <td>0.10</td>\n",
                     "      <td>0.20</td>\n",
                     "      <td>0.26</td>\n",
                     "      <td>0.21</td>\n",
                     "      <td>0.19</td>\n",
                     "      <td>0.23</td>\n",
                     "      <td>0.21</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.31</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.31</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>50%</th>\n",
                     "      <td>0.13</td>\n",
                     "      <td>0.14</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.29</td>\n",
                     "      <td>0.26</td>\n",
                     "      <td>0.30</td>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.26</td>\n",
                     "      <td>0.36</td>\n",
                     "      <td>0.36</td>\n",
                     "      <td>0.34</td>\n",
                     "      <td>0.35</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>75%</th>\n",
                     "      <td>0.16</td>\n",
                     "      <td>0.20</td>\n",
                     "      <td>0.38</td>\n",
                     "      <td>0.35</td>\n",
                     "      <td>0.39</td>\n",
                     "      <td>0.36</td>\n",
                     "      <td>0.30</td>\n",
                     "      <td>0.31</td>\n",
                     "      <td>0.42</td>\n",
                     "      <td>0.40</td>\n",
                     "      <td>0.38</td>\n",
                     "      <td>0.36</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>max</th>\n",
                     "      <td>0.25</td>\n",
                     "      <td>0.30</td>\n",
                     "      <td>0.47</td>\n",
                     "      <td>0.47</td>\n",
                     "      <td>0.46</td>\n",
                     "      <td>0.42</td>\n",
                     "      <td>0.36</td>\n",
                     "      <td>0.34</td>\n",
                     "      <td>0.48</td>\n",
                     "      <td>0.49</td>\n",
                     "      <td>0.47</td>\n",
                     "      <td>0.51</td>\n",
                     "    </tr>\n",
                     "  </tbody>\n",
                     "</table>\n",
                     "</div>"
                  ],
                  "text/plain": [
                     "       gpt3.5-zeroshot-context1  gpt3.5-oneshot-context1  \\\n",
                     "count                     14.00                    14.00   \n",
                     "mean                       0.12                     0.16   \n",
                     "std                        0.08                     0.09   \n",
                     "min                        0.01                     0.01   \n",
                     "25%                        0.04                     0.10   \n",
                     "50%                        0.13                     0.14   \n",
                     "75%                        0.16                     0.20   \n",
                     "max                        0.25                     0.30   \n",
                     "\n",
                     "       gpt3.5-zeroshot-context2  gpt3.5-oneshot-context2  \\\n",
                     "count                     14.00                    14.00   \n",
                     "mean                       0.29                     0.29   \n",
                     "std                        0.11                     0.10   \n",
                     "min                        0.11                     0.09   \n",
                     "25%                        0.20                     0.26   \n",
                     "50%                        0.29                     0.29   \n",
                     "75%                        0.38                     0.35   \n",
                     "max                        0.47                     0.47   \n",
                     "\n",
                     "       gpt3.5-zeroshot-context3  gpt3.5-oneshot-context3  \\\n",
                     "count                     14.00                    14.00   \n",
                     "mean                       0.28                     0.27   \n",
                     "std                        0.13                     0.12   \n",
                     "min                        0.06                     0.04   \n",
                     "25%                        0.21                     0.19   \n",
                     "50%                        0.26                     0.30   \n",
                     "75%                        0.39                     0.36   \n",
                     "max                        0.46                     0.42   \n",
                     "\n",
                     "       gpt4-zeroshot-context1  gpt4-oneshot-context1  gpt4-zeroshot-context2  \\\n",
                     "count                   14.00                  14.00                   14.00   \n",
                     "mean                     0.26                   0.25                    0.35   \n",
                     "std                      0.07                   0.07                    0.09   \n",
                     "min                      0.13                   0.15                    0.15   \n",
                     "25%                      0.23                   0.21                    0.29   \n",
                     "50%                      0.25                   0.26                    0.36   \n",
                     "75%                      0.30                   0.31                    0.42   \n",
                     "max                      0.36                   0.34                    0.48   \n",
                     "\n",
                     "       gpt4-oneshot-context2  gpt4-zeroshot-context3  gpt4-oneshot-context3  \n",
                     "count                  14.00                   14.00                  14.00  \n",
                     "mean                    0.35                    0.33                   0.34  \n",
                     "std                     0.08                    0.10                   0.08  \n",
                     "min                     0.16                    0.13                   0.21  \n",
                     "25%                     0.31                    0.25                   0.31  \n",
                     "50%                     0.36                    0.34                   0.35  \n",
                     "75%                     0.40                    0.38                   0.36  \n",
                     "max                     0.49                    0.47                   0.51  "
                  ]
               },
               "execution_count": 38,
               "metadata": {},
               "output_type": "execute_result"
            }
         ],
         "source": [
            "agreement_stats = aggreement_df.describe().round(2)\n",
            "agreement_stats"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "We can see that as the complexity of the labeler increases, the more average agreement we obtain with the human annotators with plus or minus same standard deviation. Interestingly, we can see that for some categories  and gpt4 based labelers we obtain almost $0.5$ agreement. "
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "### Curlie\n",
            "\n",
            "In this section we explore the labels of the `curlie-1000` that we relabel with the most promising GPT labelers given the label quality and cost of labeling as shown in the section above."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 39,
         "metadata": {},
         "outputs": [],
         "source": [
            "# Initialise data and labeler config\n",
            "curlie_cfg = hydra.compose(config_name=\"eda\", overrides=[\"data=curlie\"])\n",
            "curlie_labeler_names = [\"gpt3.5-oneshot-context2\", \"gpt4-zeroshot-context2\"]\n",
            "curlie_labeler_cfg = {name: hydra.compose(config_name=\"eda\", overrides=[f\"labeler={name}\"]) for name in curlie_labeler_names}"
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 40,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "✅ Initialised 2 labeler(s).\n"
               ]
            }
         ],
         "source": [
            "# Instantiate data\n",
            "curlie_data = hydra.utils.instantiate(curlie_cfg.data)\n",
            "curlie_labeler = {name: hydra.utils.instantiate(curlie_labeler.labeler, data=curlie_data) for name, curlie_labeler in curlie_labeler_cfg.items()}\n",
            "print(f\"✅ Initialised {len(curlie_labeler)} labeler(s).\")"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "We expect that the LPP will be higher as the labelers have learned to replicate the human labels."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 41,
         "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>valid</th>\n",
                     "      <th>invalid</th>\n",
                     "      <th>lpp</th>\n",
                     "      <th>durations</th>\n",
                     "      <th>prompt_tokens</th>\n",
                     "      <th>completion_tokens</th>\n",
                     "      <th>estimated_cost</th>\n",
                     "      <th>cost_per_1k_page</th>\n",
                     "    </tr>\n",
                     "  </thead>\n",
                     "  <tbody>\n",
                     "    <tr>\n",
                     "      <th>GPT-3.5 + 1-shot + C2</th>\n",
                     "      <td>9190</td>\n",
                     "      <td>0</td>\n",
                     "      <td>1.60 ± 1.08</td>\n",
                     "      <td>2.09 ± 1.04</td>\n",
                     "      <td>470.16 ± 87.36</td>\n",
                     "      <td>92.99 ± 3.85</td>\n",
                     "      <td>6.029949</td>\n",
                     "      <td>0.656142</td>\n",
                     "    </tr>\n",
                     "    <tr>\n",
                     "      <th>GPT-4 + 0-shot + C2</th>\n",
                     "      <td>9190</td>\n",
                     "      <td>0</td>\n",
                     "      <td>2.03 ± 0.97</td>\n",
                     "      <td>7.80 ± 3.32</td>\n",
                     "      <td>228.16 ± 87.36</td>\n",
                     "      <td>106.16 ± 8.10</td>\n",
                     "      <td>50.235890</td>\n",
                     "      <td>5.466365</td>\n",
                     "    </tr>\n",
                     "  </tbody>\n",
                     "</table>\n",
                     "</div>"
                  ],
                  "text/plain": [
                     "                       valid  invalid          lpp    durations  \\\n",
                     "GPT-3.5 + 1-shot + C2   9190        0  1.60 ± 1.08  2.09 ± 1.04   \n",
                     "GPT-4 + 0-shot + C2     9190        0  2.03 ± 0.97  7.80 ± 3.32   \n",
                     "\n",
                     "                        prompt_tokens completion_tokens  estimated_cost  \\\n",
                     "GPT-3.5 + 1-shot + C2  470.16 ± 87.36      92.99 ± 3.85        6.029949   \n",
                     "GPT-4 + 0-shot + C2    228.16 ± 87.36     106.16 ± 8.10       50.235890   \n",
                     "\n",
                     "                       cost_per_1k_page  \n",
                     "GPT-3.5 + 1-shot + C2          0.656142  \n",
                     "GPT-4 + 0-shot + C2            5.466365  "
                  ]
               },
               "execution_count": 41,
               "metadata": {},
               "output_type": "execute_result"
            }
         ],
         "source": [
            "values = [get_statistics(labeler) for labeler in curlie_labeler.values()]\n",
            "index = [get_labeler_name(name) for name in curlie_labeler.keys()]\n",
            "labeler_statistics = pd.DataFrame(values, index=index)\n",
            "labeler_statistics"
         ]
      },
      {
         "cell_type": "markdown",
         "metadata": {},
         "source": [
            "Let's first look at the distribution of labels for each category."
         ]
      },
      {
         "cell_type": "code",
         "execution_count": 42,
         "metadata": {},
         "outputs": [
            {
               "name": "stdout",
               "output_type": "stream",
               "text": [
                  "✅ Saved figure to /Users/jonas-mika/epfl/coursework/projects/ml-project-2-mlp/report/figures/curlie-10k-dist.pdf\n"
               ]
            },
            {
               "data": {
                  "image/png": "iVBORw0KGgoAAAANSUhEUgAAAroAAAFPCAYAAACiSm0uAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8g+/7EAAAACXBIWXMAAA9hAAAPYQGoP6dpAACQRklEQVR4nOzdd1gT2dcH8G8SiiAqitgr6oIFF0Sxd1REbKiICqjYC+raEbEsKioWLKDYe8G6Ntbu2rtiR0FFEEEQkN6S8/7Bm/mBFRBIwPN5Hp9dJpPMmclk5sydc++IiIjAGGOMMcZYESNWdACMMcYYY4zlB050GWOMMcZYkcSJLmOMMcYYK5I40WWMMcYYY0USJ7qMMcYYY6xI4kSXMcYYY4wVSZzoMsYYY4yxIokTXcYYY4wxViSpKDqA/CKTyZCeng6xWAyRSKTocBhjjDHGWB4gIshkMqioqEAs/nGbbZFNdNPT0/H48WNFh8EYY4wxxvKBoaEh1NTUfjhPkU105Rm+oaEhJBKJgqNhjDHGGGN5QSqV4vHjxz9tzQWKcKIrL1eQSCSc6DLGGGOMFTHZKU3lzmiMMcYYY6xI4kSXMcYYY4wVSZzoMsYYY4yxIqnI1ugyxhhjTHkREdLT0yGVShUdClMyEokEKioqeTI8LCe6jDHGGCtQqamp+PDhAxITExUdClNSmpqaqFix4k+HD/sZTnQZY4wxVmBkMhnevHkDiUSCSpUqQU1NjR/sxAREhNTUVERERODNmzeoU6dOtoYR+x5OdBljjDFWYFJTUyGTyVC1alVoamoqOhymhDQ0NKCqqoqgoCCkpqaiWLFiuf4s7ozGGGOMsQL3K610rOjLq/2D97IfkMlk+To/Y4wxxhjLP1y68ANisRgeHv8gJCTyp/NWqVIWkyb1LICoGGOMMcZYdnCL7k+EhETizZvwn/6Lj0+CTEY5+uyczs8YY4wxxdLX18etW7dy9V47OzusWbMmV++9desW9PX1c/Xe3xm36OaR4sWLQSwWwcfnFSIikn46v66uBqyt6xRAZIwxxhhjvydOdPNYREQSQkMTFB0GY4wxxthvj0sXGGOMMcbyABFh/fr16NChAxo0aIBWrVph7dq1WeYJCwuDra0tDA0NYW1tjRcvXgivxcbGYtq0aWjUqBFatWoFV1dXJCcnf3NZHz58wOjRo/Hnn3+iQ4cOWLt2rfCUucOHD8PGxgbjxo2DiYkJjh07ln8rreQUmugGBQVh2LBhMDY2Rrt27bBp0ybhtQULFkBfXz/Lv127dikwWsYYY4yx7zt69Ci2b9+OhQsX4t9//8W4ceOwZs0aPH36VJjnyJEjMDc3x9GjR1G1alWMHz9eSFCdnZ0RFxeHvXv3wsvLC48fP8bff//91XKICOPHj4eOjg6OHDkCNzc3HD9+HOvXrxfmefDgAWrXrg0fHx+0atUq/1deSSmsdEEmk2HkyJEwNDTEkSNHEBQUhMmTJ6N8+fLo3r07AgMDMWXKFPTu3Vt4j5aWlqLCZYwxxhj7oYoVK8LNzQ3NmzcHAAwYMACenp549eoV6tevDwAwMzODra0tAGD+/Plo3bo1rl27hho1auDcuXO4ffs2SpQoAQBwdXVFr1694OTklGU5N2/eRGhoKA4cOACxWAw9PT3MmDEDTk5OGDduHABAJBJhzJgxv/SwhaJAYYluZGQk6tati3nz5kFLSws1atRA8+bNce/ePSHRHTZsGHR1dRUVImOMMcZYtjVr1gx+fn5Yvnw5AgMD8fz5c0RERGQZZ79hw4bC/2tpaaFmzZp4/fo1pFIpZDIZ2rRpk+UzZTIZgoKCskwLDAxETEwMTExMssyXnJyM6OhoAICOjs5vn+QCCkx0y5UrBw8PDwAZTfD379/HnTt3MHfuXMTHxyM8PBw1atRQVHiMMcYYYzly4MABLFq0CP369UPnzp0xY8YM2NvbZ5lHIpFk+Vsmk0FVVRVSqRQlSpTAoUOHvvrc8uXLw8/PT/g7PT0denp68PLy+mpeeWuwurp6XqxSoacUoy506NABoaGhaN++Pbp06YInT55AJBJh/fr1uHz5MrS1tTF06NAsZQzZJa97yY0vd8b88CvxMcYYY4WNVCoFEQn/CqPvxb53716MGzcOw4YNA5DRuezTp0+QyWTCe/z9/YX3xsbG4u3bt6hZsybKlSuHuLg4AEC1atUAAP7+/lizZg0WLVokvIeIULNmTYSGhqJ06dJCYnvt2jUcOXIES5YsKfTbF/jfNpZKpV/lSjnJnZQi0V29ejUiIyMxb948uLm5oX79+hCJRNDT04OtrS3u3LkDFxcXaGlpoVOnTjn67MePH+cqJg0NDdSrVy9X780Jf39/JCX9fNxdxhhjrKhQUVFBUlJSllv6hcm9e/cQGxubZVqjRo1QokQJXLlyBS1atEBCQgLWrl2LtLQ0JCQkIDExETKZDCdOnED9+vVhZGQET09PVK1aFQ0bNoRIJEKLFi0wefJkzJgxA2KxGK6urihVqhRUVFSQkpICAEhMTISxsTEqVqyIyZMnY/z48YiLi8OcOXPQtGlTpKSkIDU1FUSExMRERWyePJGSkoK0tLQso1LkhlIkuoaGhgAyVmrq1Km4f/8+2rdvD21tbQCAgYEB3r59i7179+Y40TU0NCyQltnc4qecMMYY+50kJycjKCgIGhoahbaGdPXq1V9NO336NObMmQMnJyfY2NhAR0cHXbt2RYkSJRAQEABNTU2IxWLY2tri+PHjWLp0KYyNjeHp6YnixYsDAJYvXw5XV1eMHj0aKioqaNWqFWbPng1NTU2hFEFTUxMAsH79eixYsACDBw+GpqYmzM3NMX36dBQrVgxqamoQiUTCvIWRWCyGqqoqateu/dV+IpVKs92QqdDOaA8fPoSZmZkwrXbt2khLS0N8fDzKlCmTZX49PT3cvHkzx8uRSCRKnegqc2yMMcZYXpNIJBCJRMK/wsbf3/+Hr/v4+Hz3tZ8Nk1qmTBmsXLnym681a9Ysy7KrVauGDRs2fHPePn36oE+fPj9clrKT7x+/mscpbBzdkJAQjB8/HuHh4cK0J0+eoEyZMti5cyeGDBmSZf4XL15AT0+vgKNkjDHGGGOFlcISXUNDQ9SvXx+zZs1CQEAA/vvvP7i7u2P06NFo37497ty5g82bN+Pdu3fYs2cPjh49CgcHB0WFyxhjjDHGChmFlS5IJBJ4eXnB1dUV/fv3h4aGBuzs7GBvbw+RSIRVq1Zh9erVWLVqFSpXrozly5fD2NhYUeEyxhhjjLFCRqGd0cqXL//VM6DlzMzMstTvMsYYY4wxlhMKK11gjDHGGGMsP3GiyxhjjDHGiiROdBljjDHGWJHEiS5jjDHGGCuSONFljDHGGMuGjx8/wsXFBa1atULDhg3RrVs3bN68Genp6T98n52dHdasWZOtZXTo0AGHDx/+5Vhv3br106evfvjwAbNnz0abNm1gZGSEXr164ejRo7+03MzrOnPmTMycOfOXPu9XKcUjgBljjDH2e5NJZRBLCq79LafL+/DhA2xsbKCnpwcPDw+UL18ejx8/xrJly3Dz5k14e3tDLP72561ZswaqqqrZWs7BgwcL5NG9b9++xcCBA9GoUSOsWrUKOjo6uHHjBubOnYuoqKg8eXaBs7NzHkT6azjRZYwxxpjCiSVi+Mxdhoi3Ifm+LN0aVWA9f2qO3uPq6oqqVati06ZNwiNpq1atCiMjI3Tr1g179+7FoEGDvvlebW3tbC+nTJkyOYort+bPnw8DAwOsWbNGeBRztWrVkJqaihUrVqBv374oWbLkLy2jRIkSeRHqL+FElzHGGGNKIeJtCEJfBio6jK9ERkbiwoUL8Pb2FpJcuUqVKsHKygo+Pj4YNGgQDh8+DB8fH+jo6ODmzZuYO3cuDhw4AFNTUzg6OgIAtm3bhs2bNyMhIQFWVlbw9/dH7969YWVlhQ4dOmD8+PGwsrKCnZ0dWrRogbt37+LOnTuoWLEiZs+ejdatWwMAAgIC4Obmhvv37yM9PR2GhoZwdXVFrVq1frg+YWFhuHHjBjZs2CAkuXJ9+/ZF3bp1oampiZCQEHTs2BHnz59HlSpVAGS0Tt++fRs7d+785rpmJi9bWLx4MQDg7NmzWLlyJd6/f486depg+vTpMDU1zeW3kj1co8sYY4wx9gNPnz4FEcHQ0PCbr5uYmODFixdITU0FADx48AC1a9eGj48PWrVqlWXeY8eOYfXq1Zg1axb279+PkJAQ3Llz57vLXr9+Pbp164YTJ07AwMAALi4ukMlkkMlkGD16NCpXrox//vkH+/btg1Qqhbu7+0/Xx9/f/7vro6GhgcaNG0NFJXttoT9a18xevHiBGTNmYMyYMTh27Bh69OiBESNGICgoKFvLyS1OdBljjDHGfuDz588A8N1b+fLpMTExAACRSIQxY8agVq1aX5Ui7NmzB4MHD0bXrl1Rp04dLFmyBMWKFfvustu2bQsrKytUq1YNY8aMwYcPHxAREYHk5GTY2Nhg5syZqFatGurXr4/evXsjICDgp+sTGxsLIG9KC360rplt3rwZ1tbW6N69O6pXrw57e3u0adMGe/fu/eUYfoRLFxhjjDHGfqBUqVIAMkoYKlSo8NXrHz9+BPC/WlwdHZ3vJq/+/v4YOXJkls+uWbPmd5ddo0YN4f+1tLQAAOnp6dDU1MSAAQNw9OhRPHnyBK9fv8azZ89QtmzZn66PPM7Y2Nhfrgn+0bpmFhgYCF9fX+zfv1+YlpaW9sNW4LzAiS5jjDHG2A8YGhpCIpHgyZMn30x0nzx5An19faipqQEA1NXVv/tZEokERJRl2pd/Z/at0RqICAkJCejbty9Kly6NDh06wNLSEq9fv8aWLVt+uj7169eHSCTCkydP0KZNmyyvJSYmYty4cZgxY8Y3W3y/HErtR+uamVQqxYgRI9CrV68s07OTJP8KLl1gjDHGGPuBMmXKwMzMDF5eXpBKpVle+/DhAw4ePAhra+tsfVbt2rXx9OlT4e/4+Phc1anevn0bHz9+xI4dOzB8+HC0aNECoaGhP0ya5cqUKYOWLVti+/btX81/6NAh3L17FxUrVhSS7ISEBOH1kJDcjYpRs2ZNhISEoHr16sK//fv34/Lly7n6vOziRJcxxhhj7CecnZ3x+fNnjBgxAnfv3kVoaCjOnj0Le3t7mJqaYuDAgdn6HDs7O+zYsQNnzpxBYGAgZs2ahcTExK9GP/gZbW1tJCYm4ty5cwgJCcGBAwewe/duoUPczzg5OeHRo0eYOHEiHj16hDdv3mDLli1wd3fHlClTUKpUKZQtWxYVK1bE5s2bERwcjMOHD+PSpUs5ilNuyJAhOHXqFHbs2IF3795h27Zt2LZtW5bSjPzApQuMMcYYUwq6Naoo7XLKly8PHx8feHl5YerUqYiKikLVqlVhY2ODwYMHf/dhEV/q1q0bgoKCMHfuXKSkpKB///6oXLlyth8oIWdsbIxx48Zh/vz5SElJgb6+PubMmQNnZ2eEh4f/9P21a9fGnj17sGbNGowZMwYJCQnQ09PDwoUL0b17dwCAWCzGwoUL4erqCgsLCzRv3hyjR4/OVSuskZERli5dijVr1mDp0qWoVq0ali9fjiZNmuT4s3JCRNlp4y6EpFIpHj58CCMjo6/GvMuJqVM3482bn+8wrVrVw19/9YKn5yOEhib8dP5KlYpj3LiGuY6LMcYYK4ySk5Px5s0b1KxZM0t9prI/GS2v3L59G1WrVkXFihUBZNS8NmvWDJ6enmjatGmBx6OsvrefADnL8bhFlzHGGGMKV9BJpyKSXAA4d+4cHjx4gPnz56N48eLYsWMHtLS0YGRkpJB4ijqu0WWMMcYYKyATJkxAzZo1MXToUPTs2ROvX7/Gpk2bsj16AcsZbtFljDHGGCsgWlpaWLp0qaLD+G0otEU3KCgIw4YNg7GxMdq1a4dNmzYJrwUHB2PIkCEwMjKChYUFrl69qsBIGWOMMcZYYaOwRFcmk2HkyJEoXbo0jhw5gvnz52PdunU4fvw4iAjjxo1D2bJlcejQIfTs2RPjx49HaGioosJljDHGGGOFjMJKFyIjI1G3bl3MmzcPWlpaqFGjBpo3b4579+6hbNmyCA4Oxr59+6CpqYlatWrhxo0bOHToEBwdHRUVMmOMMcYYK0QU1qJbrlw5eHh4QEtLC0SEe/fu4c6dOzA1NYWfnx/q1asHTU1NYX4TExM8fPhQUeEyxhhjjLFCRilGXejQoQMGDhwIY2NjdOnSBREREShXrlyWeXR0dBAWFqagCBljjDHGWGGjFKMurF69GpGRkZg3bx7c3NyQlJQENTW1LPOoqall+7F2mX35TOqc+JUHTWTXr8THGGOMFTZSqRREJPxj7Fvk+4dUKv0qV8pJ7qQUia6hoSEAICUlBVOnTkWfPn2QlJSUZZ7U1NSvnoyRHY8fP85VTBoaGqhXr16u3psT/v7+X60rY4wxVpSpqKggKSkJMplMmCYSiSASiQoshtwm2klJSdi6dSvOnTuHDx8+QENDAyYmJhg9ejRq1aqF0NBQWFpaZnmPuro66tatCwcHB7Rq1Qrr16/Hhg0bvruMefPmoUePHl9N37VrF/bs2YOYmBgYGRlh5syZqFat2jc/48WLFxg4cGCWaXXr1sXu3btzvM6KkJKSgrS0NLx48eKXPkehndEePnwIMzMzYVrt2rWRlpYGXV1dvH79+qv5vyxnyA5DQ8MCaZnNLX19fUWHwBhjjBWY5ORkBAUFQUNDI0sDFhFBLC7ARwDLZDlOrBMSEjBs2DAkJiZi5syZ0NfXR3R0NHbv3o2hQ4fi6NGjwjodOHAAFSpUAJCxzjt27MCUKVNw8uRJjBo1CnZ2dgCABw8eYMKECbhy5YqwnBIlSnzVuHf8+HFs3LgRy5YtQ/Xq1bF27Vr89ddfOHXq1DfX4/3796hbt26WhFpFRSVL/ydlJhaLoaqqitq1a3/zEcDZbchUWKIbEhKC8ePH47///kP58uUBAE+ePEGZMmVgYmKCLVu2IDk5WVi5e/fuwcTEJMfLkUgkSp3oKnNsykwmI4jF2T9A5XR+xhhj+UMikQitt5kTNJFIBA+PfxASEpnvMVSpUhaTJvXM8fu8vLzw6dMnnDp1CiVLlvz/z6qCxYsXIywsDNu2bcPQoUMBAGXKlMnSQDdjxgwcPnwYFy9exJAhQ6ClpQUA0NbWBoCfNubFxcVh2rRpaNeuHQBgxIgR6NmzJ6Kjo6Gjo/PV/IGBgahVq1auGgmVgXz/+NU8TmGJrqGhIerXr49Zs2bByckJ79+/h7u7O0aPHg1TU1NUrFgRTk5OGDt2LC5evIhHjx7Bzc1NUeEyJSMWi+Dj8woRET8v+9DV1YC1dZ0CiIoxxtivCAmJxJs34YoO45tkMhmOHDmC4cOHC0luZkuXLkXJkiURGfntRF1FJSPlUlVVzdXyBw0aJPx/XFwc9uzZgzp16qBMmTLfnD8wMJDvGkOBia5EIoGXlxdcXV3Rv39/aGhowM7ODvb29hCJRPDy8oKzszOsrKxQvXp1eHp6olKlSooKlymhiIgkhIYmKDoMxhhjv4F3794hKioKjRs3/ubrP2o5TUhIgLe3N9LS0tC6detfiuPgwYNwdnaGmpoaNm/e/N3yi8DAQMhkMnTv3h1xcXFo06YNpk+fLrQk/y4U2hmtfPnyWLt27Tdfq169Onbt2lXAETHGGGOMfS06OhoAUKpUKWHa9evXMW7cOOHvSpUqwdvbGwBgaWkJkUgEIkJSUhLKly8PNze373Yey64WLVrgyJEjOHToEMaOHYsjR46gatWqWeZJS0tDcHAwqlSpgkWLFiE2NhZubm6YNm0a1q1b90vLL2yUYtQFxhhjjDFlJi9XiI2NFaYZGxvj6NGjAIAzZ85g7969wmsbNmxA+fLlIRKJoKmpibJly2Z7WXPmzMHx48eFv0+ePCnc1a5UqRIqVaqEunXr4vbt2zh69OhXT41VVVXFzZs3oa6uLpRKLF68GH369EF4eLjQN+p3wIkuY4wxxthPVK9eHdra2njw4AEaNmwIIGMo0urVqwPAVx3CKlWqhCpVquRqWRMnTsSwYcOEv8uVK4ebN2+iXLly0NPTA5DRWUtPT09oaf7SlyUKtWrVAoDfLtFViiejMcYYY4wpMxUVFfTp0wfbt29HfHz8V6+Hh+ddJzodHR1Ur15d+KeiooKNGzdi27ZtwjxSqRQvXrwQEtjMAgICYGxsjODgYGHa8+fPoaKiIiTmvwtOdBljjDHGssHR0RG6urqwsbHBv//+i+DgYDx69AguLi5YvXp1roZBza6BAwfi8OHDOH78OF6/fo158+YhOTkZvXr1ApDR4S0qKgoAoKenh+rVq8PFxQUvX77E3bt34eLign79+mWpMf4dcOkCY4wxxpRClSrZr2NVxHI0NDSwc+dObN++HV5eXggKCoKamhoaNmyINWvWwMzMDCEhIXkcbYaOHTti3rx5WLt2LT58+AAjIyNs2bIFxYsXBwBs2bIFR44cwYULFyAWi7Fu3TosXLgQgwYNglgsRvfu3TF9+vR8iU2ZcaLLGGOMMYWTyWS5eojDrywvN09iU1NTw4gRIzBixIhvvl6lShX4+/tn+/OaNm2a7fn79u2Lvn37fvM1R0fHLJ3SKlas+N2RrX4nXLrAGGOMMYUryMf/KmJ5TDH4W2aMMcYYY0USJ7qMMcYYY6xI+q0SXZlUpugQGGOMMcZYAfmtOqOJJWL4zF2GiLc/7xFZp3kjdB5tXwBRMcYYY4yx/PBbJboAEPE2BKEvA386X9nquXuaCWOMMcYYUw6/VekCY4wxxhj7fXCiyxhjjDHGiiROdBljjDHGWJH029XoMsYYY4zlVmJiIjZs2IB///0XoaGh0NDQQNOmTeHo6Ig6deogJCQEHTt2zPIedXV11K9fH6NHj0bbtm2xZs2aHz61zM3NDVZWVt993c/PDzY2Njh79iyqVOE+RT/CiS5jjDHGFE4mI4jFIqVeXkJCAgYOHIjExETMnDkTBgYGiI6Oxu7du2FjY4OjR49CJMr4zAMHDqBixYoAgOTkZGzfvh3jxo3DqVOn4ODgABsbGwDAgwcP4OjoiKtXrwrLKVGixHdjSEtLw+zZsyGT8ZCp2cGJLmOMMcYUTiwWwcfnFSIikvJ9Wbq6GrC2rpPj93l6euLTp084deoUSpYsCQCoXLky3Nzc8OHDB2zbtg1Dhw4FAJQpUwa6urrCe2fMmIHDhw/jwoULGDJkCIoXLw4AKFWq1P/HpIvs2LRpE7S0tHIc+++KE13GGGOMKYWIiCSEhiYoOoxvkslkOHLkCIYPHy4kuZktXboUJUuWRGRk5Dffr6KSkXKpqqrmOoY3b95g9+7d8PT0hLW1da4/53fCiS5jjDHG2E+8e/cOUVFRaNy48TdfL1eu3Hffm5CQAG9vb6SlpaF169a5Wj4RYc6cOXB0dISOjk6uPuN3pNBENzw8HAsXLsTNmzehrq4OCwsLTJ48Gerq6liwYAF27tyZZX4XFxfY2toqKNrCQyaTQSzO/oAaOZ2fMcYY+91ER0cD+F+pAQBcv34d48aNE/6uVKkSvL29AQCWlpYQiUQgIiQlJaF8+fJwc3NDtWrVcrX8gwcPIi0tDdbW1nj//v0vrMnvRWGJLhFhwoQJKFmyJHbv3o3Pnz9j1qxZEIvFmDFjBgIDAzFlyhT07t1beA/XpGSPWCyGh8c/CAn59u2TzKpUKYtJk3oWQFSMMcZY4SUvV4iNjRWmGRsb4+jRowCAM2fOYO/evcJrGzZsQPny5SESiaCpqYmyZctme1lz5szB8ePHhb9PnjyJlStXYtu2bUJnN5Y9Ckt0X79+jYcPH+LatWvClz9hwgQsWbJESHSHDRuW7eJsllVISCTevAlXdBiMMcZYkVC9enVoa2vjwYMHaNiwIQBAQ0MD1atXB4CvygkqVaqU66G/Jk6ciGHDhgl/37p1C9HR0ejfvz+AjMZCIKPVePTo0Rg9enSulvM7UFiiq6uri02bNn11hRMfH4/4+HiEh4ejRo0aigmOMcYYYywTFRUV9OnTB9u3b0efPn2+usscHp53jUs6OjpZEmcdHR00atQoy7Ls7OywYcMG/PHHH3m23KJIYYluyZIlsxRky2Qy7Nq1C82aNUNgYCBEIhHWr1+Py5cvQ1tbG0OHDs1SxpBdUqlU+H+JRJInseelzPHlldysZ37EkZ9+h3VkjLGiSCqVgoiEf3KKuCWfefnZMX78eNy9exc2NjYYN24cGjRogKioKBw4cACHDh2CpaWl8Jlfrt/PYvjRvMWLFxeGIwMg9KupWLEiSpUqleP1KAzk208qlX51/s7J+VxpRl1wd3fHs2fPcPDgQTx9+hQikQh6enqwtbXFnTt34OLiAi0tLXTq1ClHn/v48WMAGbcX6tWrlx+h/xJ/f38kJeXdmIG5Xc+8jiM//Q7ryBhjRZmKigqSkpKEhx6IxWJoaGhAV1ejQJYvX05ycnKOH7ywYcMGYYiv4OBgqKqqokGDBnB3d0f79u0RGhoqfHZiYuJPPy8lJQUAsjWvXHJyco6WURilpKQgLS0NL168+KXPUYpE193dHdu3b8fKlSvxxx9/oE6dOmjfvj20tbUBAAYGBnj79i327t2b40TX0NBQKVty5fT19RUdAgDliSM//Q7ryBhjyi45ORlBQUHQ0NBAsWLFhOkyGeXqIQ65JZNRluXnxNixYzF27Nhvvla7du0cJWdt2rTJcTKX02UURmKxGKqqqqhdu/ZX35NUKhUaMn9G4Ymuq6sr9u7dC3d3d3Tp0gVAxu0LeZIrp6enh5s3b+b48yUSiVInusoSm7LEkZ9+h3VkjDFlJ5FIIBKJhH9yBV25UJCPG2Y5J98/fjWPU+jgqWvXrsW+ffuwYsUKdOvWTZi+atUqDBkyJMu8L168gJ6eXgFHyBhjjDHGCiuFJbqBgYHw8vLCiBEjYGJigoiICOFf+/btcefOHWzevBnv3r3Dnj17cPToUTg4OCgqXMYYY4wxVsgorHTh/PnzkEqlWLduHdatW5flNX9/f6xatQqrV6/GqlWrULlyZSxfvhzGxsYKipYxxhhjjBU2Ckt0R44ciZEjR373dTMzM5iZmRVgRIwxxhhjrChRaI0uY4wxxn5PRXHsV5Z38mr/4ESXMcYYYwVGVVUVQM7GjWW/H/n+Id9fckvhw4sxxhhj7PchkUigra2Njx8/AgA0NTUV8lQ0ppyICImJifj48SO0tbV/eWhQTnQZY4wxVqAqVKgAAEKyy9iXtLW1hf3kV3CiyxhjjLECJRKJULFiRZQrVw5paWmKDocpGVVV1Tx7yBMnuowxxhhTCGV/eikr/LgzGmOMMcYYK5I40WWMMcYYY0USJ7qMMcYYY6xI4kSXMcYYY4wVSZzoMsYYY4yxIokTXcYYY4wxViRxossYY4wxxookTnR/c9raxSGTUY7ek9P5GWOMMcYUgR8Y8ZsrXrwYxGIRfHxeISIi6afz6+pqwNq6TgFExhhjjDH2azjRZQCAiIgkhIYmKDoMxhhjjLE8w6ULjDHGGGOsSOJElzHGGGOMFUmc6DLGGGOMsSJJoYlueHg4JkyYAFNTU7Ru3Rpubm5ISUkBAAQHB2PIkCEwMjKChYUFrl69qshQGWOMMcZYIfNLiS4RISoqCjExMbl674QJE5CUlITdu3dj5cqVuHjxIjw8PEBEGDduHMqWLYtDhw6hZ8+eGD9+PEJDQ38lXMYYY4wx9hvJ8agLHz58wL59+3DlyhX4+/tDJpMBACQSCerVq4c2bdrAysoKlSpV+uHnvH79Gg8fPsS1a9dQtmxZAMCECROwZMkStGnTBsHBwdi3bx80NTVRq1Yt3LhxA4cOHYKjo2MuVpMxxhhjjP1usp3oRkVFwd3dHRcuXECLFi0wYMAA1K5dG9ra2pDJZIiOjoa/vz/u37+Pnj17on379pgxYwZ0dHS++Xm6urrYtGmTkOTKxcfHw8/PD/Xq1YOmpqYw3cTEBA8fPszdWjLGGGOMsd9OthPd4cOHw87ODn///TdUVVW/OU/jxo0xaNAgJCUl4dixYxg2bBiOHj36zXlLliyJ1q1bC3/LZDLs2rULzZo1Q0REBMqVK5dlfh0dHYSFhWU3XMYYY4wx9pvLdqK7f//+7ya4X9LQ0ED//v1hZWWV7UDc3d3x7NkzHDx4ENu2bYOamlqW19XU1JCamprtz5OTSqXC/0skkhy/P79lji+vFMR65kfcOZGbdVR0zIwxxhj7dTk5n2c70f1RkhsfHw8A0NLSyvZ7MnN3d8f27duxcuVK/PHHH1BXV/+qg1tqaiqKFSuW3XAFjx8/BpCRfNerVy/H789v/v7+SEr6+aN3s6ug1jOv486J3K6jImNmjDHGWMH7pUcAP3/+HNOnT8erV68gEomgp6cHNzc3NGzYMNuf4erqir1798Ld3R1dunQBAJQvXx4BAQFZ5ouMjPyqnCE7DA0NlbIlV05fX1/RIeRKYYy7MMbMGGOMsaykUqnQkPkzv5ToOjs7Y8KECWjTpg3S0tLg4+OD6dOn499//83W+9euXYt9+/ZhxYoVMDc3F6b/+eef2LBhA5KTk4VW3Hv37sHExCTHMUokEqVOdJU5th8pjHEXxpgZY4wxlnvZHkd3/PjxePHiRZZpCQkJqFq1KtTV1aGlpYUqVaoIZQw/ExgYCC8vL4wYMQImJiaIiIgQ/pmamqJixYpwcnLCq1evsGHDBjx69Ah9+/bN2doxlo/kQ+vl1/yMMcYY+zXZbtHt27cv5syZg3LlymH8+PEwMDDAtGnTMHjwYKiqqkImkyE9PR3z58/P1uedP38eUqkU69atw7p167K85u/vDy8vLzg7O8PKygrVq1eHp6fnT8fmZawgicVieHj8g5CQyJ/OW6VKWUya1LMAomKMMcaYXLYT3Xbt2qFdu3b477//MG/ePOjo6GD8+PG4fPkyAgMDIRaLUb16dWhoaGTr80aOHImRI0d+9/Xq1atj165d2Q2PMYUICYnEmzfhig6DMcYYY9+Q4xrdtm3bom3btrhy5Qr+/vtvlC5dGhMmTICBgUF+xMcYY4wxxliuZDvRTU1NhY+PD16/fo1KlSqhf//+2Lt3L65duwZXV1eUKlUK48ePV8ohvBhjjDHG2O8n24nu1KlTERYWhmbNmuHWrVs4deoUDh8+jJYtW6Jly5a4ceMG3NzcoKWl9VXNLWOMMcYYYwUt24nutWvXcOjQIdSoUQPp6elo2LAhoqKiUKZMGQBA8+bN0bx5c9y6dSvfgmWMMcYYYyy7sp3oNmnSBDNnzkTTpk3h7++PWrVqCUluZk2bNs3TABljjDHGGMuNbI+ju2zZMlhYWODz589o1KgRduzY8dP3pKen/1JwjDHGGGOM5Va2E117e3uUKFECzs7OGDlyJEqXLv3deVNSUnDgwAH069cvT4JkjDHGGGMsp7JdurBp0ya4u7tj8eLFaNWqFVq0aIFatWqhdOnSkEqliImJgb+/P+7du4fLly+jbdu22LBhQ37GzhhjjDHG2HdlO9EtU6YM3NzcEBISAh8fH+zatQv+/v7CY00lEgn09fXRunVrHDp0CFWrVs23oBljjDHGGPuZHD8wokqVKpg8eTImT54MqVSKz58/AwBKly4NkUiU5wEyxhhjjDGWGzlOdDOTSCTfHHmBMcYYY4wxRct2ZzTGGGOMMcYKE050GWOMMcZYkcSJLmOMMcYYK5I40WWMMcYYY0USJ7qMMcYYY6xI4kSXMcYYY4wVSZzoMsYYY4yxIokTXcZYkSOTUb7OzxhjrHD4pQdGMMaYMhKLRfDxeYWIiKSfzqurqwFr6zoFEBVjjLGCphQtuqmpqbC0tMStW7eEaQsWLIC+vn6Wf7t27VJglIyxwiQiIgmhoQk//ZedZJgxxljhpPAW3ZSUFEyZMgWvXr3KMj0wMBBTpkxB7969hWlaWloFHR5jjDHGGCukFNqiGxAQAGtra7x79+6r1wIDA1GvXj3o6uoK/zQ0NBQQJWOMMcYYK4wUmujevn0bTZs2xf79+7NMj4+PR3h4OGrUqKGYwBhjjDHGWKGn0NKFgQMHfnN6YGAgRCIR1q9fj8uXL0NbWxtDhw7NUsaQXVKpVPh/iUSS61jzS+b48kpBrGd+xJ0TuVnHvI45JzFoaxeHTEYQi0XZfo9MRiCS5Sa0354y7B+MMcbyR06O1wqv0f2W169fQyQSQU9PD7a2trhz5w5cXFygpaWFTp065eizHj9+DADQ0NBAvXr18iPcX+Lv74+kpLzrDFNQ65nXcedEbtcxL2POaQzFixfL1UgAz54pbjsXVsqwfzDGGFMOSpno9urVC+3bt4e2tjYAwMDAAG/fvsXevXtznOgaGhoqZUuunL6+vqJDyJXCGLcyxCwfCSC7lCHm3wVva8YYKxykUqnQkPkzSpnoikQiIcmV09PTw82bN3P8WRKJRKkTXWWO7UcKY9wcM/sR3taMMVb0KMU4ul9atWoVhgwZkmXaixcvoKenp5iAGGOMMcZYoaOUiW779u1x584dbN68Ge/evcOePXtw9OhRODg4KDo0xhhjjDFWSChlotuwYUOsWrUK//zzDywtLbFz504sX74cxsbGig6NMcYYY4wVEkpTo+vv75/lbzMzM5iZmSkoGsYYY4wxVtgpZYsuy0om5bFUGWOMMcZySmladNn3iSVi+Mxdhoi3IT+dt07zRug82r4AomKMMcYYU26c6BYSEW9DEPoy8Kfzla1epQCiYYwxxhhTfly6wNhvRCbLWRlMTudnjDHGlAm36DL2GxGLxfDw+AchIZE/nbdKlbKYNKlnAUTFGGOM5Q9OdBn7zYSEROLNm3BFh8GYQshkBLFYlG/zM8aUCye6jDHGfhtisQg+Pq8QEZH003l1dTVgbV2nAKJijOUXTnQZY4z9ViIikhAamqDoMBhjBYA7ozHGGGOMsSKJE13GGGOMMVYkcaLLGGOMMcaKJE50GWOMMcZYkcSJLmOMMcYYK5I40WWMsXzAT6FjjDHF4+HFGGPfpK1dnAfX/wW/w1PoeP9gjCk7TnQZY99UvHgxHlz/FxX1p9Dx/sEYU3ac6DLGfogH189/hbn1nPcPxpgy40SXMcYUjFvPGWMsf3CiyxhjSoJbRxljLG8pxagLqampsLS0xK1bt4RpwcHBGDJkCIyMjGBhYYGrV68qMELGGGOMMVbYKDzRTUlJweTJk/Hq1SthGhFh3LhxKFu2LA4dOoSePXti/PjxCA0NVWCkjDHGGGOsMFFo6UJAQACmTJkCIsoy/ebNmwgODsa+ffugqamJWrVq4caNGzh06BAcHR0VFC1jjDHGGCtMFNqie/v2bTRt2hT79+/PMt3Pzw/16tWDpqamMM3ExAQPHz4s4AgZY4wxxlhhpdAW3YEDB35zekREBMqVK5dlmo6ODsLCwgoiLMYYY0ypFNbh5xhTNKUcdSEpKQlqampZpqmpqSE1NTXHnyWVSoX/l0gkvxxbXssc3/cU1rjzU262SV7HXBDfC8ecO7x/FAxl2M45VRhjBjLizunwc4qOWyQS5zg5J+JHYbOfy8m+rZSJrrq6OmJiYrJMS01NRbFixXL8WY8fPwYAaGhooF69enkRXp7y9/dHUtL3D1yFNe6cUFVVRb169aGikr+JQV7GXFDfy+8ec27kdj15W+eMMmznnCqMMQP/izunw88pw7bOaXL+7JlitzUrepQy0S1fvjwCAgKyTIuMjPyqnCE7DA0NlbJFVE5fX1/RIeRKXsctkUjg4fEPQkIifzqvsXEtDBrULsfLKIzbmmMuOIUxbo65YBTGmAHliDunybkyxMyUn1QqFRoyf0YpE90///wTGzZsQHJystCKe+/ePZiYmOT4syQSiVInusoc24/kR9whIZF48yb8p/NVrqyTq88vjNuaYy44hTFujrlgFMaYgcIZd2GMmSk3hY+j+y2mpqaoWLEinJyc8OrVK2zYsAGPHj1C3759FR0aY4wxxhgrJJQy0ZVIJPDy8kJERASsrKxw7NgxeHp6olKlSooOjRVhMil3gmCMMcaKEqUpXfD398/yd/Xq1bFr1y4FRcN+R2KJGD5zlyHibchP563TvBE6j7YvgKgYY4wxlltKk+gypgwi3oYg9GXgT+crW71KAUTDGGOMsV+hlKULjDHGCp5MxuU7jLGihVt0GWOMAQDEYnG+D/PHGGMFiRNdxhhjgvwe5o8xxgoSly6wfMEjGBQM3s6MMcbY93GLLssXPIJBweDtzBhjjH0fJ7os3/AIBgWDtzNjjDH2bVy6wBhTejwaAGOMsdzgFl3GmNLj0QAYY4zlBie6jLFCgUcDYIwxllNcusAYK3CFcbSIwhgzY4z97rhFlzFW4ArjaBGFMWbGGPvdcaLLGFOIwjhaRGGMmTHGfmdcusAYY4wxxookTnQZY4yxX5TTGm6u+VZOMhnl6/ys4HHpAmOMMfaLclLDrVujCqznTy2AqFhOicUi+Pi8QkRE0k/n1dXVgLV1nQKIiv0KTnQZY4yxPJDdGu7CSiaTQSwu+jeCIyKSEBqaoOgwWB7hRJcxxhhjP8UPbmGFESe6jDFWRMmkMoglRb8FjhUcfnALK2w40WWMsSLqdxj793e5nc4Yyx2lTnTPnj2L8ePHZ5nWpUsXrF69WkERMcZY4VLUx/7l2+mMsR9R6kQ3ICAA7du3h6urqzBNXV1dgRExxhhTNnw7nTH2PUqd6AYGBuKPP/6Arq6uokNhjDHGGGOFjFIXNgUGBqJGjRqKDoMxxhhjjBVCStuiS0R48+YNrl69Cm9vb0ilUpibm2PChAlQU1PL9udIpVLh/yUSSX6E+ksyx/c9hTFujjlvFMWYgcIZN8ecN7Kzf+REQaxjfu3TOdkW+f35+RVDTuV1zDmlDNuZ/VxOtrnSJrqhoaFISkqCmpoaPDw8EBISggULFiA5ORmzZ8/O9uc8fvwYAKChoYF69erlV7i55u/vj6Sk7z+BpTDGzTHnnaIWM1A44+aY887P9g9VVVWoqGTv1FSsWDHo6enlVWjflV/7dFBQEJKTk386X27X82dx50RB7U95GXNO5XYdFRkz+zmlTXQrV66MW7duoVSpUhCJRKhbty5kMhmmTZsGJyenbF91GRoaKmWrhpy+vr6iQ8iVwhg3x1wwCmPMQOGMuyjGLAIgVrJjdl5vZ60y2pDJZPmepBfF/UMZFcaYCzupVCo0ZP6M0ia6AKCtrZ3l71q1aiElJQWfP39GmTJlsvUZEolEqRNdZY7tRwpj3BxzwSiMMQOFM+6iGrOyjf2b19u5WAmtAhkWrajuH8qmMMb8O1HaRPfKlSuYOnUqLl26BA0NDQDA8+fPoa2tne0klzHGWOFT1Mf+leNh0VhekMkIYrEo3+Yv7JQ20TU2Noa6ujpmz56NcePGITg4GEuXLsXw4cMVHRpjjDHGmFIQi0Xw8XmFiIif1wnr6mrA2rpOAUSlPJQ20dXS0sLmzZuxaNEi9OnTB8WLF4eNjQ0nuowxxlgekEllEEuUepRRlk0REUkIDU1QdBhKSWkTXQCoU6cOtm7dqugwGGOMsSJHLBErXT00Y3lNqRNdxhhjjOWf36Uemv2++J4FY4wxxhgrUDIZ5ev8ctyiyxhjjDHGClRBdaLjRJcxxhhjjBW4guhEx6ULjDHGGCuSZDKZokNgCsYtuowxxhgrkgriCXRMuXGiyxhjjLEii59A93vj0gXGGGOMMSXB5RZ5i1t0GWOMMcaUBJdb5C1OdBljjDHGlEhhLLeQyWQQi5WvUIATXcYYY4wx9kuUtSWaE13GGGOMMfbLlLElWvnamBljjDHGGMsDnOgyxhhjjLEiiRNdxhhjjDEmkEmLzhBnXKPLGGOMMcYEYokYPnOXIeJtSLbmr9O8ETqPts/nqHKHE13GGGOMMZZFxNsQhL4MzNa8ZatXyedoco9LFxhjjDHGWJHEiS5jjDHGCoXCWDtaGGMuSpS6dCElJQXz58/HmTNnUKxYMTg4OMDBwUHRYTHGGGNMAXJSO6osdaOFMeaiRKkT3aVLl+LJkyfYvn07QkNDMWPGDFSqVAnm5uaKDo0xxhhjCpDd2lFlqhstjDEXFUqb6CYmJuLAgQPYuHEj6tevj/r16+PVq1fYvXs3J7qMMcYYY+ynlLZG98WLF0hPT4exsbEwzcTEBH5+fpDJuN6FMcYYY4z9mNImuhEREShdujTU1NSEaWXLlkVKSgpiYmIUFxhjjDHGGCsUlLZ0ISkpKUuSC0D4OzU19afvJyJhXolEAgCQSCQoV7s6JGqqP31/mSoVIJVKUb26LlRUfn49UKGCNqRSKcqVKwaJhH46v45OMUilUkil0p/OWxjj5pg55qIWN8fMMRe1uDlmjjkvYlZE3PL/ynO9HxFRduZSAF9fXyxYsADXrl0TpgUGBsLCwgK3bt2Ctrb2D9+fmpqKx48f53OUjDHGGGNMEQwNDb9qFP2S0rboli9fHtHR0UhPT4eKSkaYERERKFasGEqWLPnT96uoqMDQ0BBisRgikSi/w2WMMcYYYwWAiCCTyYT88EeUNtGtW7cuVFRU8PDhQzRu3BgAcO/ePSF5/RmxWPzTLJ8xxhhjjBVdStsZTUNDA7169cK8efPw6NEjnDt3Dlu2bIG9PQ+kzBhjjDHGfk5pa3SBjA5p8+bNw5kzZ6ClpYVhw4ZhyJAhig6LMcYYY4wVAkqd6DLGGGOMMZZbSlu6wBhjjDHG2K/gRJcxxhhjjBVJnOgyxhhjjLEiiRNdxhhjjDFWJHGiyxhjjDHGiiROdBljjDHGWJHEiW4Bk8lkig4hW6RSqaJDKPQKy3f9u/pyZEUeabFg8Hb+Nbz9CodvnUPz87v71mfzvpKBE90ClJ6eLjy+OCoqSsHR/JhEIkFoaCguXLgAoPAnbQV90JFKpcJ3/eDBA7x48QJv377Nt+UVhKJ00JRKpRCJRACAU6dO4eHDhxCJREqxn38ZQ1Hd7pkpwzoW9DEiNzJvP26MUF5SqRQSiQQAcOXKFdy4cQNv3rz55r6fV8uTf3Z4eDgSEhKQkpKSb8vLL986/ubFMVnllz+BZQsRQUVFBUSEiRMnIiQkBDo6Opg/fz4qVaqk6PC+6fDhw9i7dy86dOggJG2FkUwmEw46N2/eRHJyMlq0aAE1NTUAGd9NXh4QiEhY3owZM3Dt2jUUL14c5ubmGD9+PFRUVArNASjztiksMf9M5pPQuXPnsG/fPqipqWHp0qUoU6YMZDJZge/v4eHhKF++PNLS0qCqqgoAiI+Ph5aWVpHZ7pl/hxs2bEBISAiSkpLg7u6u8HXMvE88ffoURIQ//vhD+K0qYp/4kvy48v79eyxduhRxcXHo0KEDbG1tFRpXXsn8HRR28vUYNWoUnj17hpIlS6Jdu3aYMmUKxGJxnp9z5Mtbu3Ytjh8/jmLFiqFnz55wcHDIs2XkJ/n2EIvF8Pf3x71796Cqqgpzc3OUKFHilz+fE90CIhKJEBsbC3t7e5QpUwYTJkxAaGgodHR0FB0agG8fZKytrXH16lWcPHkS3bp1U1BkvybzCWrq1Km4dOkSpFIp6tevj+nTp6Nhw4Z5csCJi4tDfHw8KlSoIHze8uXL8ezZM+zduxepqakoXbo0VFVVhVaivD7Y5bXM8f3333+4fv06VFRUYGxsDDMzMwVHl3sSiQQJCQkYPHgwVFVVERsbi6ioKLi5uWHRokXCd1RQ382pU6dw6NAhTJgwAX/++Seio6Mxbdo0pKWloUaNGpg1axbU1dULJJb8JP8djhs3Dk+ePIGFhQVUVFTw8eNHlCtXTqGxyY99o0ePxvv37xEVFQUDAwMYGxtj3LhxCklyiQhpaWlQU1MTjmOPHj2Cg4MD2rZti0aNGhX4vppf5Oef9PR0BAQEIDU1FQ0bNhReL2zrKJPJsHz5cnz+/BlHjhyBWCyGWCzO10YDDw8PHD58GHPmzIFMJkODBg3yfBn5Rb49jh49ivnz56NOnTooXbo06tWrh/r16wP4tX2AE90CFBQUBC0tLSxevBjlypXDkydPcP78eURERKBBgwYwMTEp0HgSEhIQExODypUrCwf6Fy9eoHLlyihRogTU1dVRuXJlvHr1CkDhvOIWi8WQyWQ4evQoYmJisHfvXhARZs6ciU2bNsHR0RF16tT55eWcOXMGDx8+hKurKxISElC8eHHExsaicePGqFq1Km7fvo0TJ07g/v370NDQwNKlS1GyZEmlPoDL49q0aRPWrVuHLl26QEtLC46Ojhg5ciQcHBxQqlQpBUeZOxs2bICqqiq8vb0BAGfPnoWvry/c3d0xa9YsiESiAvtu1NXVIRKJsH//fojFYri7u0NdXR0mJibYunUrEhISMHfu3Dxp2VCEzMeNO3fuICwsDMeOHUOpUqUQFRWF2NhY3LhxA40aNVJYQk9EWLBgAaKjo7F+/XpUrlwZo0ePxr59+2BmZgYDA4MCj+nNmze4fv06bG1thUT7/Pnz6N69O+bOnQsACAwMxPXr16GhoYFatWoV2t+jRCJBUFAQRo0ahdKlS+PZs2ewsLBA48aN0adPH6U9Rsp9eawQi8X48OED6tati7Jly+LcuXO4efMmbt++jbJly2LevHmoVq1anixT/t9Xr15hyJAhMDMzw5s3b3D79m28ePEC5cqVKxQtuy9fvsT69euxZMkSdO7cGVFRUdDQ0EBYWFiWBqTc4EQ3n3wrKYyNjcWTJ09w7NgxPHr0CIGBgfj8+TNKliwJX19fzJs3r0APqG/fvoWLiwu2bduGkiVLYu3atdi+fTu6dOmCYcOGoWbNmujatSucnJxgZ2enNK3POZGcnIxNmzbh+PHjaNmypZDULlu2DGPGjMHevXsxfPjwXy4fqVevHpydnZGQkIDg4GC4uLigZMmS+Pfff9GvXz8EBQWhdu3a0NPTg5+fH2bOnAkvLy+lP4CHh4fD19cXS5cuRceOHZGamoq9e/ciPT290Fz0fHkSkkql8Pf3R8OGDVGyZEkAQLdu3aCqqopdu3Zh165dsLW1zdfvRiaT4f79+2jcuDE6duyIuLg4nDx5Elu2bEHp0qWxbNkyqKqqwtLSEj169EClSpUwbty4QtWyK9/uEokEjx49wqtXr1CrVi08ffoUr169gr+/P06fPo3AwEB8+vQJvXr1wpw5c6CpqVlgscmlpqbC398fDg4OqFy5MrZt24Zbt25h8eLFkEgkePHiRYEnu/7+/jh16hS0tLQQEREBGxsbfPz4EX5+fnj48CE2bdqEd+/e4fXr19DV1YWtrS2GDBlSaH6XmcXGxmLq1KkwNTWFi4sLwsPDYW9vj/DwcHTs2BHa2tqKDvG7Mp/rU1NToaqqCpFIhDp16mD79u24ffs2IiIiUL9+fXTu3Bm+vr5YtmwZVq9e/UvLFYlEiI6OhqenJ7p3747ixYvj5MmTCA0NxaVLl6Curg4tLS08f/4cAJQu2f3yNygv2SpVqhQiIiKwadMm+Pn5ITAwEGZmZnBzc8v1sgpv4aUSy7zjHzx4EN7e3rh06RJatmyJrl274siRIwgKCsKgQYOwY8cOLFmyBNHR0UhJSSnQOA0MDKCtrY3+/fujS5cuMDQ0hLOzM2JjY9G/f38cOXIEtWvXRrdu3XDo0KECjS23vixcL1asGKpWrQoVFRXhBw8Aenp6mD17Ni5cuIATJ07kunOgvENI3bp10b9/f5w6dQqVK1dGw4YN0atXLwwcOBCVKlWCm5sbli5digULFqB3795QUVFRio5PX/qy841UKgURoVmzZrh27Rpat24t1H55eXkpfafK9PT0rxJWiUSC8uXL4+nTp8K0YsWKoX379ihWrBi2bNkidMLML2KxGHv27MGYMWMAAE2aNEGTJk3w5MkT+Pv7Cwd9PT09eHh4YMOGDTh27Fih6oAk3+4vXrzA2LFjERAQgMqVK6NDhw4YMmQIli1bhjp16mD+/PnYv38/jh49ioCAgHyP68sOcUQEdXV16Orq4uPHj3B2dsbmzZuxevVqdOnSBevXr8f169fzPa4vmZmZQV9fH25ubli/fj3U1dUxbNgwpKamYvDgwYiMjETv3r1x+fJldOzYEb6+vkrXee57vozz06dP+Pz5M5ycnKCqqoqVK1dCRUUFM2bMwOvXrxEZGamgSH8uc9352LFj4ebmhlevXmHw4MGYNGkSTExM4OnpiaVLl2L8+PHo168fiAjp6ek5XtaX54yPHz/i0qVLSExMhLW1NapUqYKLFy+iX79+wu+qb9++ePLkiVKdb+S/wcwxaWpqQldXF9OnT0fr1q3x6NEj6Ovrw9XVFUeOHMG5c+dyvTxu0c0H8h1/xIgRQrlCfHw82rVrBzc3N8THx0NdXV04mQUEBGSp3ynIOAcOHIjx48dDR0cHbdu2BQD07NkTq1atwpEjR3Dw4EGUKVMGxYsXB6DctVKZLzCePHmC9PR0VK5cGT179gQAeHt7Y/Xq1ZgwYQIAoHXr1hg9ejTc3NxQsWJFWFpa5mjdMneuWbt2LR4/fox+/frh2LFjuHbtGlq2bIlatWoByGgZFYlESElJwcmTJ2FsbKzwzi1fkn+3cXFxICKULFkSYrEYHz9+xKJFi3D27FkMHz4cI0eOBBHB19cXlSpVUtrOMPIOoEBG/VpycjKqV6+OAQMGwNzcHB4eHtiyZYvQ0lGiRAmUK1cOiYmJ2LVrFypXrgx9ff08jenOnTuIiopCly5dMGTIEAwePBht27aFgYEBlixZguTkZJw9exb79u2DjY0NgIyEZ8qUKXB1dUXp0qULVX20r68vjh49ij///BMzZswAAKxcuRJPnz6Fnp6e0FIXGRkJQ0PDfG+xztxR1NvbG8HBwYiJiYGzszN0dXWxfv16lChRAnv27EHVqlUBZLQ2BgUF5WtcX8YovyUdGhoKsVgMY2NjvHz5Eg0aNMChQ4cQExODKlWqCOtSvXp1fPz4EWlpacI+r6zS09O/ilFXVxe6urq4fPkyNmzYADU1NWzfvh1isRjz58/HlClTULZsWQVF/G2ZzzdLlizB8ePH0aZNGzx48ADXr1/HunXrYGNjA6lUisDAQMhkMoSFhcHHxwcdOnTI1fckP2e8efNGOD5ZWFhg0aJFOHbsGBo3bgyZTIbU1FQUK1YMAPDq1SuYmpoq1flGIpHg+fPn2LFjB1JSUlC1alUMHjwY7u7uuHz5MsqWLYsGDRqgXLlySEtLQ6NGjYQcJDeU+xdRiO3YsQPh4eE4ceIE1NTUEB4ejoiICCQnJ6Nq1arYsmULHj58iLJly8LX1xddu3bNUnyfX77sPVyyZEk4ODjgwIEDWLBgAWbPng2RSIRJkybh0aNHOH/+vFDH2LlzZ/z555/5HmNOyQ848oPOnDlzcO3aNaSkpKB06dJo1aoVpk+fjrCwMJw5cwaVKlVC3759AQA2NjZIS0tDmzZtsp3kZu4hCmRcyZ89exbz58+HkZEREhISMGvWLHh7e8PAwAAvXrzAxo0b8eDBA6ipqaFChQqYPXt2/myMXyASiXDu3DmsWrUKRITBgwejX79+GDFiBNzc3DBnzhwMHDgQQEZioqWlhT/++EPBUX+bfJ+QyWQYPHgwPn78iNKlS+PcuXN4/fo1nJ2dcePGDZw+fRqpqamwsrLCmzdv8OrVK/Tu3Ru3bt3CkydP8jTRjYmJwZYtW5CSkoKKFSuiYcOGwrFh8ODBwt2VyMhInDt3Drq6uujYsSOAjItmPz8/3L9/X6kT3S+PL+XKlROOfY8ePULDhg2hrq6Ohg0b4saNG7hx4waaNm2KZcuWQVdXV7gwzC/y3/isWbNw/fp1DBw4EJGRkXj58iUGDx4MPz8/VKlSBZ8+fUKVKlXw5s0bhISEoGvXrvka15cxvnz5EiNHjsSAAQMwbtw4eHt7Y+vWrRgzZgxq166N9PR0LFmyBGXLlkVCQgK2bt2Kv//+GxoaGgUWZ27IZDJh9KF58+YhLS0NJUqUgL29PYoXL46ZM2eiRYsW8PT0BJBxkREcHKyULboSiQRJSUnw8fGBTCbD+vXr0aBBA7x48QJr1qzB1KlTsWPHDkRFRWHt2rW4f/8+ihUrBkNDQ0ydOjXXy92zZw/c3NwwfPhwTJw4ESNHjoS/vz8OHjyIfv364fHjx9izZw8iIyMRHx8vdLxVJteuXcO4cePQr18/VK5cGYGBgejQoQPOnj0La2trhIeH48GDBzAyMsLu3bsREhKC8uXL53p5nOjmgW+1cqalpaF8+fKIjo7GnTt3cPHiRdy8eRNlypRB37590a1bNzx//hxJSUmYPHky+vXrl+9xZr6SfvnyJUqUKIF69erB1NQUpqamGDt2LGrUqCG00DVs2BANGzaEoaEhDh06hPPnz6NBgwZKUwO2Z88edOvWLUsHjA0bNuD27dtYv3490tLS8O7dOzg5OUFTUxNDhw7Fp0+fcPDgQZQvXx6tW7cGANjZ2WVreREREdDV1YVIJEJ6ejpkMhkcHR3x8eNH2NnZwcjICACwYsUKWFtbY+7cuVi7di0MDAwwYMAAdO/eHcnJyTA3N8/zbZFbmVslQkND4eHhAXNzc3z48AErV64EEaFz58549OgRvLy8oKmpidKlS8PX1xdSqRTVq1dX8Br8z5gxY+Dg4IAmTZoIwzAFBASgSpUq2Lx5M1JSUnDlyhUsX74cNWvWxF9//YWdO3fCy8sLR48eRUREhFDnePHixTy/ja6trQ0HBwds27YN27dvx/jx4/H333/D398f7u7uMDIyQqNGjTBw4EB4e3vj8OHD0NXVFS6A165dm6fx5DX5vpSamork5GSIRCKYmJhg1qxZWLBgAU6ePIkKFSqgXLlySE9Ph6qqKi5cuIAnT57AyMgIrq6uAPL/rlFYWBgCAgKwYsUKNGrUCEBGbaWamhomTJiAbdu2YeLEiShXrhxCQ0PRokULWFlZ5Vs8X3r69ClWrVoFMzMz2NvbQ0NDAz169ICPjw927dqFiRMnQkdHBzo6Ojh//jyKFy8OLy8vtGrVqsBizC2xWIzPnz/D1tYW2traaNasGZ4+fQqpVAp7e3sEBgZCX18fQUFBqF69Oq5evQp1dXXUrFlT0aF/08ePH4XaUfkFqIGBAYYMGYLly5dj5syZWLlyJZydnfH06VMUL14cTZs2zdEyvuzv07p1a6xevRre3t548+YNpk2bBj09PYSEhAAAqlWrhmbNmuHly5coWbKkUB6lTC5evIhevXrB2dkZUqkUDg4OqFq1qlDOce/ePXh6ekImk0Emk2HDhg3Q09PL/QKJ/ZL09HTh/1NSUkgmkxER0alTp6hly5bUrFkzatiwIU2ePJk2btxIa9euJUtLS4qLiyuwGKVSaZa/R40aRb1796bOnTvToEGD6ObNm0REtGvXLqpXrx5dvnyZiIgSExOF92zfvp2GDh1aYDH/TFBQEI0aNYr8/f2zTJ8+fTqtWLEiy7Tz58+TgYEB3bhxg8LDw2ny5MnUq1cv+vTpU7aWJZPJaMyYMdSzZ0/aunVrlte2bNlCDRo0IGdn5yzfaXJyMrVq1YpGjBhBa9asIW9v79ytaD7KvF/cuXOHDh06RDNmzBCmubu7k4WFBZ09e5ZSUlJozpw51KtXL7K0tCRbW1uKiopSRNjftX79+izf6ZIlS0hfX5969+5Nnz9/JiKimJgY2rp1K7Vu3ZrOnj1LRESfPn2i58+f0/Pnz4mIKC4ujgYOHEiHDh3Ks9gyb2tfX18aOnQozZ8/X4hrxowZ1LhxY4qIiCAiomvXrtGYMWNozJgxFBQUlGdx5Lfnz59Tr169aMCAAWRmZkbbtm2jhIQEOnLkCPXp04c2b95M8fHxwvyJiYkUExMj/J2Wlpan8ciPx5k9evSI9PX16enTp0SUcQyXyWR0//59WrNmDYWEhND169fp8OHDdO7cuTyN50cxymQy+vz5M7m5udGff/5JixcvzjLfnj17aMiQIbRixQqKjo6msLAwIso47xBl7GPfWl9lc+nSJRowYADFxsYSUdbfxvr168nBwYGMjY1p8ODB9Oeff9Lu3bsVFWoWX55H5dv62rVrpK+vT1u3bhW+i7S0NDp37hxZWlqSi4vLLy87NDSUnj17Jnz+mjVryM3NjSwtLWn27Nnk6upKjRs3/up8qAwy50hEGfurnZ0dnT59mhISEqhTp07k4OBAERERtHv3buH7fv/+Pfn5+VFycvIvx8CJ7i/IfFDx8vKiKVOmkJ2dHR05coSSk5Pp+fPndPToUXr37p1wQjt58iTZ2tpSUlJSvsfn4+OT5e+0tDRydHSk/v37U3BwML17945sbGyoSZMmFB0dTURECxcuJENDQ1qzZg25uLgIB9Nr165RmzZt6P379/ke98/ExcWRTCYTfvSZE4G+ffvSzJkzhb+lUilJpVKaPHkyTZs2jYiIbty4Qbdu3cr28j5//kyjRo2iHj16UMuWLcnBwYF8fX2F15cvX06dOnWi06dPZzlRBwYGUv/+/cnMzCxHyysI8oO2VCqlvn37UosWLahZs2bUu3dvSk1NFeabOHEi9enThx49ekRERElJSRQSEqKQmLPr5MmTlJCQQEQZFz6mpqb0/Plz4fcaFhZG7u7uZGZmRnfu3CEiog8fPtDff/9NHh4e1LlzZ7K3t8+zpOHLAz1RRtIyaNAgWr16NUmlUoqNjSVbW1uysLAQ5tm5cydNnz5d+A0qE5lM9tV6BQQEUIsWLWjBggUUHBxMO3fupMaNG9OqVauIKON3YmNjQ4cPH/7mtv0ykfhVmT8vLCyMUlNTKT09nZKSkmjgwIG0du3aLPv63bt3SV9fnwIDA/M0jh/5VmIfEhJCU6ZModatW9Pjx4+zvLZmzRrq06cP6evr09ixY4Xpyprgfmvf9/b2pubNmwt/y4/jBw4coP3799OnT5/o2LFjdPToUXry5EmBxfojmfclHx8fcnd3JxcXF+FiaefOnVS/fn26dOmSMG9CQgLt2rWL+vfvT9HR0Tn6jr6cd+7cudSoUSPavHkzRUdH0/Xr12ny5MkUHBxMq1atolmzZpG+vj7NmDEjzy8Wf4X8+09NTaV3794JSeuiRYuoQ4cOZGpqSvPnzxeO17NmzaJx48bl+bFAeaqTCyH57TUnJyfs2bMHzZs3h4mJidCbuk6dOujevTtCQkJw//59PH36FOvXr0elSpXyvdPFgwcPsGXLFsTFxQk9tWUymTCMS5UqVXDixAm8fv0ac+bMwfv37wFk1K5ZW1vjxIkTqFOnDsqXL4+PHz/i5MmTKFGihDAck6IEBQXh2LFjSE9Ph5qaGoKDgzFixAisXLkSAISSkKtXrwKAMFB3Wlqa8BnNmjWDqalptpdZsmRJNGrUCGlpadi+fTtSU1OxcOFCODo64v3795g8eTKMjY2xadMmPHz4UHifnp4evL29cfTo0RwtryCIxWJER0fj2LFj0NPTw5YtWzBx4kT4+/vjn3/+EeZzd3eHhoYGli1bhidPnqBYsWKoXLmyAiP/GmXqwZ2cnIxFixZh5MiRACDUMS5evFio8ytfvjz69OmDWrVqYd++fQAyevzKh3Hq2bMntm/fnie3zzMPw3bkyBEcPnwYERERGDBgAFq2bIlbt25h7969KFGiBBYtWiTUCw8fPhyVK1eGq6vrL9Wm5Zdbt25l2deBjA4y+vr6cHZ2RpUqVXD16lVUrVoVFhYWeP/+Pf766y/UqFEDu3btwo0bN776zLzsLENEwuetXbsWAwYMgL29PTZv3oxixYrB1NQUt2/fxsmTJ4X3lChR4pc7veSEvF41JiYGCxcuxOzZs7Ft2zaoqKhg8uTJaNCgATw8PBARESG8Z9iwYXBycsLixYuFOlZAOZ9amHnfDwsLE47BTZo0QYUKFXDmzBkAEJ5QGRgYiC1btqBUqVLo3r07evbsKTwsQNHk+5KLiwtWrVqFpKQkJCYmwsHBAZs3b4atrS169OgBZ2dn+Pv7A8g4pvTo0QM7duyAtrZ2tr+jzKOCxMTEAADmzZuHWbNm4datW5g4cSIaNmyIqKgo7NmzBxMmTEDv3r1RtmxZvHz5EklJSXm/AXJJIpEgMDAQ5ubmGDVqFAYMGIDY2Fgh3mrVqmUZUjA5ORkVKlTI+45zeZo2/4Y+ffpEgwYNEq48r169Sg0bNqRNmzYJt1Fnz55NXbt2JTMzsyytjflJ3pJJREKLUHh4OLVr144ePXpErq6u1KpVK7p+/TrJZDJydHSkbdu2Ce/PfDuRiOjWrVtZShkU5fHjx6Svr08+Pj60ZcsWOnbsGG3atInat29PFy9epE+fPpGjoyMNHz5c+E6io6PJxsaGNm7cSEQ5a/2Qz5ucnEzNmzen/fv3U3JyMu3atYvatWtHbdu2pbVr11JsbCz17duXpk2bJlzlK7PY2FiaO3cumZqa0po1a4TpmzZtovr162dpgQ4ODqYuXbqQn5+fIkL9oW+1GAUEBFDz5s2FW4aRkZHUuHHjr8pL3r1799V787olgSij1c7GxobatWtHJiYm5ODgQA8ePKDk5GRydXUle3t74Q5BQEAATZgwgf766688jyMvbdmyherVq0ehoaF05MgRioyMpFOnTpGdnR29fPmSLCwsyM7OjiIiIujatWs0efJkIvpfa3pe3I7MjgsXLlDLli3pyJEjNHXqVOrXrx9t3ryZ0tLSaNasWWRjY0NOTk7k6+tLvXr1ohEjRhRoi1hgYCC1bt2axowZQ+PGjaOBAwdSixYtKDg4mG7dukVDhgyhqVOnfjcmZWq9yyzzMXbcuHHUoUMHGjVqFPn6+lJKSgqNHTuWJk+eLNxRISLat28fTZ48OV9+g3nh6dOn1KNHD6HF/+3bt6Svr0979uwRvgc7OzuytLTMUblR5m2V+f/d3d1p6NChZGdnR4sXL6b09HQKCQmh8ePHk5WVFW3evJnMzc2F4/Lnz5+Vopzsy3LOyZMnk6urK506dYp69uxJAwYMoOjoaDp06BB16dKFHB0dac+ePeTm5kbGxsZ048aNPI+JE90c+FaC9OrVKzI2NqaoqCjavXs3GRsb07Zt2ygqKorc3NwoMDCQ4uLiKCwsrMDqZzLHef36ddLX16cXL14QUUbSbWBgQF27dqWPHz8SUUbSY2lpmSXRlftWIqEo8vXavXs31a1bV6hpjI6OpsWLF1P79u3p9evXFBgYSBMmTCBDQ0OytbWl9u3bk62t7S/d3ktJSaG5c+fS0KFDhQuHT58+kYGBAenr65ODgwMtWbKETE1NycXFRSgFURZffo9paWl04cIF6ty5M40YMSLLay4uLtSiRYssiWDmukplIf8+379/T66urrRs2TI6evQoEWVcmNWvX582bdpERET379+nBg0a0Jo1a75KsvJzH09OTqZRo0aRg4MDxcbG0sePH8nCwoIcHR0pKCiIIiIiaOrUqTR69Gi6fv16vseTlxwcHMjQ0JC6d+9OSUlJdOPGDerXrx8ZGRnRjBkzhITl7NmzZGRkRG/evMny/vxMaNLT02nUqFE0d+5coYQrLi6OvL29qXfv3uTr60tJSUn0zz//UPfu3Wn48OHk5OSUb/F8z8qVK2nChAnC3zExMTRu3DgyNzenlJQU8vX1pYEDB9LcuXMLPLbckv8upVIpLVy4kGxsbOjAgQM0ZswYsrKyojt37lBQUBANGzaMbG1tycPDg3x8fMjU1JTWr1+v4Oi/z8/PjywtLUkqldK+ffvI2NiY1qxZQ8HBweTu7k5EGcfVxo0bZylt+xl5OWDm3/3UqVOFcrhTp05RmzZtaODAgcJvZtmyZTR69GjS19en5cuXF0gpZHZk/k1fvHiRrl+/TiNGjKC3b98SUcY5s23btjRz5kxKSkqiO3fu0MCBA8nOzo4GDRqUb40pnOhm07fqveRXcePGjaPevXtT06ZN6cqVK0SUsfM2adKEjh8/XqBxfnmFHxcXR5MmTaI2bdpQcnIyvXr1imxtbWn06NFElLFeISEhZG5uTv/++2+BxpoTmWsCXV1dqVmzZtSkSRMhgQ8KCqIpU6ZQjx49hCTz5MmT5OPjQwcOHMiTGHx9falBgwb0/v17SklJISsrK+rduzedOnWK+vbtS2ZmZkJrs6JlTuozH0AvXbpEN2/epJcvXxIR0fHjx6lly5a0bt06YZ7U1FRycHCgFi1aKGWCm9mLFy+EuulRo0aRoaEhzZ07l969e0dHjhwhAwMDodPZvn37SF9fv0BrjD98+EDm5ub06tUrIiI6cuQImZqaUo8ePcjFxYViY2PJ39+fbG1taeLEiRQbG6u0LVpEWY+D5ubmpK+vn6WTqouLC3Xp0oUuXbokTNuyZQsNHjw4XzvgZu7QJTdx4kTS19ennTt3CtNCQ0Np8eLFZGVlJVxYJCUlCTWC+SnztktPT6fU1FSysbEhNze3LLFHRERQ586dacWKFSSTyWjt2rXk6Ogo9PMoDN6+fUv29vbk7OwsdPJ8/fo1ubi4UP/+/SkoKIiCg4Np/fr1ZGlpSYMHD/6qo68ifKv2XO7BgwfUrVs3mjFjBpmamtLJkyeJKONOaYMGDb7ZiftnTp8+TY0bNxaOD/J6fRsbG6G1+9SpU2RsbExbtmzJcuzy8/OjYcOGkZeXV67WNa9l3r///vtv0tfXp7Zt25K+vr6wDxARPXnyhP7880/y8PAgooycJT09XajVzg+c6H7H924nrFmzhtq3b082NjZCT/pNmzZRmzZtsuxwkZGR1Lt37wLthCTf0aKjo2n58uW0cOFCunTpEgUEBJCNjQ3179+fiIj+++8/6tq1K7Vv356GDx9OzZo1K7CSipzK/OP5spfu0KFDqWvXrkLL45MnT2jIkCHk4OCQb/H079+fevXqRS1atKBRo0ZRZGQkEWW03P3777/08OHDfFt2TshbCTJf+AwZMoTMzMyoXbt21LJlS1q5ciVFR0fT1q1bqW3btnTq1Clh3k+fPtHChQsLPO6c8PPzoxUrVmTZd69evUq9e/empUuXUmJiIrm7u1OTJk2EcpLQ0NB8i+dbJ8iIiAiys7Oj27dv07lz52jMmDF0584doZVT/h34+fkVaCeoX+Hn50dXrlyhlJQUevfunXAXgyjjwnrEiBHUv39/6tevH02fPp0MDQ2FpCA/ZD5GZE5YpVIpDR06lHr27Jklyfb39ycXFxfq3r17lhNwfvqyAUJ+TnFxcSEbGxvhJC+f7uTkROPHjycioqioKKUtUZCTfwfy+KOioqhTp05ZRrcgyrizMmHCBBo+fLhwZywuLq5ALjR+5OXLl1+VnB0+fFg4vsg7Bbq4uJC+vr6Q1BJllNNZWloK56Gc3Dl8/vw5OTo6krW1tbCPPn/+nNq3b08ymYy8vb2pUaNGdOjQIUpMTKQJEyZkuYhUNjKZjE6fPk1jx46lt2/f0q1bt2jMmDHUo0ePLOWQZ8+e/eoiND9xovsd8iunzCevb9V7bdy4kaRSKTk7O5O1tTVNnjyZ9u3bRz169KBBgwbl61XKtwQGBlLLli1p3LhxNHjwYFq2bBm9efNGaPmSDx8VFxdH69evpy1btmRpdVam1qTMB4yjR49Snz59aMyYMcLtaCKiTp060dChQ4XWjlu3blGXLl1o1qxZeR6PVCqlNWvWkL6+Ps2dOzfLsD7K5MtWgtTUVFq0aBENGjSIoqKi6O3bt3Tx4kWqX78+eXl5UWRkJC1evJg6deqklHW4cpn3h+TkZBo8eDDVr1//q3rWffv2UcuWLYUkxsbGhkaNGiV8Rl71UP9eq/mdO3fo0aNHwq36gIAAio+Pp9mzZ9Pq1aspOjqaoqOjqWnTptS5c+evhpFSZgkJCeTq6krm5ub033//ERHRlStXqF69esLv8tOnT3Tt2jWaP38+LV68WBixIz9k/g527txJDg4ONGLECHJ1daWkpCQKDw+njh070qRJk7K87/r16+Ti4kLBwcH5FtuXMcbHx9OUKVNo6tSpNH36dAoNDSVfX1+ys7Oj1atXZ3mPs7MzzZ8/P8s0ZTvOyGVOwlNSUoQ4AwMDqWnTpjRv3rws81+6dIlGjBhBDg4OBX5+/JaTJ0+Svr5+ltI9V1dXMjU1pZUrV9LYsWOpX79+tHDhQoqPjydzc3MaOnQo7dq1i27evEm2trZkZ2eXo/KBzMeLBw8ekK2trXCMIso4ZrVo0YLatGkjtOyGhYVR586d6eDBg3mw1nlLPrSdfPhUeaMaUcbFjZ2d3VclcmvXri2wUhVOdL/hwoULZGRkRHFxcZSenk5paWk/rfdKTk6mEydOkJWVFY0ePZrmzJmjkNhXrVpFI0eOFP7OPK7o9evXqXHjxsJB9csTvrIeSNesWUNGRkbk5uZGc+fOpXbt2tGRI0eIKOPH36hRI3J2dqbg4GA6efIkXbhwId+GZHr58iUZGhrSrl27iEg5t9m3WgmGDh361Vi+vr6+pK+vT48fP6aoqCgaO3YsmZubF1hHoZzIfDKVnyTev38vdNaQl2LIWVpa0tKlS4mI8q1+7Xut5j169CAzMzPq3bu3sM1DQkLIxMREaI25fPkyDRs2rEDGaf0V32qlfv78Oc2ePZsGDRoklA7t3r2b6tWrR2fPnqW4uLivWqfz8ncSEhLy1bFrxYoV1Lx5c9q+fTudP3+eTE1NycHBgT59+kRPnz4VWs8zK8hW0ujoaLK0tKQRI0aQm5sb9enTh7p06UJnz56lFStWUP/+/cnFxYWuX79Ou3btIhMTkxzVeSpK5u91/vz5NHLkSBo6dCjt2bOHiDKGpaxXr55wvJTbt28fLVmyROH16N7e3lSvXr0sjT0hISHUr18/4QLt6dOnVK9ePeGObUxMDP31119kaWlJPXv2FFres+NbY7enp6fTpUuXyNLSUqjFPnXq1FcXwZ8+fSILCwvhAlPRMn/38v9/9+4dTZ8+ndq3by+UBqWlpdGlS5eoR48eX128FRROdL8hNjb2qx7Zjo6OP6z3unbtGhFltDQV1FXqlzVf6enpNGLECJo9ezYRZU1k9+3bR8eOHaN//vmH9PX1lfr2R2ZBQUHUvXt3unr1KhFlDPZuZGRErVq1Euqhb968SY0aNaJGjRrR1KlT8y35lG/PmTNnUs+ePZWiNSKzzOv94MEDGjRokNBK0L17d1q2bBkR/W9fIcroUOTs7ExEGaUfXyaMykC+3cPDw2nWrFn0119/0datW+nDhw8UEBBA/fr1I09PTwoPDyeijO0wZMiQrx74kJcn1S9bzVNSUsjFxYX69etHsbGxlJSURE5OTqSvr0/v3r2j8PBwGjJkCE2aNIk8PT3J1NRUKR8i8i0ymYy2bNmSZdqtW7do3LhxNGrUKOHkvXjxYqpfvz6ZmJjkS0uNTCajS5cuUZMmTYSSIaL/PeBD3sfg9evXZGJiQqtXrxbmO3PmDOnr63+VcOW12NhYOnToEB05ckQ4xt65c4c2bdpEo0ePznKLfuTIkWRvb09PnjyhY8eOkbm5OXXv3p06d+5Mx44dy9c481JycjLZ2tpS//79ycfHh44cOUL6+vq0atUqkkqltGvXLqpbt65wvCZSjtEi5GPPymtF5fz8/KhJkyaUmppKJ06coEaNGpGHhweFhYXR2rVr6fPnz5SSkkKJiYk5uiNw7tw5atSoEc2ePZtOnjxJcXFxwjjO8oeqtGvXTsgxvL29ydTUlBwdHcnd3Z26dOlCDg4OSjFmcuZj6cmTJ2nRokX077//UkpKCr1+/ZpGjRpFEydOFC6EExIS6NixY2RsbFxg5QqZcaL7Aw8fPqQBAwYIf9vZ2VHv3r0VXu9F9P2arxUrVpCZmZlwQJVKpZScnEwDBgwgT09PkkqlQr2Pssr8Q7506RJ17NiR4uLi6NWrVzRy5EhycXEhV1dX6ty5s1Bz+ezZswJrHdu0aRO1bdtW6EmqaJk7jMmT3ZSUFLp27Rp1796dli9fTj4+PtS5c2fhNph8+Dn5aBHK7uPHj9SpUycaO3YsTZ8+naytrcnKyooCAwPp8uXL1Lt3b5o5cyb9888/tG3bNjIyMsrX+vhvtZqPGjVKSLZWr15NpqamdOLECbp//z59/vyZfH196a+//qI+ffoILV6FwcuXL0lfX/+r2+unT58mW1tb4cKaKKPEKL87YwYEBBDR/46Bnz59IhsbGwoLC6NDhw6RkZERrV69mpKSkrIMYbVmzZp8vVXq5+dHbdq0oUGDBlHXrl1JX1+frl+/Tjt37hQ65sgvxogySopatWolPMkxJSWFYmJihCGi8rLMJi99GdPTp0/J0tJS6AR84sQJoTFFvi6LFi0iAwMDpbiQTkpKoj59+pC5uTmNGDGCHB0ds9TchoWF0ZgxY2j8+PHUqFEjYSSXN2/ekJGRUZZ5cyIxMZGMjIxIX1+fWrZsSS1btiRHR0e6ePEihYeHU3p6Om3evJnat28vLOPSpUs0evRomjx58ldP/FSUzA0qLi4u1KhRI+rfvz+1bNmSVq1aRampqXT9+nUaOnQozZ49Wxjd6dOnT/TPP/8oZAg0TnQz+bLF5/nz52RkZERTp04loowOZm3btlVovRfR/3a0+Ph4mj59Ok2ePJkcHR3p2bNndPv2bRo2bNhXncumT58uPKFITtG3jTL7Xue/+Ph4WrBgAb1//56WLl1KCxcupMjISDp//jzp6+tT586dvxq2KL/Fx8dnGf9Rka5evUqOjo5CyyIRZWlptrW1JWNjY5o5cyYtWbKEevfuTYGBgRQfH0/v3r0jCwsLpRgl4kfu3bsn7OPydXv+/DnNnDmT+vTpQwkJCbRv3z4yMjISTl75leT+qNV80KBBtHXrVnJ1daU2bdoI40E6OztnuQUp71SpjL53N+TUqVNUt25doWRIbvHixdS4cWOhTCSzvD6+ZL649/Pzo1GjRtHHjx9JJpORhYUFmZubU4sWLYTEhIioV69etHz58jyN41v8/PyoZcuW5OHhQUlJSRQXF0cPHjwQGhy8vb2pVatWdPHixSxPY1u3bh3179+fEhISsmwvZUxwib4euvLly5d048YNsre3J6KM4SubN29O58+fp+DgYBo1apRQCjVu3DiFHjflIyp07tyZbG1tiSjjAnrw4ME0ZcoUoX9Ceno6TZs2jRo1apTl8cMhISHUo0ePXxon/ePHj2RgYEBbt26l9evX05gxY8jY2JhMTU3J1dWV1q1bRzNnzqQOHToIF3TKcOfwW2P07tq1iywsLIQGvkWLFlGbNm2Euyb//PMPDR48mJYsWaLwcjh+Mtr/y/wUl6CgILx79w4GBgbw8vLCmTNnsHbtWujo6GDNmjW4ePEiVq1aJby3efPmmDNnDqpUqVIgsYrFYoSFhaFXr16IiopC5cqVERsbi8mTJyMgIAAdO3ZEYGAgJkyYgP/++w+7d+/GhQsX0KBBgyyfI19fRSMi4UkwFy5cwKxZszB9+nTs378fEokEs2bNgkgkwqVLl9CmTRvo6OhAU1MTTZo0QcWKFfHx48cCjbd48eJo3LhxgS7ze0xMTPD8+XMcOnQIYWFhADKeNJSamoohQ4YgMjISNjY2uHv3LlRUVNCoUSP06NED9vb2sLa2hqGhIfr166fgtchKJpMByNgvkpKS8PTpU9y4cQPh4eHCU5QMDAzQp08faGho4MSJE+jfvz+sra1RqVIlWFhYCE+ikz8V8FclJCQAyPjtyeOrV68exo4di9DQUHh4eGDAgAFYtWoVzp8/jxMnTqBZs2YgIoSGhmaJo0SJEnkSU16jTE8TO3z4MGbNmgUnJyfcvHkTXbt2xfTp0zF79mzcvXtXeE+NGjWgp6eH4OBgfPz4MctT6vLy+CKVSqGioiL8XbVqVVy9ehXu7u4QiUSYP38+Pn78CCsrK/Ts2RMAEB8fD5lMViBP17p79y4aN26MiRMnolixYtDS0oKRkRE0NTXx6dMnDB8+HCYmJti7dy9evnwpvC82NhY1atSApqZmlu2ljE86A/4X14kTJzB27FjExMRAV1cX9+7dQ9euXfHkyRNs3rwZHTp0ABHh3r17wv6yZs0ahR43RSIRJBIJNm/ejJ07dwIAdHV14ejoiJCQEPzzzz94/fo1JBIJpkyZgjp16uDGjRvw9PTErVu34OTkBDU1NVSvXj3XMejq6mLHjh3w9vZGo0aN4OXlBR8fH0yaNAmBgYHYtGkT/v33X7x//x5jx45FbGyscMxTlCVLlmDixIno3r077O3tcfz4cQDAw4cPYWJiAgMDAzx48ADXr19HxYoVcfz4cZw+fRo9evRAixYtcO3aNTx+/Fih66Dy81mKPiKCiooK0tPTMXz4cERERCA8PBxeXl5o1qwZ5s6di1mzZqFmzZro1q0bFi9ejEmTJqFs2bIYNGgQAGQ5COeluLg4nD17FmKxGKVLl0bbtm0BAOfPn0e1atWwceNGYd558+bhzJkzGDRoEKZOnQp3d3d4enoiISEBc+bMQYcOHfIlxl8lP3hu27YNq1evRt++faGtrQ0PDw/cu3cP06ZNw/v37xEaGopatWohPT0de/bsQbVq1TB37lyFHwgUqVixYti7dy/69u2L0qVLY+TIkQgPD4ednR2qVasGHx8fJCcnQyQS4eTJk1i/fj169OiB9+/fQ1NTU9iflEV6errwWxKJRNDQ0ICVlRUiIyOxY8cO3L9/H40aNQIANG7cGGpqarh//z6sra0xceJEODk54fTp0yhVqhTat2+fJ8nWtWvXsH//fkyYMAG1a9eGWCxGamoq1NTU0KJFC5QqVQrbt29Hz549YW9vj0uXLiE2NhYSiQRRUVEICwtDp06dfjmO/CSTyYQkd9u2bVi3bh26du2K4OBgTJgwAcOGDUO/fv0QHByMiRMnYsmSJahUqRKuXLkCMzMzDB48ON9+h1KpVPgez507B1VVVdStWxcnTpyAhYUFqlWrBgcHB0ycOBGLFi1CTEwMypUrhzt37kAkEqF58+b5Ehfwv4v0Dx8+QEtLS5ju7++PR48eYffu3fjw4QP09PRgZWWFHTt2wN3dHXp6eqhRowZ8fHwwefLkfIsvPxw/fhxbt27F2LFj0aRJEwDAxIkTsXz5cuzduxd169YFAISHh6NcuXKoUaMGAOVJ3uUNUjKZDCKRCCYmJrC3txceP2xjY4Py5ctj7ty5OH36NPbv348rV65AW1sb27Zt++XH0zZp0gTjx4/HjBkzsGnTJtSqVQu1a9dGnz598PHjR1y+fBlnz55FZGQkUlNT82KVcyUtLQ1DhgxBfHw8RowYAVVVVTx69Ah//vknAKBhw4ZISEhAQEAA9uzZg969e8PBwQF9+vSBt7c30tLSMHLkSLRs2VLhj3LmRBcZP8CkpCSMGjUK6urqWLZsGdLS0lCzZk2IRCJYWVnh/fv3mDZtGvT09GBubo4xY8YgPj4+X+N69OgRHB0dUbVqVURFReH169dYu3YtzMzM8PbtW0gkEshkMuGZ6c7Ozhg4cCDOnj2LJUuWYO/evUhNTUVqaiq0tbWF1hZlOeBkPoHFx8fj/PnzWLBgASwsLCCVSrFlyxaULl0aKioqqFatGpo2bQoLCwuUK1cOWlpa2Llz52+d5MqVLVsWa9aswZQpUxATE4NDhw7B0tISU6dOhYaGBkqUKAErKyuEh4dj9OjROHjwIBo2bKjosL8i349jYmKwZMkSJCUloWzZsrC2tsbw4cPx+fNnLF++HB4eHtDV1QWQ0bJXqlQpEBE0NTUxffp0TJo0CadOnYKJiQlKliz5y3GZmJhg3rx5OHToEAYPHowKFSoIreYjR47Ep0+fYGNjg7Nnz8Lc3BxNmjRBly5dYGBggPfv36Nt27YYMGDAL8eRn+Qn71OnTuHu3btwcnJCr169AABbtmzBiRMnUKZMGcyYMQOhoaGYOnUqRCIR6tSpgxEjRgDIemcmr8hkMuEYMXjwYISHh0Mmk+HPP//EwoULsXTpUkybNg1VqlSBvb09dHR0cOHCBTx+/BjVq1eHq6trnsbzJfn6VqpUCevXr0eFChXw8eNH3Lp1C0FBQahXrx46d+6MkJAQHD9+HEOGDMH27dvh5+cHCwsLuLq6omvXrvka46/K/L2mpKQgODgYISEhePbsmTBP37598fjxY4wZMwaWlpYoV64cdu/ejTZt2qBy5cqKCv2HxGKxcE60sLBAWFgY/v33X5QsWRLW1taoW7cu6tati2HDhiElJQVly5bNs2UPGjQIT58+hYuLC5YvX44KFSpARUUFVapUwcCBAzFw4MAsF/0FiYgglUqxZMkSaGlpYd26dcJxtEOHDlBVVcXdu3chkUgwevRoHDp0CPHx8WjRogUAoGbNmrh//z6OHDmC5s2bKzzJBYDftkb3yxqowMBA6tGjh9DB6Pr167Rx40YaO3YsnT59mkJDQ2nWrFlkYGBQIE+o+VbNl5+fn9DxyMPDg9q2bSvU1Mnrvvbs2UNt27al6Ohopa75ksednJxMfn5+FBUVRX369KGXL1/S9evXydTUlGbOnElRUVHk7u4ujEV67Nixr+oEWQb5eJDOzs7C9515H7h79y7NmDEjS4cYZRMUFETt2rUjR0dHmjt3Lo0cOZLq169P169fp6dPn9KYMWOoS5cu5O3tTRs3biQDAwOhE6J8n3r06FGOnjWfHREREdS2bVthpISwsDDq1KkTDRs2THi079KlS6lDhw708uVL8vPzo5MnTyr96CaZjwsvXrygLl26kL6+/ldPE3RxcaFevXoJf9+/f59u3rwp/J2X9bgvX76k4OBgoXOTfHzOfv36UWJiIr1584bevn0rfN+enp5Uv379LPEooq7RycmJOnXqRE2bNqVBgwbR4cOHs3T6tba2pq1bt9LLly+pXbt2tHr1aiFOZeovIScfG5UooyOVfJi+pKQkWr16NbVp0yZLPTRRRoc/+QMhNmzYUOAxf8vPzn2Z69IXLlxIgwYNon/++Se/wyKZTEZWVlY0bdq0LB3clWHIyrS0NOrRo4fQaTZzTFu3bqX69etT48aN6datW7R48WLhMdZJSUk0YsQI2rFjR74+CTGnfttEV+7ixYtElPFFtm/fnmxtbWnYsGHUoUMHsrS0JHt7e2rcuDE9fPiQIiMjadSoUQXSc3Tz5s00ceLEb74WERFBcXFx1KtXL6GjnJyHh0eWZ6crs48fP5KFhQWtXLmS3r9/T7179yZHR0dq0qQJbdy4UZjPzMzsq4507NtWrlxJtra2dO/ePWFa5oOUsl3wfGnTpk1ZxoEmyhifs1WrVvT+/Xt6/Pgx9ejRgxo1akQrVqwQhvUjyv91e/ToEXXq1ImWLFlCpqam9Pfff2dJZAICAmjKlCnUoUMHhfQszonM2yoyMlJIYu7evUsWFhY0f/78LEN4xcTEkJGRUZbtLZdXSVpKSgqNGDGCevfuTW3btqXGjRsLHU1nzZr11bGOKGPIpoSEBJo/fz4ZGRkJwxkpaj9/9+4dffz4Mcv3L+9EN3XqVGEs1gMHDlDHjh1px44dStU58VsXCOfPn6fhw4dTv379yNXVlW7cuEGpqak0d+5c6t+//zc7mCn6SWeRkZF05cqVr4YJ/Z7Mo9VMnTqV+vfvn2U4tPySnJxMLVq0oKVLlypFpzO558+fU/369YWxwuVmz55NTZo0oYMHD5Krqyt16tSJpkyZQiNGjKC//vqLunTpQsOGDVO688xv3RktIiICEydOxOTJkyEWizF//nyIRCKoqKjAyckJa9aswfbt21G+fHk8fvwYOjo68PLyQp06dfItJvr/Wynfqvk6cOAAevXqBUtLS4waNQp9+/bFyZMnMX36dGzZsgX79u3Dzp07hY44yiwwMBBOTk6oVasWhg4dikqVKsHW1hZnzpyBvb09hg8fDiCjpKFUqVKoV6+egiMuHCZOnIiSJUti27ZtCAgIAIAsNWXKUrYC/K/Tmfz/iQj3798XbpOlp6cDAObMmQNdXV14eHigQYMGcHR0RPXq1ZGcnCzcLktPT8/3dTM0NMSkSZOwZcsWdOrUCbNnz4aGhobQ0axWrVoYMGAAmjRpgrS0tHyN5VfJt9XBgwcxYMAA9OvXD4cOHYKJiQlGjBiBZ8+e4dKlS8J6REZGolq1at/siJMXddAfP35E9+7dAQCurq5YtmwZ7O3thc+Oi4tDUlISgP/tK2lpaThw4AA2btyIOXPmoH79+govz6patSp0dXWzLF9FRQUPHz7EixcvhHNH37590alTJ2zZsgUxMTEKifVLjx8/Rrdu3bJ0ljt+/DgmT56MFi1awNbWFhKJBEOGDEFERAT69++PypUrw8vLC8HBwQD+95vV0NBQyDoAGZ2kOnXqhBUrVqBHjx44deoUgKzHmy+JxWJIpVKoqanhr7/+Qu3atQvknKOuro7Dhw9j586diIqKyvflZVfp0qVRrlw53L59W5gWGxsLPT097Nq1C3369IGNjQ309fXx7t07GBoaQiKRoHPnzti0aZNSnWeA36xGN3NNKADo6Ohg7dq1mDBhAjZt2oThw4ejdevWAIBPnz5BR0cHL168QFpaGqpWrQoAv1yI/jPZqfnq1KkTQkJC8O+//2LatGm4ffs2Tp06BZFIBCcnJ/Tp0ydfY8yNL7d9QEAA3r17Bw0NDZQqVQoAYGVlBT8/P+zduxcikQjVqlXD5cuXER0dLXRwYD8mEomwcuVKDBw4EFu3bsWYMWMKbDSQ7IiOjkbp0qUB/O+3lJqaCpFIBFVVVdSqVQsnT55EYmIiNDU1kZaWBlVVVbRv3x7Xrl2DVCqFqakpoqKisGHDBlSpUgV2dnYFVstmYWGBly9f4t69e3jw4AEaNWok1MqLxWKYmJigUaNGSnegl8v8O3z79i28vb0xePBg3L9/H7t370ZaWhpsbGwQEhKCffv24e7du2jdujWOHz8OiUQi/FbzCv1//efRo0dhYGCAFStWCPFl7qE/atQo9OvXDydPnkS3bt0AAKqqqkKCAgC7du3K09hy6+XLl1i6dCm6d+8OdXV1fP78GcuWLYOVlRXMzMyE+WbMmAErKyvh3KJIsbGxWLp0Kf744w/88ccfwnQ/Pz8MGTIEQ4cORXR0NLy9vdGmTRuIxWL88ccfsLe3h6enJ2bPno0NGzZAXV0dgOIuNGQyGU6fPg1ra2vMnDkTTk5OWLduHSwsLLKcuzN3vpST/44rVaqEBQsWFFjM5cuXx507d4RtpwxKlSqFEiVK4Pbt2+jevTskEglKliyJwYMHC9utdu3aCAoKQq1ateDo6JgvNfp55bdKdCUSCaKjo/Hy5UsYGxtDTU0NzZs3x6xZs+Di4oIqVarA3NwcFy9ehKurK+rXr4/bt2/DwsKiwHunDx06FK9evcKxY8cQGxuL2rVrY/To0TA3Nxeulm1sbJCeno5169YhPT0dCQkJwonoWz9kRcl8cn348CHU1NTQpEkTjB49GvPmzcOOHTtgb28PAJg/fz7KlCmD+/fv48KFC9DW1sbBgweF5Ij9nJqaGjZs2ID+/ftj1KhRig5HcODAAeHizMDAAElJScIBUltbG4sXL0afPn1w69YtuLq6ws3NDaqqqgCAxMREVKlSRTjgdurUCe/fv8fGjRvRtWtX6OjoFNhBduLEiRg/fjy2bduGkiVLCiMxyCnrwT5zx64rV64gJCQE7du3h62tLWxtbbFgwQIcO3YMZcuWxfjx4xETE4Ndu3YhLCwM1apVg6enp9CBJ6/WUf45ISEh0NbWFrZjdHQ0wsLC8M8//yA0NBSmpqaYMGECZsyYAU1NTVSrVg3q6uoIDg5Gq1atAORPZ7jcqF69OgwMDODm5gYtLS2ULVsW06ZNg7W1NYD/tSyKxeJ8vTuYEzKZDFFRUWjatCmSkpJw9OhRDBgwAI8fP0aPHj3w7t072NjYoFWrVpgzZw6OHTuGmJgYjB07Fubm5oiIiFCKRE0sFiM4OFho/be2tsaHDx+EUVJu3ryJUqVKoW7dut/cXxR1zlSGbZdZsWLFMGHCBIwfPx5NmzYVhuzLvH1evHgBVVVVWFhYAFCe39+3iIgyDXxYhKSlpWHHjh0oXrw4ypYtK1xJz5w5Ew8fPsSCBQuEFoPk5GSsW7cOmzdvxrFjx6Cnp4d9+/YhJSUFlSpVUujQQMHBwShWrBhUVFSEZE/eG3P69OnQ09PD6NGjhfmVdWeTSqUYPHgwYmNjERQUhDFjxqBTp044c+YM9u7dC1dX1ywXE1KpFJ8/f0aZMmUUGHXhJj+4K4vAwECMHDkS7dq1Q79+/TBv3jxoa2ujZs2auHjxIsqVK4cdO3bg1KlT8PLyQrVq1dClSxdER0dj9erVmDNnjjASAACEhYUJLTAFLTU1FQMHDoS+vr7StZp/i/zCNyEhAb1790Z6ejo+fPgAU1NTeHp6QktLC4mJiXByckJ8fDwmTJgAfX19TJkyBerq6rCzs4OxsfFXd2byirOzM8LCwjBhwgRERETg5MmTuHbtGhITE1GxYkVERESgR48eiIqKwrNnz6CmpobExES0aNECixcvzvN48sKHDx+grq4OmUwm9NhXpgaIL505cwaTJ09G8eLF0bRpU6xevRrbtm0ThkibNGkSHBwcIBaLMWnSJKSmpsLLyyvf9oncunbtGqZPn47atWtDIpEgNDQUW7ZsgUQigZubG9LT07FmzRqlPE8qEyLCokWL4OPjAzc3NzRr1gxlypRBeno6kpKSMG3aNMTGxsLT01PpG6KKZKL79u1bDB48GFWqVEFaWhrCw8PRrVs3TJ8+HampqXBwcECJEiUwceJEGBgYAACioqLg4OCAlJQUHDx4EMWLF1fwWmQVExMDbW1t4e+HDx/CxcUFkyZNQseOHRUXWDYQEebNm4fAwECsXr0aERERSExMhLGxMT59+gRvb29cvHgR69evR61atYTb1azwk48DTUSoW7cu0tPTMWHCBHTs2BGRkZHw8PAAESEgIAAODg7o0KED/v77b9y7dw/u7u6QyWRIS0vDsGHDYGlpCUB5koWoqCj0798fmzdvRrVq1RQdzk+Fhobi2bNnuHz5MkaMGIGzZ89i+fLlcHd3R6dOnaCqqoqQkBDMnz8fEokETk5OSE9Px7x581C9enXY2Nh89dCZvPLp0yf07dsXiYmJ+Pz5M0qWLImOHTvC3t4eNWrUQHx8POzt7eHk5IQ6deogPDwcaWlpwjiuyihzo4OyNkBk9vbtW5ibm6NYsWLYsWMHGjZsiAcPHgi/0S1btgglQtOmTYOOjg5mzpypNL9HubS0NLx8+RLHjx/HwYMHQUTQ0NBAhQoV8ObNGxgaGmLKlCmoUKECypQpo1RJurJJTU3FokWLcPr0adSsWROmpqaIj4/Ho0ePIBKJsH37dhQrVkzRYf5cwfV7KxiXLl2i+vXr099//01ERFFRUXTmzBlq0aKF8FjQd+/eUZcuXWj27NlCr8z09HQaM2YM1atXj8aNG0dEytND3d/fn4YNG0ZHjx4lX19f2rdvHzVu3JgWLVqk6NCyJSUlhWxtbWn79u1fvebn50cXLlygqVOnUufOnZVqSBL2a/z8/KhNmzY0aNAgMjc3J319fXr8+DHt37+fGjZsSJ07d6aIiAgiyuj1fOXKFWrQoIEwhFdaWhqlpKQIvdJlMpnS/CbllKmn9I/ExcWRra0tGRkZZRnRRP4I37t37wqjJ9y8eZNsbW2F0WUuX75MHTp0oF27duXr9g8NDSVfX1+6cOECPXr06KvXhw4dSh4eHvm2/N/Nl6NlxMbG0unTp2nUqFHUuXNnYajNI0eOkK2tLfXo0YNWrlxJ48aNoyZNmgiPflU28n30+fPnZGxsTNevX6fDhw+Tp6cnDRw4kPT19ally5bUq1evfHtUeFFz8uRJWrx4Mdna2tKiRYu+eS5XZkUq0b148SLp6+vT1q1bs0yPiIigAQMG0IMHD4QT0507d6hdu3bk4eFBz549o/fv39Pw4cMpICBAKcaxyyw5OZnc3d2padOm1LFjR+rfvz/t379feF2Z4v3eUEN9+/al2bNnZ5kvODiYrKys6Pr163T//n0aPXq0MJwQK9y+NQ70/fv3hWGHvLy8qG3btuTr6yv8JlNTU8nHx4f09fW/GitZ2RJcZffl71AqldLVq1epbdu25OLikuW1UaNGkbm5OQUGBgrbWT5et5yiEgJ5PEFBQWRra1sgQz79DjLvHxcvXqTLly/Tw4cPhWldu3YlW1tb4SLT39+fZs+eTX/99RdNnjxZqcfilrtx4waZmZllOXacOXOGjIyM6MaNG3To0KFCc6HKfk2RKl1ITk5G8+bNMWnSJNjY2EBFRQUSiQSpqalo3rw5/vzzT0RERMDOzg7W1tbw9fWFh4cHpFIpoqKi0KVLF7i5uSl6Nb5L2Wu+MtdqXbp0Cerq6lBVVUXjxo2xfft2+Pr6YtiwYVlqnrt3745BgwbBxsaGSxaKkC1btuDRo0fw8PD46rVPnz6hTJkymDJlCiIiIoQOD0DGb3j58uVISkoq0J7PRcmXj8zV0tJC6dKloa+vj6NHj8LJyUkYFUCuW7duUFFRwbZt24R6O8poCCnw40tgYCCOHz8OBwcHxMXFIS4uDjNmzEDZsmWxfv16Pkb8osznDPlQYaqqqkhISEDLli0xb9484XxoZmaGOXPmKHS4sNwKDAzEmDFjMG3aNOGcc+DAAVy6dAmenp4Kjq5wIgUP35dbRWrUhWLFisHX1xft27eHtrY2evbsiZCQEAwdOhR//PEH2rVrh+fPn2PhwoVITU2Fra2t0NFBLBYrfa1rhQoVvqr5UpYkN3OP7swHz7i4OHTr1g1dunTB8+fPsXv3bmhpacHU1BSvXr2CTCYThtfhE1jhJ98vvzUO9KNHj4SOLXp6erCyssKOHTuwZ88elCpVCgYGBihWrBimTZumVJ3olB1lqv+U/w6TkpJga2uL+Ph4EBEiIiLg6OgIS0tLTJo0CTNmzEDNmjWFmtudO3diw4YNWTqViEQihZzQNDU1ce3aNezatQulSpWCmpoaGjZsiCVLlhR4LEXFzZs3UaJECVSoUAE6OjpIT0/HypUrkZCQgIMHD0JDQwPLly/Hxo0b0a9fP9SvXx8bN26EnZ0dKleujBEjRijNyACZk63M+/6XneJKliwJIyMjrF27FpcuXUK5cuWwdetWTJ48WSFxFwWFLcGVK1ItunJ3797F2LFjMXr0aGzYsAFdu3bFzJkzhR+qo6Mj3r59i+PHjys40sLtewfP27dvY8eOHdDQ0MCyZcuwadMmXLhwAZ8+fcKBAwdw+PBhGBoa4vXr1+jatSvmzZun6FVheWzr1q1Yv3497OzsvhoHun79+ggJCYFUKkWPHj2wfft21KlTB3/99VeWMUWV6W6FMouLi4OKiorQ6ibvQBYVFYWlS5ciJSUFV69excKFCzFmzBh0794dy5Ytw9WrV3Hy5MmvxsYlJeg4lZycjAsXLqBYsWLChTHLudTUVIwcORKBgYFISUlB6dKlsWnTJlStWhUzZsxAgwYNYGdnBw8PD+zduxezZ89GxYoVoaGhgfr162P//v2YO3cuzp49q/DxfmUyWZaLr8uXL+Ps2bOIiYnB9OnTvxlfYGAgbt68CV9fX5QoUQKWlpbCOMzs91EkE10Awg902LBhmDZtGoD/Dbe0Zs0a+Pn5wdPTU2muUguTnB48nZ2dUblyZZQuXRp6enq4desWkpKSoKamJjzZihU9s2bNwt27d4VxoPv06ZNlHOj+/fvD0tISDRs2xPz587F+/XqUK1dOwVEXHtHR0XBxcUFMTAyioqJQqlQpeHp6okyZMhg0aBDMzc1hZ2cnzL9161Z4enpi3759qFChAgYMGAAVFRUcOXJEgWvB8svbt29ha2uLunXrYuLEiQgNDcW2bdsQFxeH48ePY/To0TAwMEB8fDzOnj2LJUuWoFmzZliyZAnevHkDLy8viMVivH//HpUrV1bYegQHBwtJrPzid8eOHVi2bBl69OgBqVSKPn36ZHnAyJcXyenp6ZDJZHyn6DdVpEoXMuvfvz+ePn2KR48eITw8HOXKlYOamho+fvyI06dPo2PHjpzk5kLmg+e6deuEg+fYsWNx/PhxfP78GZ8+fcKCBQtw9uxZrFq1Sjh4vn37Fp6enkI9JivaFi1a9MNxoKtVq4a4uDj8+eef2Lt3L/8ec+DVq1cYOnQomjZtikGDBiEhIQEPHjxAbGwsVFVVhe0OACkpKVBXV8fQoUPxzz//wMfHB7NmzcLatWtRRNs5fnuXL1/G2LFjMWDAADg7OwMAGjRoAHV1dUybNg1v375Fv3794OjoiPLly8PX1xeampoAMobNK126tJAoKmKcarn4+HgMGTIEtWvXhre3N8RiMcLDw3H27FmsXLkSHTt2RGpqKsLDw3Hz5k3UrVsXJUuW/OpOUEE9OZEppyJ9X3D+/PlITEzEsmXLIBKJ8PDhQ3Tv3l24Tcpy5vLly7C0tETXrl2xceNGNGjQAJ07d8aoUaMQHh4uHDw3bNiA8+fPw9fXF82aNQOQcfDM/OQj9nuoWrUqdHV1s9wKV1FRwcOHD/HixQvhcaNqamqcdGUDESEtLQ2bN2+Gubk5lv9fe/cek+P/x3H8mVOi7qVlZSiHTWbTHxSL6A9txsyhVTdNK5PI5lBOjX9kzodly7IVItZmRjFzyzCEspFGG2vccuhE4TaSnL5/tK7fz2Hf3w/fr+vu9nr82e67ve/rrvf1uq7rc9ixg0mTJhEdHc2GDRvo378/Xl5eTJ8+nczMTBoaGnB3d+fjx4+0trbi5eVlTGQNDAxkwIABxha64hpKS0tJTk5m7ty5RshtaWkB2rZ29ff3x9PTk9DQUBISEmhububBgwc0NDTw+PFjbt++zcCBA43fZ+YwFk9PTzIyMigvLycrKwtom8tx7949qqqquH79OosXLyYpKYnExESSkpKw2+2m1SvOyaVTh5ubGwUFBVy9epXExEQSEhKYOXMmmZmZZpfW4fzTzVP+HFVVVSxfvpzjx49z+vRpDh8+zLx58xgzZoyxY6FZE586Gjc3N5qbm6moqDAuIuE/E3TaJ+NER0cTEhLCnDlzaGpq4tWrV9jtdmpra78Zy6gF813LiBEjjO+4rq4OaJuo/eHDB3Jzc4G2CX8Wi8UYsx0fH09iYiJWq5Xg4GCSk5NNq79d+1bJ4eHhrF69muzsbM6cOYOPjw+xsbHk5OQwe/ZsOnfuTFxcHFevXuXRo0ecO3fO5MrF2bj8/Xx3d3eKioqIiIhg+/btxr7M8mO+bp59+vT5bvPs0aMHKSkpvH//nvj4ePz8/HA4HIwfP94pmqf8foGBgQwdOpRNmzbh6emJr68vK1asIDY2FtCksx/15s0bGhsbjUfKX6+60KlTJy5evMiUKVOw2WxMnjwZPz8/GhoaiIqKYtKkSWaWL/8yd3d3CgoKiImJoWfPniQnJ1NfX09iYiLV1dUEBAQwY8YM3N3dmTp1KvPnz2f8+PH4+vricDgICwsz+yN8saxdaWkpdrsdX19f0tLSKCoqIjU1lQkTJuDt7U3fvn2Ni7WRI0c6/Xa08vu57GS0r7VPRJOf9+zZM2JiYoiLi/tu83RzczOa57Rp07hz545TNU8xl7OvA90RfP78mSdPnpCUlMScOXOIjY397vHLzc2loKCA4uJiysrKePfuHRaLxRgfr+Pu+m7fvs2yZcuIjIzk6NGjTJs2jaioKDw9PSkpKeHSpUtUVFTw4sULIiIiyMrKMvUc+fjxYzw8PIzeAHDu3DmWLVtGcnIyzc3NlJeXc//+fYqLi/H29qaqqorCwkJGjx6NzWbj8uXLHDx4kEGDBpn2OcT5/DFBV/4ZHa15ivP47zuPzrCEVUe2cOFCampqyM/P/2J5sPbjevHiRdasWcPRo0fx8/P74r0KuX+OU6dOkZaWRnR0NOvWrfvme6+rq+PSpUuEhYUREBBgUpVta+BGRESwatWqLzYy2bJlC3V1dcbGMy9evCA9PZ2amhpOnjxJfX09K1euNG5kbdu27Zu/dxEFXflhHaV5iriqmpoaYmNjCQ8PZ926dcaKFe0rWhw8eJCzZ8+Sm5urC80/3M6dO7lx4wapqamMGDEC+HZzBTMVFRURFhZGt27djGEH7cE1PT2djx8/sm3bNuP1T58+xWq1EhYWxsaNGwF4+fIl3t7eZpQvHYAu6+WHTZ48mQULFvDw4UMqKiqMn7fP3u7Tpw9Wq1UhV+Rf0rdvXzIyMrDZbKxfv57Kykqg7X/w7t27HDp0iJCQEIVcYcmSJVgsFvbv38+9e/cA55mAeOvWLdLT03E4HPTq1YvW1laioqI4cuQIACEhIRQXF1NfX2+8x8fHh6CgII4dO2ZsE66QK39HQVd+ijM3T5E/QWRkJNnZ2Vy5coWlS5ditVpZsmQJCxYsYNy4cSxatMjsEsUJuLm5kZmZSW1tLXl5eTx58sTskgwOhwOLxULv3r2BtmUGR40axY4dOygtLSU6OprQ0FDmzZvHmzdv+PTpE126dMHf35+UlBRjC2uRv6OhC/LTWltbiYuLIygoiJSUFPr162d2SSJ/HLvdjt1up6SkhMDAQAYPHkxERATgXI+oxVzPnz/HarWyd+9ep3naVlZWxqpVqzhx4gQ9evSga9euAKSlpXHz5k1ycnLo3r07ixcvpqWlhaFDhxqrihw7dsxpPoc4NwVd+SXO2DxFRJPO5Ftmrz5UXl5O165dGT58OAA2m40NGzZw4cIFunTp8sWFWVRUFN26dSM7OxuLxUJOTg6NjY20tLSwcOFC3ViR/5uCrvwys5uniIg4v4SEBKqrq8nPzycwMJAjR46wZ88eiouLjde0n0/evn1LZGQkoaGhbN26VecY+Wm63JdfpgYkIiL/y4EDB/Dy8iIjI4PXr1/z6dMnHj58SEZGBnl5eTQ1NRnnEw8PD/Ly8jh9+jSbN2+mtbXV5Oqlo9IdXREREfktGhsbmThxIrNmzcLLy4t9+/YxZMgQ7ty5Q+fOnfH392fixImEh4cTHBzMtWvXsNvtzJo1y+zSpYNS0BUREZHf5vr168THx+Ph4cHUqVNZu3YtTU1NlJSUcP78eSorK6mtrSUgIIDCwkJ69uxpdsnSgSnoioiIyG9ls9lITU1l7Nix7N69+4shcA6Hg7KyMoYNG0b//v1NrFJcgYKuiIiI/HZZWVkcOHCAffv2ERwcbHY54qIUdEVERMQUKSkpVFdXs2vXLgYPHmx2OeKCFHRFRETENBMmTMDHx4f8/Hw8PDzMLkdcjIKuiIiImMbhcFBZWcnYsWPNLkVckIKuiIiIiLgkbRghIiIiIi5JQVdEREREXJKCroiIiIi4JAVdEREREXFJCroiIiIi4pIUdEVERETEJSnoioiIiIhLUtAVEREREZekoCsiIiIiLklBV0RERERckoKuiIiIiLikvwCsRcO7MxQOZgAAAABJRU5ErkJggg==",
                  "text/plain": [
                     "<Figure size 700x300 with 1 Axes>"
                  ]
               },
               "metadata": {},
               "output_type": "display_data"
            }
         ],
         "source": [
            "# Plot label distribution\n",
            "label_dist = get_label_dist(curlie_labeler)\n",
            "original_curlie_dist = {\n",
            " 'Business': 27.6,\n",
            " 'Society': 13.9,\n",
            " 'Arts': 9.3,\n",
            " 'Home': 8.4,\n",
            " 'Health': 7.4,\n",
            " 'Recreation': 8.4,\n",
            " 'Computers': 6.2,\n",
            " 'Shopping': 7.4,\n",
            " 'Reference': 4.3,\n",
            " 'Science': 4.8,\n",
            " 'Sports': 6.8,\n",
            " 'Kids_and_Teens': 1.1,\n",
            " 'News': 1.1,\n",
            " 'Games': 1.7\n",
            "}\n",
            "for category, freq in original_curlie_dist.items():\n",
            "    label_dist.loc[len(label_dist)] = [\"Original Curlie\", category, freq]\n",
            "\n",
            "renamer2 = {\"Original Curlie\": \"Original Curlie\", \"GPT-3.5 + 1-shot + C2\": \"GPT-3.5\", \"GPT-4 + 0-shot + C2\": \"GPT-4\"}\n",
            "label_dist[\"labeler\"] = label_dist[\"labeler\"].apply(lambda x: renamer2[x])\n",
            "\n",
            "label_dist[\"labeler\"] = pd.Categorical(label_dist['labeler'], [\"Original Curlie\", \"GPT-3.5\", \"GPT-4\"], ordered=True)\n",
            "labeler_dist = label_dist.sort_values(\"labeler\")\n",
            "\n",
            "fig, ax = plt.subplots(figsize=(7, 3))\n",
            "fig.tight_layout(pad=1.0)\n",
            "sns.barplot(\n",
            "    label_dist,\n",
            "    y=\"freq\",\n",
            "    x=\"category\",\n",
            "    hue=\"labeler\",\n",
            "    order=label_dist.groupby(\"category\").sum({\"freq\": \"sum\"}).sort_values(\"freq\", ascending=False).index,\n",
            "    palette={\"Original Curlie\": \"#852D48\", \"GPT-3.5\": \"#49498D\", \"GPT-4\": \"#6D6DD3\"},\n",
            "    ax=ax\n",
            ")\n",
            "ax.get_legend().set_title(\"Labeler\")\n",
            "ax.set_xlabel(\"\")\n",
            "ax.set_ylabel(\"(%)\")\n",
            "ax.set_xticks(ax.get_xticks())\n",
            "ax.set_xticklabels([rename_dict.get(x.get_text(), x.get_text()) for x  in ax.get_xticklabels()], rotation=40)\n",
            "\n",
            "path = os.path.join(FIGURE_DIR, \"curlie-10k-dist.pdf\")\n",
            "fig.savefig(path, bbox_inches=\"tight\")\n",
            "print(f\"✅ Saved figure to {path}\")"
         ]
      }
   ],
   "metadata": {
      "kernelspec": {
         "display_name": "ml-project-2-mlp-a6NSXBdT-py3.10",
         "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": "3.10.13"
      }
   },
   "nbformat": 4,
   "nbformat_minor": 2
}