3_racism_detection_model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training HuggingFace models in google colab\n",
    "\n",
    "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/10kGk_-X2F2gEOpyGpJLm-viy7tp72-II?usp=sharing)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import pandas as pd\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "plt.rcParams.update({'font.size': 14})\n",
    "\n",
    "from src.utils import load_dataset\n",
    "from src import utils, huggingface\n",
    "\n",
    "pd.set_option('display.max_colwidth', None)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Raw labels\n",
    "\n",
    "First evaluate the raw labels, without the aggregation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>epoch</th>\n",
       "      <th>f1_validation</th>\n",
       "      <th>f1_evaluation</th>\n",
       "      <th>f1_average</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0.789536</td>\n",
       "      <td>0.867925</td>\n",
       "      <td>0.828730</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>0.819090</td>\n",
       "      <td>0.842105</td>\n",
       "      <td>0.830598</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>0.790974</td>\n",
       "      <td>0.905660</td>\n",
       "      <td>0.848317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>0.799534</td>\n",
       "      <td>0.888889</td>\n",
       "      <td>0.844211</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   epoch  f1_validation  f1_evaluation  f1_average\n",
       "0      1       0.789536       0.867925    0.828730\n",
       "1      2       0.819090       0.842105    0.830598\n",
       "2      3       0.790974       0.905660    0.848317\n",
       "3      4       0.799534       0.888889    0.844211"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# for each model and epoch, get the predictions on both validation_set and evaluation_set and compute f1_score\n",
    "all_values = []\n",
    "for epoch in [1,2,3,4]:\n",
    "    values = huggingface.evaluate_pipe('raw_label', epoch)\n",
    "    all_values.append({'epoch': epoch, **values})\n",
    "    \n",
    "pd.DataFrame(all_values)\n",
    "# everything matches"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Majority vote"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>epoch</th>\n",
       "      <th>f1_validation</th>\n",
       "      <th>f1_evaluation</th>\n",
       "      <th>f1_average</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0.735586</td>\n",
       "      <td>0.884615</td>\n",
       "      <td>0.810101</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>0.771800</td>\n",
       "      <td>0.867925</td>\n",
       "      <td>0.819862</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>0.734531</td>\n",
       "      <td>0.857143</td>\n",
       "      <td>0.795837</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>0.767486</td>\n",
       "      <td>0.851852</td>\n",
       "      <td>0.809669</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   epoch  f1_validation  f1_evaluation  f1_average\n",
       "0      1       0.735586       0.884615    0.810101\n",
       "1      2       0.771800       0.867925    0.819862\n",
       "2      3       0.734531       0.857143    0.795837\n",
       "3      4       0.767486       0.851852    0.809669"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_values = []\n",
    "for epoch in [1,2,3,4]:\n",
    "    values = huggingface.evaluate_pipe('m_vote_strict', epoch)\n",
    "    all_values.append({'epoch': epoch, **values})\n",
    "    \n",
    "pd.DataFrame(all_values)\n",
    "# everything matches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "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>epoch</th>\n",
       "      <th>f1_validation</th>\n",
       "      <th>f1_evaluation</th>\n",
       "      <th>f1_average</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0.799353</td>\n",
       "      <td>0.857143</td>\n",
       "      <td>0.828248</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>0.825040</td>\n",
       "      <td>0.827586</td>\n",
       "      <td>0.826313</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>0.830861</td>\n",
       "      <td>0.825397</td>\n",
       "      <td>0.828129</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>0.841941</td>\n",
       "      <td>0.866667</td>\n",
       "      <td>0.854304</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   epoch  f1_validation  f1_evaluation  f1_average\n",
       "0      1       0.799353       0.857143    0.828248\n",
       "1      2       0.825040       0.827586    0.826313\n",
       "2      3       0.830861       0.825397    0.828129\n",
       "3      4       0.841941       0.866667    0.854304"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_values = []\n",
    "for epoch in [1,2,3,4]:\n",
    "    values = huggingface.evaluate_pipe('m_vote_nonstrict', epoch)\n",
    "    all_values.append({'epoch': epoch, **values})\n",
    "    \n",
    "pd.DataFrame(all_values)\n",
    "# everything matches"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Regression on weighted majority vote"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "all_values = []\n",
    "for epoch in [1,2,3,4]:\n",
    "    values = huggingface.evaluate_pipe('regression_w_m_vote', epoch)\n",
    "    for t, t_values in values.items():\n",
    "        all_values.append({'epoch': epoch, 'threshold': t, **t_values})\n",
    "    \n",
    "df = pd.DataFrame(all_values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Weighted majority vote strict/nonstrict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>epoch</th>\n",
       "      <th>f1_validation</th>\n",
       "      <th>f1_evaluation</th>\n",
       "      <th>f1_average</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0.780316</td>\n",
       "      <td>0.867925</td>\n",
       "      <td>0.824120</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>0.834862</td>\n",
       "      <td>0.852459</td>\n",
       "      <td>0.843661</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>0.845201</td>\n",
       "      <td>0.847458</td>\n",
       "      <td>0.846329</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>0.821830</td>\n",
       "      <td>0.888889</td>\n",
       "      <td>0.855359</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   epoch  f1_validation  f1_evaluation  f1_average\n",
       "0      1       0.780316       0.867925    0.824120\n",
       "1      2       0.834862       0.852459    0.843661\n",
       "2      3       0.845201       0.847458    0.846329\n",
       "3      4       0.821830       0.888889    0.855359"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_values = []\n",
    "for epoch in [1,2,3,4]:\n",
    "    values = huggingface.evaluate_pipe('w_m_vote_strict', epoch)\n",
    "    all_values.append({'epoch': epoch, **values})\n",
    "    \n",
    "pd.DataFrame(all_values)\n",
    "# everything matches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>epoch</th>\n",
       "      <th>f1_validation</th>\n",
       "      <th>f1_evaluation</th>\n",
       "      <th>f1_average</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0.846954</td>\n",
       "      <td>0.813559</td>\n",
       "      <td>0.830257</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>0.836923</td>\n",
       "      <td>0.900000</td>\n",
       "      <td>0.868462</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>0.841782</td>\n",
       "      <td>0.881356</td>\n",
       "      <td>0.861569</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>0.828250</td>\n",
       "      <td>0.896552</td>\n",
       "      <td>0.862401</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   epoch  f1_validation  f1_evaluation  f1_average\n",
       "0      1       0.846954       0.813559    0.830257\n",
       "1      2       0.836923       0.900000    0.868462\n",
       "2      3       0.841782       0.881356    0.861569\n",
       "3      4       0.828250       0.896552    0.862401"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_values = []\n",
    "for epoch in [1,2,3,4]:\n",
    "    values = huggingface.evaluate_pipe('w_m_vote_nonstrict', epoch)\n",
    "    all_values.append({'epoch': epoch, **values})\n",
    "    \n",
    "pd.DataFrame(all_values)\n",
    "# everything matches"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Performance evaluation\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Threshold analysis\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>Test</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Thr</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0.10</th>\n",
       "      <td>labels_racism_0.1</td>\n",
       "      <td>0.685185</td>\n",
       "      <td>0.719512</td>\n",
       "      <td>0.685979</td>\n",
       "      <td>0.685185</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.10</th>\n",
       "      <td>evaluation_sample</td>\n",
       "      <td>0.627907</td>\n",
       "      <td>0.650602</td>\n",
       "      <td>0.627907</td>\n",
       "      <td>0.627907</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.10</th>\n",
       "      <td>Average</td>\n",
       "      <td>0.656546</td>\n",
       "      <td>0.685057</td>\n",
       "      <td>0.656943</td>\n",
       "      <td>0.656546</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.15</th>\n",
       "      <td>labels_racism_0.1</td>\n",
       "      <td>0.692398</td>\n",
       "      <td>0.795115</td>\n",
       "      <td>0.790823</td>\n",
       "      <td>0.790761</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0.15</th>\n",
       "      <td>evaluation_sample</td>\n",
       "      <td>0.627907</td>\n",
       "      <td>0.739726</td>\n",
       "      <td>0.710526</td>\n",
       "      <td>0.710526</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   Test         1         2         3         4\n",
       "Thr                                                            \n",
       "0.10  labels_racism_0.1  0.685185  0.719512  0.685979  0.685185\n",
       "0.10  evaluation_sample  0.627907  0.650602  0.627907  0.627907\n",
       "0.10            Average  0.656546  0.685057  0.656943  0.656546\n",
       "0.15  labels_racism_0.1  0.692398  0.795115  0.790823  0.790761\n",
       "0.15  evaluation_sample  0.627907  0.739726  0.710526  0.710526"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "thr_analysis = pd.read_csv('./model/thr_analysis_w_m_vote.csv')\n",
    "thr_analysis.set_index('Thr', inplace=True)\n",
    "thr_analysis.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "thr_analysis[\"Test\"].replace({\"labels_racism_0.1\":\"validation\", \n",
    "                              \"evaluation_sample\": \"evaluation\", \"Average\": \"average\"}, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 648x648 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "\n",
    "fig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2,2,sharex=True, sharey=True,figsize=(9,9))\n",
    "\n",
    "thr_analysis.groupby('Test')['1'].plot(legend=True, ylim=[0.4,0.95], xlim=[0.0,1.0], ax=ax1)\n",
    "ax1.set_title('Epoch 1')\n",
    "ax1.set_ylabel('F1-score')\n",
    "thr_analysis.groupby('Test')['2'].plot(legend=True, ylim=[0.4,0.95], xlim=[0.0,1.0], ax=ax2)\n",
    "ax2.set_title('Epoch 2')\n",
    "thr_analysis.groupby('Test')['3'].plot(legend=True, ylim=[0.4,0.95], xlim=[0.0,1.0], ax=ax3)\n",
    "ax3.set_title('Epoch 3')\n",
    "ax3.set_ylabel('F1-score')\n",
    "ax3.set_xlabel('Threshold')\n",
    "thr_analysis.groupby('Test')['4'].plot(legend=True, ylim=[0.4,0.95], xlim=[0.0,1.0], ax=ax4)\n",
    "ax4.set_title('Epoch 4')\n",
    "ax4.set_xlabel('Threshold')\n",
    "\n",
    "plt.tight_layout()\n",
    "fig.savefig('model/thr_analysis_w_m_vote.png')\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Error analysis\n",
    "\n",
    "\n",
    "Evaluation sample translated has the english message and the toxicity scores generated with the src/perspective_api.ipynb."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluation sample: 59\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>message</th>\n",
       "      <th>english_message</th>\n",
       "      <th>label</th>\n",
       "      <th>toxicity</th>\n",
       "      <th>severe_toxicity</th>\n",
       "      <th>raw</th>\n",
       "      <th>m_vote_nonstrict</th>\n",
       "      <th>w_m_vote_nonstrict</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Feméride' (DCV): hoy se cumplen 60 años del estreno de \"libertad bajo palabra\" de Alfredo Bettanín, con Lautaro Murúa, Ubaldo Martínez, Enrique Fava y Raúl Parini. Mil vidas en veinticuatro horas. Subastas, parques negros, conductas más o menos honorables. Regreso y hasta pronto.</td>\n",
       "      <td>Feméride' (DCV): today marks the 60th anniversary of the premiere of \"freedom on parole\" by Alfredo Bettanín, with Lautaro Murúa, Ubaldo Martínez, Enrique Fava and Raúl Parini. A thousand lives in twenty-four hours. Auctions, black parks, more or less honorable conduct. I'll be back and see you soon.</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>0.310055</td>\n",
       "      <td>0.310310</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>non-racist</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Pegan una paliza a dos policías en Zaragoza Pegan paliza a jóvenes en Tenerife Intento de atraco a taxista en Barcelona Roba y choca un coche contra otros 6 turismos en Barcelona Pero tranquilos que no tenemos problemas con los magrebíes, vienen a pagar pensiones.</td>\n",
       "      <td>They beat up two policemen in Zaragoza They beat up young people in Tenerife Attempted robbery of a taxi driver in Barcelona Steals and crashes a car against 6 other cars in Barcelona But don't worry, we don't have problems with the North Africans, they come to pay pensions.</td>\n",
       "      <td>racist</td>\n",
       "      <td>0.466007</td>\n",
       "      <td>0.451909</td>\n",
       "      <td>racist</td>\n",
       "      <td>racist</td>\n",
       "      <td>racist</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>@A_ausdenRuthen @SimonLevyMx Pinches musulmanes, todos son iguales. 😂</td>\n",
       "      <td>@A_ausdenRuthen @SimonLevyMx Fucking Muslims, they're all the same. 😂</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>0.923079</td>\n",
       "      <td>0.914049</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>non-racist</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>@Puri95702837 @yeilouondeflo @PLAyer__slaYER @liv_june @NoaGresiva No se como ha hecho esa relación, yo hablaba respecto a menas e inmigrantes ilegales, no otra cisa</td>\n",
       "      <td>@Puri95702837 @yeilouondeflo @PLAyer__slaYER @liv_june @NoaGresiva I don't know how he has made that relationship, I was talking about ores and illegal immigrants, not another cisa</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>0.310055</td>\n",
       "      <td>0.297385</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>racist</td>\n",
       "      <td>non-racist</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>@HamzaMaca a quien juzgo? a ti? has empezado a juzgar que hace mal o bien mi padre sin saber nada de él cosa que no te incumbe y has opinado igual yo solo te he dicho que los musulmanes que juzgan a otros no son buenos musulmanes</td>\n",
       "      <td>@HamzaMaca who do I judge? to you? you have begun to judge what my father is doing wrong or right without knowing anything about him which is none of your business and you have the same opinion I have only told you that Muslims who judge others are not good Muslims</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>0.417203</td>\n",
       "      <td>0.202464</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>racist</td>\n",
       "      <td>non-racist</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                                                                                                                                                                                                                    message  \\\n",
       "0  Feméride' (DCV): hoy se cumplen 60 años del estreno de \"libertad bajo palabra\" de Alfredo Bettanín, con Lautaro Murúa, Ubaldo Martínez, Enrique Fava y Raúl Parini. Mil vidas en veinticuatro horas. Subastas, parques negros, conductas más o menos honorables. Regreso y hasta pronto.   \n",
       "1                  Pegan una paliza a dos policías en Zaragoza Pegan paliza a jóvenes en Tenerife Intento de atraco a taxista en Barcelona Roba y choca un coche contra otros 6 turismos en Barcelona Pero tranquilos que no tenemos problemas con los magrebíes, vienen a pagar pensiones.   \n",
       "2                                                                                                                                                                                                                     @A_ausdenRuthen @SimonLevyMx Pinches musulmanes, todos son iguales. 😂   \n",
       "3                                                                                                                     @Puri95702837 @yeilouondeflo @PLAyer__slaYER @liv_june @NoaGresiva No se como ha hecho esa relación, yo hablaba respecto a menas e inmigrantes ilegales, no otra cisa   \n",
       "4                                                     @HamzaMaca a quien juzgo? a ti? has empezado a juzgar que hace mal o bien mi padre sin saber nada de él cosa que no te incumbe y has opinado igual yo solo te he dicho que los musulmanes que juzgan a otros no son buenos musulmanes   \n",
       "\n",
       "                                                                                                                                                                                                                                                                                                 english_message  \\\n",
       "0  Feméride' (DCV): today marks the 60th anniversary of the premiere of \"freedom on parole\" by Alfredo Bettanín, with Lautaro Murúa, Ubaldo Martínez, Enrique Fava and Raúl Parini. A thousand lives in twenty-four hours. Auctions, black parks, more or less honorable conduct. I'll be back and see you soon.   \n",
       "1                            They beat up two policemen in Zaragoza They beat up young people in Tenerife Attempted robbery of a taxi driver in Barcelona Steals and crashes a car against 6 other cars in Barcelona But don't worry, we don't have problems with the North Africans, they come to pay pensions.   \n",
       "2                                                                                                                                                                                                                                          @A_ausdenRuthen @SimonLevyMx Fucking Muslims, they're all the same. 😂   \n",
       "3                                                                                                                           @Puri95702837 @yeilouondeflo @PLAyer__slaYER @liv_june @NoaGresiva I don't know how he has made that relationship, I was talking about ores and illegal immigrants, not another cisa   \n",
       "4                                      @HamzaMaca who do I judge? to you? you have begun to judge what my father is doing wrong or right without knowing anything about him which is none of your business and you have the same opinion I have only told you that Muslims who judge others are not good Muslims   \n",
       "\n",
       "        label  toxicity  severe_toxicity         raw m_vote_nonstrict  \\\n",
       "0  non-racist  0.310055         0.310310  non-racist       non-racist   \n",
       "1      racist  0.466007         0.451909      racist           racist   \n",
       "2  non-racist  0.923079         0.914049  non-racist       non-racist   \n",
       "3  non-racist  0.310055         0.297385  non-racist           racist   \n",
       "4  non-racist  0.417203         0.202464  non-racist           racist   \n",
       "\n",
       "  w_m_vote_nonstrict  \n",
       "0         non-racist  \n",
       "1             racist  \n",
       "2         non-racist  \n",
       "3         non-racist  \n",
       "4         non-racist  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "IMPORT_PATH = os.path.join(\"data\", \"toxicity_scores\")\n",
    "EVAL_PATH = os.path.join(IMPORT_PATH, \"evaluation_sample_translated.csv\")\n",
    "\n",
    "pred_tags = ['raw', 'm_vote_nonstrict', 'w_m_vote_nonstrict']\n",
    "thr = 0.8\n",
    "\n",
    "\n",
    "eval_df = load_dataset(EVAL_PATH)\n",
    "\n",
    "predictions = {}\n",
    "for pred_tag in pred_tags:\n",
    "    predictions[pred_tag] = load_dataset(os.path.join(IMPORT_PATH, f\"evaluation_sample_{pred_tag}.csv\"))\n",
    "    eval_df[pred_tag] = predictions[pred_tag][\"predicted_label\"]\n",
    "\n",
    "n_total = eval_df.shape[0]\n",
    "print(f\"Evaluation sample: {n_total}\")\n",
    "eval_df.head()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## False positives"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "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>FP</th>\n",
       "      <th>High Tox</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>raw</th>\n",
       "      <td>3.389831</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>m_vote_nonstrict</th>\n",
       "      <td>11.864407</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>w_m_vote_nonstrict</th>\n",
       "      <td>10.169492</td>\n",
       "      <td>1.694915</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                           FP  High Tox\n",
       "raw                  3.389831  0.000000\n",
       "m_vote_nonstrict    11.864407  0.000000\n",
       "w_m_vote_nonstrict  10.169492  1.694915"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "fp_table = {}\n",
    "fp_index = {}\n",
    "\n",
    "for pred_tag in pred_tags:\n",
    "    \n",
    "    fp = eval_df.loc[(eval_df.label==\"non-racist\")&(eval_df[pred_tag]==\"racist\")]\n",
    "    fp_tox = fp.loc[fp.toxicity>thr]\n",
    "\n",
    "    # save percentage of FP and FP with high toxicity\n",
    "    fp_table[pred_tag] = {'FP':(fp.shape[0]/n_total)*100, \n",
    "                            'High Tox':(fp_tox.shape[0]/n_total)*100}\n",
    "\n",
    "    # save indexexs\n",
    "    fp_index[pred_tag] = {'FP': fp.index.to_list(), 'High Tox': fp_tox.index.to_list()}\n",
    "\n",
    "fp_table = pd.DataFrame.from_dict(fp_table).T\n",
    "fp_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>message</th>\n",
       "      <th>english_message</th>\n",
       "      <th>label</th>\n",
       "      <th>toxicity</th>\n",
       "      <th>severe_toxicity</th>\n",
       "      <th>raw</th>\n",
       "      <th>m_vote_nonstrict</th>\n",
       "      <th>w_m_vote_nonstrict</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>Otra bufonada del ladrón y narcocriminal colombiano @NicolasMaduro Moros.</td>\n",
       "      <td>Another buffoonery from the Colombian thief and drug criminal @NicolasMaduro Moros.</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>0.92509</td>\n",
       "      <td>0.920679</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>racist</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                      message  \\\n",
       "18  Otra bufonada del ladrón y narcocriminal colombiano @NicolasMaduro Moros.   \n",
       "\n",
       "                                                                        english_message  \\\n",
       "18  Another buffoonery from the Colombian thief and drug criminal @NicolasMaduro Moros.   \n",
       "\n",
       "         label  toxicity  severe_toxicity         raw m_vote_nonstrict  \\\n",
       "18  non-racist   0.92509         0.920679  non-racist       non-racist   \n",
       "\n",
       "   w_m_vote_nonstrict  \n",
       "18             racist  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Only w m vote has FP according to the ground truth with a high toxicity score, let's see\n",
    "\n",
    "eval_df.loc[eval_df.index.isin(fp_index[pred_tags[-1]]['High Tox'])]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## False negatives"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>FN</th>\n",
       "      <th>Low Tox</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>raw</th>\n",
       "      <td>5.084746</td>\n",
       "      <td>1.694915</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>m_vote_nonstrict</th>\n",
       "      <td>1.694915</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>w_m_vote_nonstrict</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                          FN   Low Tox\n",
       "raw                 5.084746  1.694915\n",
       "m_vote_nonstrict    1.694915  0.000000\n",
       "w_m_vote_nonstrict  0.000000  0.000000"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fn_table = {}\n",
    "fn_index = {}\n",
    "\n",
    "for pred_tag in pred_tags:\n",
    "    \n",
    "    fn = eval_df.loc[(eval_df.label==\"racist\")&(eval_df[pred_tag]==\"non-racist\")]\n",
    "    fn_tox = fn.loc[fn.toxicity<(1.0-thr)]\n",
    "\n",
    "    # save percentage of FP and FP with high toxicity\n",
    "    fn_table[pred_tag] = {'FN':(fn.shape[0]/n_total)*100, \n",
    "                            'Low Tox':(fn_tox.shape[0]/n_total)*100}\n",
    "\n",
    "    # save indexexs\n",
    "    fn_index[pred_tag] = {'FN': fn.index.to_list(), 'Low Tox': fn_tox.index.to_list()}\n",
    "\n",
    "fn_table = pd.DataFrame.from_dict(fn_table).T\n",
    "fn_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>message</th>\n",
       "      <th>english_message</th>\n",
       "      <th>label</th>\n",
       "      <th>toxicity</th>\n",
       "      <th>severe_toxicity</th>\n",
       "      <th>raw</th>\n",
       "      <th>m_vote_nonstrict</th>\n",
       "      <th>w_m_vote_nonstrict</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>Detenidos 12 polizones marroquíes que alcanzaron Motril (Granada) en un ferri desde Melilla</td>\n",
       "      <td>Arrested 12 Moroccan stowaways who reached Motril (Granada) on a ferry from Melilla</td>\n",
       "      <td>racist</td>\n",
       "      <td>0.142734</td>\n",
       "      <td>0.281967</td>\n",
       "      <td>non-racist</td>\n",
       "      <td>racist</td>\n",
       "      <td>racist</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                        message  \\\n",
       "45  Detenidos 12 polizones marroquíes que alcanzaron Motril (Granada) en un ferri desde Melilla   \n",
       "\n",
       "                                                                        english_message  \\\n",
       "45  Arrested 12 Moroccan stowaways who reached Motril (Granada) on a ferry from Melilla   \n",
       "\n",
       "     label  toxicity  severe_toxicity         raw m_vote_nonstrict  \\\n",
       "45  racist  0.142734         0.281967  non-racist           racist   \n",
       "\n",
       "   w_m_vote_nonstrict  \n",
       "45             racist  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Only the weighted model has no FN, but there are some with low toxicity in the raw, let's see\n",
    "eval_df.loc[eval_df.index.isin(fn_index[pred_tags[0]]['Low Tox'])]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Interesting case of implicit bias:implicit racism in news (which is hard for annotators to detect, and even for the Perspective API\n",
    "# )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "7d2d92c880370c324d499b28a936b865890665b043fe3fd75e8367b595773a30"
  },
  "kernelspec": {
   "display_name": "Python 3.8.5 ('base')",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.12"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}