deeplow_paper_results_new.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Experiment Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Result per Task\n",
    "\n",
    "### English"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "baseline = [52.1, 50.7, 50.5, 63.5, 50.0, 54.9, 54.1]\n",
    "rnn = [50.0, 50.1, 50.2 ,50.0, 50.1 ,84.4 ,50.0]\n",
    "gru = [50.4, 50.1, 52.7, 66.1 , 50.0, 82.7, 53.1]\n",
    "lstm = [50.4, 50.8, 51.3 , 50.1 , 50.0, 75.1 , 50.0]\n",
    "bert = [99.7, 100, 100, 84.4, 99.7, 87.5, 90.0]\n",
    "index = [\"task1\", \"task2\", \"task4\", \"task3\", \"task5\", \"task6\", \"task7\"]\n",
    "results_eng = np.stack([baseline, rnn, gru, lstm, bert]).T\n",
    "columns = [\"Baseline\", \"RNN\", \"GRU\", \"LSTM\", \"BERT\"]\n",
    "df_eng = pd.DataFrame(results_eng,columns=columns,  index=index).sort_index()\n",
    "\n",
    "df_eng.loc[:,\"Recurrent Models\"] = df_eng.iloc[:,1:4].mean(1)\n",
    "df = df_eng[[\"Baseline\", \"Recurrent Models\", \"BERT\"]]\n",
    "df = pd.melt(df).rename(columns={\"variable\":\"model type\", \"value\":\"accuracy\"})\n",
    "\n",
    "ax, fig = plt.subplots(figsize=(10,5))\n",
    "sns.boxplot(x=\"model type\", y=\"accuracy\", data=df);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Portuguese"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmQAAAE9CAYAAACleH4eAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAcLElEQVR4nO3de5SddX3v8fcnXEqAyjXGlDTG41AttdVCyrLVWhSk0nqKWqvSi1FUzmk5RNtylvSUs9RWWzy9nc65tFJF46V4qwjiQYmpWrVSCRcBEchUiR3KJQGihAQIyff8sZ/RSUhgJ2Tv32T2+7XWrL2f334u35n1zMxn/36//TypKiRJktTOnNYFSJIkjToDmSRJUmMGMkmSpMYMZJIkSY0ZyCRJkhozkEmSJDW2b+sCHo8jjzyyFi9e3LoMSZKkx3TVVVetq6p5O3ptrw5kixcvZtWqVa3LkCRJekxJ1uzsNYcsJUmSGjOQSZIkNWYgkyRJasxAJkmS1JiBTJIkqTEDmSRJUmMGMkmSpMYGFsiSXJDkriQ3TGs7PMmKJKu7x8O69iQZTzKR5Lokxw6qLkmSpJlmkD1k7wNetF3bOcDKqjoaWNktA5wCHN19nQH8zQDrkiRJmlEGdqX+qvqnJIu3az4VOKF7vhz4AvDmrv39VVXAFUkOTbKgqm4fVH2SJA3D+Pg4ExMTrcvYxuTkJAALFy5sXMm2xsbGWLZsWesymhj2rZPmTwtZdwDzu+dHAf82bb3Jru0RgSzJGfR60Vi0aNHgKpUk7XVOP/10br99Zr2Xf/DBB9m6dWvrMrYxVc8999zTuJJt3XDDDVx22WWty9jGggULuOCCCwZ+nGb3sqyqSlK7sd35wPkAS5Ys2eXtJUmz1/r169mwccPMulNzgH1aF7FjW/bZ0rqEbWxhC5sf2ty6jB94uHdODcOwT9k7p4YikywA7urabwN+dNp6C7s2SZL6tnDhQtZmLVtPmFk9Uto7zfnCHBYeNZxh3WFf9uISYGn3fClw8bT2V3eftnw28F3nj0mSpFExsB6yJBfSm8B/ZJJJ4C3AecBHk7wOWAO8olv9/wG/BEwAG4HXDqouSZKkmWaQn7I8bScvnbiDdQs4c1C1SJIkzWReqV+SJKkxA5kkSVJjBjJJkqTGDGSSJEmNGcgkSZIaM5BJkiQ1ZiCTJElqzEAmSZLUmIFMkiSpMQOZJElSYwYySZKkxgxkkiRJjRnIJEmSGjOQSZIkNWYgkyRJasxAJkmS1JiBTJIkqTEDmSRJUmMGMkmSpMYMZJIkSY0ZyCRJkhozkEmSJDVmIJMkSWrMQCZJktSYgUySJKkxA5kkSVJjBjJJkqTGDGSSJEmNGcgkSZIaM5BJkiQ1ZiCTJElqzEAmSZLUmIFMkiSpMQOZJElSYwYySZKkxgxkkiRJjRnIJEmSGjOQSZIkNWYgkyRJasxAJkmS1FiTQJbkjUluSPKNJG/q2g5PsiLJ6u7xsBa1SZIkDdvQA1mSZwBvAI4Hngm8OMkYcA6wsqqOBlZ2y5IkSbNeix6yHwf+pao2VtXDwBeBlwGnAsu7dZYDL2lQmyRJ0tC1CGQ3AD+f5IgkBwK/BPwoML+qbu/WuQOY36A2SZKkodt32Aesqm8meSdwOXA/cC2wZbt1KkntaPskZwBnACxatGjA1UqSJA1ek0n9VfWeqjquqp4H3AvcAtyZZAFA93jXTrY9v6qWVNWSefPmDa9oSZKkAWn1Kcsndo+L6M0f+3vgEmBpt8pS4OIWtUmSJA3b0IcsO/+Q5AhgM3BmVa1Pch7w0SSvA9YAr2hUmyRJ0lA1CWRV9fM7aLsbOLFBOZIkSU15pX5JkqTGDGSSJEmNGcgkSZIaM5BJkiQ1ZiCTJElqzEAmSZLUmIFMkiSpMQOZJElSYwYySZKkxgxkkiRJjRnIJEmSGjOQSZIkNWYgkyRJasxAJkmS1JiBTJIkqTEDmSRJUmMGMkmSpMYMZJIkSY0ZyCRJkhozkEmSJDVmIJMkSWrMQCZJktSYgUySJKkxA5kkSVJjBjJJkqTGDGSSJEmNGcgkSZIaM5BJkiQ1ZiCTJElqzEAmSZLUmIFMkiSpMQOZJElSYwYySZKkxgxkkiRJjRnIJEmSGjOQSZIkNWYgkyRJasxAJkmS1Ni+rQuQJGmPWg9zvmB/w6Pa0D0e3LSKmW89cNRwDmUgkyTNGmNjY61L2CusXr0agKOPOrpxJTPcUcM7p5oEsiS/C7weKOB64LXAAuDDwBHAVcBvVdVDLeqTJO2dli1b1rqEvcLUz2l8fLxxJZoy9D7dJEcBy4AlVfUMYB/gVcA7gb+qqjHgXuB1w65NkiSphVaD7PsCc5PsCxwI3A68APh49/py4CWNapMkSRqqoQeyqroN+HPgO/SC2HfpDVGur6qHu9UmGdo0OkmSpLZaDFkeBpwKPAX4EeAg4EW7sP0ZSVYlWbV27doBVSlJkjQ8LYYsTwK+XVVrq2oz8AngOcCh3RAmwELgth1tXFXnV9WSqloyb9684VQsSZI0QC0C2XeAZyc5MEmAE4Ebgc8DL+/WWQpc3KA2SZKkoWsxh+xf6E3ev5reJS/mAOcDbwZ+L8kEvUtfvGfYtUmSJLXQ5DpkVfUW4C3bNX8LOL5BOZIkSU31FciSfIJej9VlVbV1sCVJkjR7jI+PMzEx0bqMbUxdqX+mXUh3bGxsxtU0LP0OWf5f4NeB1UnOS/K0AdYkSZIGaO7cucydO7d1GZqmrx6yqvoc8LkkhwCndc//Dfg74IPdpyUlSdJ2RrXHR7um70n9SY4AXkPvHpTXAH8NHAusGEhlkiRJI6LfOWQXAU8DPgD8x6q6vXvpI0lWDao4SZKkUdDvpyzHq+rzO3qhqpbswXokSZJGTr9DlsckOXRqIclhSX5nQDVJkiSNlH4D2Ruqav3UQlXdC7xhMCVJkiSNln4D2T7dbY4ASLIPsP9gSpIkSRot/c4h+wy9Cfzv6pb/U9cmSZKkx6nfQPZmeiHst7vlFcC7B1KRJEnSiOn3wrBbgb/pviRJkrQH9XsdsqOBPwWOAQ6Yaq+q/zCguiRJkkZGv5P630uvd+xh4PnA+4EPDqooSZKkUdJvIJtbVSuBVNWaqnor8MuDK0uSJGl09Dup/8Ekc4DVSf4LcBtw8ODKkiRJGh399pC9ETgQWAYcB/wmsHRQRUmSJI2Sx+wh6y4C+8qqOhvYALx24FVJkiSNkMfsIauqLcBzh1CLJEnSSOp3Dtk1SS4BPgbcP9VYVZ8YSFWSJEkjpN9AdgBwN/CCaW0FGMgkSZIep36v1O+8MUmSpAHp90r976XXI7aNqjp9j1ckSZI0Yvodsrx02vMDgJcC/77ny5EkSRo9/Q5Z/sP05SQXAl8eSEWSJEkjpt8Lw27vaOCJe7IQSZKkUdXvHLL72HYO2R3AmwdSkSRJ0ojpd8jyhwddiCRJ0qjqa8gyyUuTHDJt+dAkLxlcWZIkSaOj3zlkb6mq704tVNV64C2DKUmSJGm09BvIdrRev5fMkCRJ0qPoN5CtSvKXSZ7aff0lcNUgC5MkSRoV/Qays4CHgI8AHwYeAM4cVFGSJEmjpN9PWd4PnDPgWiRJkkZSv5+yXJHk0GnLhyX57ODKkiRJGh39Dlke2X2yEoCquhev1C9pFlq3bh1nnXUWd999d+tSJI2QfgPZ1iSLphaSLGbbK/dL0qywfPlyrrvuOpYvX966FEkjpN9A9ofAl5N8IMkHgS8CfzC4siRp+NatW8dll11GVXHZZZfZSyZpaPoKZFX1GWAJcDNwIfD7wKYB1iVJQ7d8+XKqep3/W7dutZdM0tD0O6n/9cBKekHsbOADwFsHV5YkDd+KFSvYvHkzAJs3b+byyy9vXJGkUdHvkOUbgZ8B1lTV84GfBtY/+iY7luRpSa6d9vW9JG9Kcnj3ac7V3eNhu7N/SdpdL3zhC9lvv/0A2G+//Tj55JMbVyRpVPQbyB6oqgcAkvxQVd0EPG13DlhVN1fVs6rqWcBxwEbgInrXOVtZVUfT643zumeShmrp0qUkAWDOnDksXbq0cUWSRkW/gWyyuw7ZJ4EVSS4G1uyB458I/GtVrQFOBaYmbCwHXrIH9i9JfTvyyCM55ZRTSMIpp5zCEUcc0bokSSOi3yv1v7R7+tYknwcOAT6zB47/KnofEgCYX1W3d8/vAObvgf1L0i5ZunQpt956q71jkoYqU58oGvqBk/2Bfwd+oqruTLK+qqbfDeDeqnrEPLIkZwBnACxatOi4NWv2REedJEnSYCW5qqqW7Oi1focsB+EU4OqqurNbvjPJAoDu8a4dbVRV51fVkqpaMm/evCGVKkmSNDgtA9lp/GC4EuASYGqMYClw8dArkiRJaqBJIEtyEPBC4BPTms8DXphkNXBStyxJkjTr9TWpf0+rqvuBI7Zru5vepy4lSZJGSsshS0mSJGEgkyRJas5AJkmS1JiBTJIkqTEDmSRJUmMGMkmSpMYMZJIkSY0ZyCRJkhozkEmSJDVmIJMkSWrMQCZJktSYgUySJKkxA5kkSVJjBjJJkqTGDGSSJEmNGcgkSZIaM5BJkiQ1ZiCTJElqzEAmSZLUmIFMkiSpMQOZJElSYwYySZKkxgxkkiRJjRnIJEmSGjOQSZIkNWYgkyRJasxAJkmS1JiBTJIkqTEDmSRJUmMGMkmSpMYMZJIkSY0ZyCRJkhozkEmSJDVmIJMkSWrMQCZJktTYvq0L0I6Nj48zMTHRuoxtTE5OArBw4cLGlWxrbGyMZcuWtS5DkqTdZiBT3zZt2tS6BEmSZiUD2Qw1E3t8pmoaHx9vXIkkSbOLc8gkSZIaM5BJkiQ11iSQJTk0yceT3JTkm0l+NsnhSVYkWd09HtaiNkmSpGFr1UP218BnqurpwDOBbwLnACur6mhgZbcsSZI06w09kCU5BHge8B6AqnqoqtYDpwLLu9WWAy8Zdm2SJEkttOghewqwFnhvkmuSvDvJQcD8qrq9W+cOYH6D2iRJkoauRSDbFzgW+Juq+mngfrYbnqyqAmpHGyc5I8mqJKvWrl078GIlSZIGrUUgmwQmq+pfuuWP0wtodyZZANA93rWjjavq/KpaUlVL5s2bN5SCJUmSBmnogayq7gD+LcnTuqYTgRuBS4ClXdtS4OJh1yZJktRCqyv1nwV8KMn+wLeA19ILhx9N8jpgDfCKRrVJkiQNVZNAVlXXAkt28NKJw65FkiSpNa/UL0mS1JiBTJIkqTEDmSRJUmMGMkmSpMYMZJIkSY0ZyCRJkhozkEmSJDVmIJMkSWqs1ZX6Z5Tx8XEmJiZalzHjrV69GoBly5Y1rmTmGxsb8+ckSeqbgQyYmJjgmutvZOuBh7cuZUbLQwXAVf96R+NKZrY5G+9pXcJeYya+GZqcnARg4cKFjSvZliFfmt0MZJ2tBx7OA8e8uHUZmgUOuPHS1iXocdi0aVPrEiSNIAOZpGZmYo/PVE3j4+ONK5E0SpzUL0mS1JiBTJIkqTEDmSRJUmMGMkmSpMYMZJIkSY0ZyCRJkhozkEmSJDVmIJMkSWrMQCZJktSYgUySJKkxA5kkSVJj3stSGhHj4+NMTEy0LmPGW716NTAz77M504yNjflzkvYQAxkwOTnJnI3f5YAbL21dimaBORvvZnLy4dZlPMLExAS33HA1iw7e0rqUGW3/zb2BgwduvbJxJTPbdzbs07oEaVYxkEkjZNHBWzh3yYbWZWgWePuqg1uXIM0qBjJg4cKF3PngvjxwzItbl6JZ4IAbL2Xhwie1LkOStBdxUr8kSVJjBjJJkqTGDGSSJEmNGcgkSZIaM5BJkiQ1ZiCTJElqzEAmSZLUmIFMkiSpMQOZJElSYwYySZKkxgxkkiRJjRnIJEmSGjOQSZIkNWYgkyRJamzfFgdNcitwH7AFeLiqliQ5HPgIsBi4FXhFVd07rJrmbLyHA268dFiH2yvlge8BUAc8oXElM9ucjfcAT2pdhiRpL9IkkHWeX1Xrpi2fA6ysqvOSnNMtv3kYhYyNjQ3jMHu91avvA+Dopxo2Ht2TZuQ5NTk5yf337cPbVx3cuhTNAmvu24eDJidblyHNGi0D2fZOBU7oni8HvsCQAtmyZcuGcZi93tTPaXx8vHEl2l0Pbglr7tundRnft3lr2Fqtq9g7zAnsN2fm/LAe3BIOal2ENIu0CmQFXJ6kgHdV1fnA/Kq6vXv9DmD+jjZMcgZwBsCiRYuGUWsT4+PjTExMtC5jG6tXrwZmXoAdGxubcTXNRCeccMKMO6cmJyfZtGlT6zL2CnPnzmXhwoWty9jGTOwJlvZWrQLZc6vqtiRPBFYkuWn6i1VVXVh7hC68nQ+wZMmSmfN2cQTMnTu3dQl6HAytkjRzNQlkVXVb93hXkouA44E7kyyoqtuTLADualHbTOE/T0mSRsfQL3uR5KAkPzz1HDgZuAG4BFjarbYUuHjYtUmSJLXQoodsPnBRkqnj/31VfSbJlcBHk7wOWAO8okFtkiRJQzf0QFZV3wKeuYP2u4ETh12PJElSa16pX5IkqTEDmSRJUmMGMkmSpMYMZJIkSY0ZyCRJkhozkEmSJDVmIJMkSWrMQCZJktRYqvbe+3MnWUvvqv4aniOBda2LkAbM81yjwPN8+J5cVfN29MJeHcg0fElWVdWS1nVIg+R5rlHgeT6zOGQpSZLUmIFMkiSpMQOZdtX5rQuQhsDzXKPA83wGcQ6ZJElSY/aQSZIkNWYgm+WSbElybZKvJ7k6yc/t4f2/L8nLu+fvTnLMnty/Zo9p5+INST6V5NDWNT2WJCfs7HcmyWuSVJKTprW9pGt7+S4e49LHu460s7/3SRYn2dS9NvX16u61W5Ncn+S6JF9M8uQkR0xb744kt01b3r/tdzl77du6AA3cpqp6FkCSXwT+FPiFQRyoql4/iP1q1ph+Li4HzgTeMYgDJdm3qh6ethx6UzS27uKuTgA2AP+8k9evB14FfK5bPg34+i4eQ9pTHu3v/b9OvbYDz6+qdUneBpxbVW8ApvbzVmBDVf35YEuXPWSj5QnAvQBJDk6ysnsXdX2SU7v2g5J8unuHdUOSV3btx3Xvnq5K8tkkC7bfeZIvJFnSPd+Q5B3dfq5IMr9rn5fkH5Jc2X09Z2jfvWaSrwJHTS0k+a/d+XBd909hqv3VXdvXk3yga/t+r2y3vKF7PCHJl5JcAtzY9QrcnOT9wA3AjyY5OclXu/P+Y0kO7ra9Ncnbpv0+PD3JYuA/A7/b9Qz8/A6+jy8BxyfZr9vXGHDttNpOTHJNt88LkvxQ1/6iJDcluRp42bT1D+rW+1q33anbHzDJL0zrrbgmyQ/v8k9fo+D7f+93wTa/lxoue8hmv7lJrgUOABYAL+jaHwBeWlXfS3IkcEX3j+xFwL9X1S8DJDkkyX7A/wJOraq1XUh7B3D6oxz3IOCKqvrDJP8DeAPwduCvgb+qqi8nWQR8FvjxPf1Na+ZKsg9wIvCebvlk4GjgeCDAJUmeB9wNnAv8XPfu/fA+dn8s8Iyq+nYXqI4GllbVFd15fi5wUlXdn+TNwO8Bf9Rtu66qjk3yO8DZVfX6JH/Lo/cOFL3esV8EDgEuAZ7SfV8HAO8DTqyqW7pg+NvdPv+O3u/iBPCRafv7Q+Afq+r09IZ0v5bkc2zrbODMqvpKFwIf6OPnotGws7/3AE/tXptyVlV9abvtXwR8csA1aicMZLPf9C7snwXen+QZ9P7x/Un3j28rvXdF8+kNwfxFkncCl1bVl7r1nwGs6I38sA9w+2Mc9yFgas7LVcALu+cnAcd0+wF4QpKDq2rD4/9WNcNN/bM4CvgmsKJrP7n7uqZbPphekHom8LGqWgdQVff0cYyvVdW3py2vqaoruufPBo4BvtKdf/vT6xGY8onu8Sqm9Vr14cPAMnqB7PeB/9a1Pw34dlXd0i1PDdN+oWtfDZDkg8AZ3TonA7+S5Oxu+QBg0XbH+wrwl0k+BHyiqiZ3oVbNbjv7ew+PPmT5+e4Nzwbgvw+hTu2AgWyEVNVXu16CecAvdY/HVdXmJLcCB3Tv5I/tXn97kpXARcA3qupnd+Fwm+sH11TZwg/OtTnAs6vKd/WjZ1NVPSvJgfR6Rs8Exum9OfjTqnrX9JWTnLWT/TxMN90iyRx6wWrK/dutO305wIqqOm0n+32we5x+vj6mqvpakp8ENna/P/1uuiMBfrWqbt6msRvy7453XpJP0/sd/UqSX6yqmx7PQTX7bPf3/rE8H1gPfAh4G72eYw2Zc8hGSJKn0+vdupveu/m7ujD2fODJ3To/Qu8fyweBP6M3BHQzMK97x0U3X+YndrOMy4Hv/6NNsrN3bJqlqmojvR6l30+yL71wdvq0+VxHJXki8I/AryU5omufGrK8FTiue/4rwH59HvoK4DlJxrr9HZTkxx5jm/uAfuZoncMPesam3Awsnjoe8FvAF4Gbuvandu3TA+JngbPSpbokP739gZI8taqur6p3AlcCT++jPo2Y7f7eP6buQzBvAl7d5/QA7WH2kM1+U8NE0Hv3vbSqtnTDHZ9Kcj2wit4/CYCfBP4syVZgM/DbVfVQN4l6PMkh9M6b/wl8YzfqWQb8nyTXdfv5J3oTpzVCquqa7hw4rao+kOTHga92OWQD8JtV9Y0k7wC+mGQLvSHN19Cbf3Vxkq8Dn+GRvWI7O+baJK8BLpyaXE9vTtktO9+KTwEf7ybX72jOzdS+L9tB2wNJXgt8rAueVwJ/W1UPJjkD+HSSjfQ+GDAV+v6Y3u/WdV3v37eBF2+36zd1b6K20vsdfMSxNbJ29vceHjmH7IKqGp++cVXdnuRCer3XfzyUivV9XqlfkiSpMYcsJUmSGjOQSZIkNWYgkyRJasxAJkmS1JiBTJIkqTEDmaSRk969K4/cnXWSbH+9MUl63AxkkrRrDGSS9jgDmaQZL8niJDcleV+SW5J8KMlJSb6SZHWS47v1Dk/yySTXJbkiyU917UckuTzJN5K8m95FM6f2/ZtJvpbk2iTv6m5+vrM6zqO7+GZXwx8ledO019+R5I1JTkjyT0k+neTmJH/bXeiVJCcn+WqSq5N8bOoOBZJGm4FM0t5iDPgLercKejrw68BzgbP5Qa/V24Brquqnurb3d+1vAb5cVT9B796siwC6OwS8EnhOd+PlLcBv7KyAqjqH7p6cVfUbwAXAq7t9zQFeBXywW/14ercJOwZ4KvCybgj0XOCkqjqW3l0yvG+gJG+dJGmv8e2quh4gyTeAlVVV3e2/FnfrPBf4VYCq+seuZ+wJwPOAl3Xtn05yb7f+ifTui3lld3uZucBd/RZUVbcmubu75+R8emHw7m5fX6uqb3X1XtjV9gC9gPaVbp39ga/uzg9D0uxiIJO0t3hw2vOt05a3svt/ywIsr6o/eBx1vZvePTafRK/HbMr296Wr7ngrquo0JGkahywlzSZfohtyTHICsK6qvkfvJva/3rWfAhzWrb8SeHmSJ3avHZ7kyY9xjM1J9pu2fBHwIuBngM9Oaz8+yVO6ocxXAl8GrgCek2SsO95BSX5sd79ZSbOHPWSSZpO3AhckuQ7YCCzt2t8GXNgNdf4z8B2AqroxybnA5V1w2gycCax5lGOcD1yX5Oqq+o2qeijJ54H1VbVl2npXAv+b3ty3zwMXVdXWJK/pavmhbr1zgVse7zcuae+Wqu171SVJ/eqC3NXAr1XV6q7tBODsqnpxy9ok7T0cspSk3ZTkGGCC3gcMVreuR9Leyx4ySZKkxuwhkyRJasxAJkmS1JiBTJIkqTEDmSRJUmMGMkmSpMYMZJIkSY39f7xW/Yr72oK8AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "baseline = [53.9, 49.8, 50.9, 61.7, 49.9, 58.9, 55.4]\n",
    "rnn = [50.1, 50.0, 50.0, 50.0, 50.1, 66.4, 51.6]\n",
    "gru = [50.2, 50.0, 50.4, 70.6, 50.8, 79.7, 51.6]\n",
    "lstm = [50.0, 50.0, 50.0, 50.1, 50.0, 67.2, 51.1]\n",
    "bert = [99.9, 99.9, 100, 78.7, 99.8, 79.1, 82.7]\n",
    "\n",
    "index = [\"task1\", \"task2\", \"task4\", \"task3\", \"task5\", \"task6\", \"task7\"]\n",
    "results_pt = np.stack([baseline, rnn, gru, lstm, bert]).T\n",
    "columns = [\"Baseline\", \"RNN\", \"GRU\", \"LSTM\", \"BERT\"]\n",
    "df_pf = pd.DataFrame(results_pt,columns=columns,  index=index).sort_index()\n",
    "df_pf.loc[:,\"Recurrent Models\"] = df_pf.iloc[:,1:4].mean(1)\n",
    "\n",
    "df = df_pf[[\"Baseline\", \"Recurrent Models\", \"BERT\"]]\n",
    "df = pd.melt(df).rename(columns={\"variable\":\"model type\", \"value\":\"accuracy\"})\n",
    "\n",
    "ax, fig = plt.subplots(figsize=(10,5))\n",
    "sns.boxplot(x=\"model type\", y=\"accuracy\", data=df);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Using different pre-trained versions of BERT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "results = {}\n",
    "\n",
    "# chi -> pt \n",
    "\n",
    "SN = [50.0, 50.0, 99.2, 98.7, 99.4, 99.4, 98.7, 99.3, 33.5, 99.4]\n",
    "BC = [50.0, 50.0, 50.0, 99.3, 99.6, 99.5, 50.0, 99.4, 99.4, 99.5]\n",
    "DD = [50.0, 50.0, 50.0,99.9, 99.8, 99.7, 99.9, 99.8, 100,100]\n",
    "Q = [50]*7 + [62, 76.1, 50.0]\n",
    "Comp = [50]*7  + [49.6, 50.0, 50]\n",
    "Count = [50, 64.6, 50, 65, 50, 50, 71.3, 66.7, 75.1, 68.9]\n",
    "M = [51.1, 48.9, 51.1, 51.1, 51.1, 54.3, 51.1, 48.9, 48.9, 48.9]\n",
    "\n",
    "bert_chi =np.stack([SN, BC, Q, DD, Comp, Count, M])\n",
    "results[\"BERT_chi\"] = bert_chi.flatten()\n",
    "\n",
    "# eng -> pt\n",
    "\n",
    "SN = [50.0, 50.0, 99.7, 99.9, 50.0, 99.9, 99.8, 99.9, 99.9, 99.9]\n",
    "BC = [99.4, 99.7, 50.0, 99.8, 99.9, 99.8, 99.9, 99.9, 99.9, 99.9]\n",
    "DD = [50.0, 50.0, 50.0, 99.3, 100, 50.0, 100, 100, 50.0,100]\n",
    "Q = [50.0, 5.0, 62.5, 62.7, 82.9, 79.1, 80.6, 79.8, 62.5, 78.7]\n",
    "Comp = [50.0]*5 +[99.8, 99.6,99.0, 100, 99.8]\n",
    "Count = [68.1, 72.2, 75.1, 79.5, 74.8, 75.5, 74.9, 76.0, 79.4, 79.1]\n",
    "M = [48.9, 48.9, 51.1, 51.1, 83.0, 79.6, 51.1, 81.9, 83.5, 82.7]\n",
    "\n",
    "bert_eng =np.stack([SN, BC, Q, DD, Comp, Count, M])\n",
    "results[\"BERT_eng\"] = bert_eng.flatten()\n",
    "\n",
    "# multi -> pt \n",
    "SN = [98.8, 99.5, 50.0, 99.9, 99.7, 99.9, 100, 100, 100, 100]\n",
    "BC = [50.0, 50.0, 99.9, 100, 100, 100, 99.7, 100, 100, 99.9]\n",
    "DD = [99.7, 100, 99.8] + [100]*7\n",
    "Q = [73.9, 62.9, 73.2, 74.2, 75.4, 78.7, 75.4, 77.9, 79.5, 75.4]\n",
    "Comp = [79.5, 98.6, 93.0,99.7, 99.2, 99.2, 96.3, 100, 99.8, 99.5]\n",
    "Count = [64.6, 76.6, 67.6, 66.3, 50.0, 71.8, 81.4, 87.4, 77.7, 82.4]\n",
    "M = [48.9, 51.0, 51.1, 88.3, 83.6, 82.6, 63.6, 85.9, 92.8, 90.3]\n",
    "\n",
    "bert_mult =np.stack([SN, BC, Q, DD, Comp, Count, M])\n",
    "results[\"BERT_mult\"] = bert_mult.flatten()\n",
    "\n",
    "df = pd.DataFrame(results)\n",
    "df = df[[\"BERT_mult\", \"BERT_eng\", \"BERT_chi\"]]\n",
    "df = pd.melt(df).rename(columns={\"variable\":\"Pre-trained model\", \"value\":\"Accuracy\"})\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(10,5))\n",
    "sns.boxplot(x=\"Pre-trained model\", y=\"Accuracy\", data=df, palette=\"tab20\");\n",
    "ax.set_xlabel(\"Pre-trained model\", fontsize=12);\n",
    "ax.set_ylabel(\"Accuracy\", fontsize=12);\n",
    "plt.savefig(\"acc_box_plot_tasks_pt.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}