tutorials/tutorial-medical-classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# IR Lab Tutorial (Research Oriented): Classification of Medical Queries and Documents\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# This is only needed in Google Colab, in the dev container, everything should be installed already\n",
    "!pip3 install tira trectools python-terrier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tira.third_party_integrations import ensure_pyterrier_is_loaded\n",
    "from tira.rest_api_client import Client\n",
    "import pandas as pd\n",
    "\n",
    "import pyterrier as pt\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "ensure_pyterrier_is_loaded()\n",
    "tira = Client()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ClueWeb09 Web Track Results\n",
    "\n",
    "Over all web tracks, 16 queries are health related and 184 are not health-related.\n",
    "\n",
    "Purely lexical models (Dirichlet and BM25) models seem to perform better than semantic neural models on health related topics compared to not health-related topics."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "clueweb_dataset_ids = [\n",
    "    \"irds:clueweb09/en/trec-web-2009\",\n",
    "    \"irds:clueweb09/en/trec-web-2010\",\n",
    "    \"irds:clueweb09/en/trec-web-2011\",\n",
    "    \"irds:clueweb09/en/trec-web-2012\",\n",
    "]\n",
    "all_datasets = [pt.get_dataset(dataset) for dataset in clueweb_dataset_ids]\n",
    "all_topics = [dataset.get_topics(variant='query') for dataset in all_datasets]\n",
    "all_query_health_classifications = [tira.pt.transform_queries('ir-benchmarks/fschlatt/query-health-classification', dataset) for dataset in all_datasets]"
   ]
  },
  {
   "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>qid</th>\n",
       "      <th>query</th>\n",
       "      <th>mean_health_score</th>\n",
       "      <th>median_health_score</th>\n",
       "      <th>mean_medical_score</th>\n",
       "      <th>median_medical_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>196</td>\n",
       "      <td>sore throat</td>\n",
       "      <td>411.7695</td>\n",
       "      <td>411.7695</td>\n",
       "      <td>99.3674</td>\n",
       "      <td>99.3674</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>88</td>\n",
       "      <td>forearm pain</td>\n",
       "      <td>354.0813</td>\n",
       "      <td>354.0813</td>\n",
       "      <td>310.8832</td>\n",
       "      <td>310.8832</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>82</td>\n",
       "      <td>joints</td>\n",
       "      <td>331.5549</td>\n",
       "      <td>331.5549</td>\n",
       "      <td>240.9360</td>\n",
       "      <td>240.9360</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>128</td>\n",
       "      <td>atypical squamous cells</td>\n",
       "      <td>272.0022</td>\n",
       "      <td>277.4749</td>\n",
       "      <td>452.8779</td>\n",
       "      <td>399.4028</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>67</td>\n",
       "      <td>vldl levels</td>\n",
       "      <td>267.6323</td>\n",
       "      <td>267.6323</td>\n",
       "      <td>434.0579</td>\n",
       "      <td>434.0579</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>173</td>\n",
       "      <td>hip fractures</td>\n",
       "      <td>233.0622</td>\n",
       "      <td>233.0622</td>\n",
       "      <td>308.1734</td>\n",
       "      <td>308.1734</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>toilet</td>\n",
       "      <td>226.5945</td>\n",
       "      <td>226.5945</td>\n",
       "      <td>24.5363</td>\n",
       "      <td>24.5363</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>78</td>\n",
       "      <td>dieting</td>\n",
       "      <td>220.2939</td>\n",
       "      <td>220.2939</td>\n",
       "      <td>245.9618</td>\n",
       "      <td>245.9618</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>89</td>\n",
       "      <td>ocd</td>\n",
       "      <td>218.6667</td>\n",
       "      <td>218.6667</td>\n",
       "      <td>328.5885</td>\n",
       "      <td>328.5885</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>30</td>\n",
       "      <td>diabetes education</td>\n",
       "      <td>203.9913</td>\n",
       "      <td>203.9913</td>\n",
       "      <td>243.6227</td>\n",
       "      <td>243.6227</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>63</td>\n",
       "      <td>flushing</td>\n",
       "      <td>201.4836</td>\n",
       "      <td>201.4836</td>\n",
       "      <td>94.9256</td>\n",
       "      <td>94.9256</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>123</td>\n",
       "      <td>von willebrand disease</td>\n",
       "      <td>200.9725</td>\n",
       "      <td>226.8965</td>\n",
       "      <td>279.1960</td>\n",
       "      <td>335.1087</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>168</td>\n",
       "      <td>lipoma</td>\n",
       "      <td>181.9301</td>\n",
       "      <td>181.9301</td>\n",
       "      <td>496.2743</td>\n",
       "      <td>496.2743</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>182</td>\n",
       "      <td>quit smoking</td>\n",
       "      <td>177.5335</td>\n",
       "      <td>177.5335</td>\n",
       "      <td>190.5365</td>\n",
       "      <td>190.5365</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>26</td>\n",
       "      <td>lower heart rate</td>\n",
       "      <td>168.4160</td>\n",
       "      <td>130.3137</td>\n",
       "      <td>198.0859</td>\n",
       "      <td>184.8707</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>57</td>\n",
       "      <td>ct jobs</td>\n",
       "      <td>167.2068</td>\n",
       "      <td>167.2068</td>\n",
       "      <td>179.6955</td>\n",
       "      <td>179.6955</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>125</td>\n",
       "      <td>butter and margarine</td>\n",
       "      <td>136.5154</td>\n",
       "      <td>125.0510</td>\n",
       "      <td>52.6185</td>\n",
       "      <td>54.7226</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44</th>\n",
       "      <td>195</td>\n",
       "      <td>pressure washers</td>\n",
       "      <td>129.7544</td>\n",
       "      <td>129.7544</td>\n",
       "      <td>126.1504</td>\n",
       "      <td>126.1504</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>61</td>\n",
       "      <td>worm</td>\n",
       "      <td>129.3127</td>\n",
       "      <td>129.3127</td>\n",
       "      <td>99.3838</td>\n",
       "      <td>99.3838</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>111</td>\n",
       "      <td>lymphoma in dogs</td>\n",
       "      <td>123.1295</td>\n",
       "      <td>39.4860</td>\n",
       "      <td>228.4353</td>\n",
       "      <td>198.1618</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>121</td>\n",
       "      <td>sit and reach test</td>\n",
       "      <td>121.9814</td>\n",
       "      <td>78.1871</td>\n",
       "      <td>80.8871</td>\n",
       "      <td>54.8896</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>69</td>\n",
       "      <td>sewing instructions</td>\n",
       "      <td>114.4124</td>\n",
       "      <td>114.4124</td>\n",
       "      <td>38.1157</td>\n",
       "      <td>38.1157</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>34</td>\n",
       "      <td>cell phones</td>\n",
       "      <td>114.1950</td>\n",
       "      <td>114.1950</td>\n",
       "      <td>194.8462</td>\n",
       "      <td>194.8462</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>162</td>\n",
       "      <td>dnr</td>\n",
       "      <td>112.1133</td>\n",
       "      <td>112.1133</td>\n",
       "      <td>193.3705</td>\n",
       "      <td>193.3705</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>118</td>\n",
       "      <td>poem in your pocket day</td>\n",
       "      <td>107.4421</td>\n",
       "      <td>34.8868</td>\n",
       "      <td>40.2603</td>\n",
       "      <td>48.4656</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>119</td>\n",
       "      <td>interview thank you</td>\n",
       "      <td>91.8932</td>\n",
       "      <td>1.8060</td>\n",
       "      <td>19.2758</td>\n",
       "      <td>6.6938</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>55</td>\n",
       "      <td>iron</td>\n",
       "      <td>88.1704</td>\n",
       "      <td>88.1704</td>\n",
       "      <td>106.5617</td>\n",
       "      <td>106.5617</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>158</td>\n",
       "      <td>septic system design</td>\n",
       "      <td>87.3629</td>\n",
       "      <td>74.0658</td>\n",
       "      <td>199.0964</td>\n",
       "      <td>117.9455</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>177</td>\n",
       "      <td>best long term care insurance</td>\n",
       "      <td>86.1771</td>\n",
       "      <td>47.4360</td>\n",
       "      <td>83.9122</td>\n",
       "      <td>44.2507</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>104</td>\n",
       "      <td>indiana child support</td>\n",
       "      <td>84.3150</td>\n",
       "      <td>57.3328</td>\n",
       "      <td>56.8610</td>\n",
       "      <td>75.9244</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42</th>\n",
       "      <td>193</td>\n",
       "      <td>dog clean up bags</td>\n",
       "      <td>83.9026</td>\n",
       "      <td>81.7193</td>\n",
       "      <td>43.2959</td>\n",
       "      <td>40.5533</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>175</td>\n",
       "      <td>signs of a heartattack</td>\n",
       "      <td>78.6511</td>\n",
       "      <td>33.5019</td>\n",
       "      <td>66.7424</td>\n",
       "      <td>49.2636</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>152</td>\n",
       "      <td>angular cheilitis</td>\n",
       "      <td>76.9094</td>\n",
       "      <td>76.9094</td>\n",
       "      <td>201.6837</td>\n",
       "      <td>201.6837</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>getting organized</td>\n",
       "      <td>74.7692</td>\n",
       "      <td>74.7692</td>\n",
       "      <td>22.8042</td>\n",
       "      <td>22.8042</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>70</td>\n",
       "      <td>to be or not to be that is the question</td>\n",
       "      <td>68.8508</td>\n",
       "      <td>59.1396</td>\n",
       "      <td>70.9016</td>\n",
       "      <td>80.7897</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>178</td>\n",
       "      <td>pork tenderloin</td>\n",
       "      <td>66.2645</td>\n",
       "      <td>66.2645</td>\n",
       "      <td>41.0316</td>\n",
       "      <td>41.0316</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>83</td>\n",
       "      <td>memory</td>\n",
       "      <td>64.2118</td>\n",
       "      <td>64.2118</td>\n",
       "      <td>145.4128</td>\n",
       "      <td>145.4128</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>117</td>\n",
       "      <td>dangers of asbestos</td>\n",
       "      <td>63.6375</td>\n",
       "      <td>48.6819</td>\n",
       "      <td>89.2323</td>\n",
       "      <td>53.1681</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>132</td>\n",
       "      <td>mothers day songs</td>\n",
       "      <td>60.7894</td>\n",
       "      <td>57.7526</td>\n",
       "      <td>98.9898</td>\n",
       "      <td>52.5622</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42</th>\n",
       "      <td>143</td>\n",
       "      <td>arkadelphia health club</td>\n",
       "      <td>60.6634</td>\n",
       "      <td>0.4916</td>\n",
       "      <td>62.4479</td>\n",
       "      <td>0.5497</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>9</td>\n",
       "      <td>used car parts</td>\n",
       "      <td>57.5809</td>\n",
       "      <td>75.1587</td>\n",
       "      <td>49.4786</td>\n",
       "      <td>32.9048</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>7</td>\n",
       "      <td>air travel information</td>\n",
       "      <td>56.5040</td>\n",
       "      <td>44.4720</td>\n",
       "      <td>55.5602</td>\n",
       "      <td>23.8385</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>161</td>\n",
       "      <td>furniture for small spaces</td>\n",
       "      <td>55.1753</td>\n",
       "      <td>39.9681</td>\n",
       "      <td>52.5583</td>\n",
       "      <td>59.6951</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>189</td>\n",
       "      <td>gs pay rate</td>\n",
       "      <td>54.8410</td>\n",
       "      <td>23.4032</td>\n",
       "      <td>123.0190</td>\n",
       "      <td>101.9324</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>115</td>\n",
       "      <td>pacific northwest laboratory</td>\n",
       "      <td>54.6976</td>\n",
       "      <td>2.7565</td>\n",
       "      <td>70.4378</td>\n",
       "      <td>11.0099</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>59</td>\n",
       "      <td>how to build a fence</td>\n",
       "      <td>54.1619</td>\n",
       "      <td>38.2836</td>\n",
       "      <td>45.7611</td>\n",
       "      <td>45.3592</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41</th>\n",
       "      <td>92</td>\n",
       "      <td>the wall</td>\n",
       "      <td>51.8888</td>\n",
       "      <td>51.8888</td>\n",
       "      <td>64.8467</td>\n",
       "      <td>64.8467</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>184</td>\n",
       "      <td>civil right movement</td>\n",
       "      <td>50.0213</td>\n",
       "      <td>73.1061</td>\n",
       "      <td>40.7264</td>\n",
       "      <td>53.3459</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>176</td>\n",
       "      <td>weather strip</td>\n",
       "      <td>48.1532</td>\n",
       "      <td>48.1532</td>\n",
       "      <td>18.7130</td>\n",
       "      <td>18.7130</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>112</td>\n",
       "      <td>kenmore gas water heater</td>\n",
       "      <td>46.8892</td>\n",
       "      <td>54.1085</td>\n",
       "      <td>54.4762</td>\n",
       "      <td>61.0496</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    qid                                    query  mean_health_score  \\\n",
       "45  196                              sore throat           411.7695   \n",
       "37   88                             forearm pain           354.0813   \n",
       "31   82                                   joints           331.5549   \n",
       "27  128                  atypical squamous cells           272.0022   \n",
       "16   67                              vldl levels           267.6323   \n",
       "22  173                            hip fractures           233.0622   \n",
       "3     4                                   toilet           226.5945   \n",
       "27   78                                  dieting           220.2939   \n",
       "38   89                                      ocd           218.6667   \n",
       "29   30                       diabetes education           203.9913   \n",
       "12   63                                 flushing           201.4836   \n",
       "22  123                   von willebrand disease           200.9725   \n",
       "17  168                                   lipoma           181.9301   \n",
       "31  182                             quit smoking           177.5335   \n",
       "25   26                         lower heart rate           168.4160   \n",
       "6    57                                  ct jobs           167.2068   \n",
       "24  125                     butter and margarine           136.5154   \n",
       "44  195                         pressure washers           129.7544   \n",
       "10   61                                     worm           129.3127   \n",
       "10  111                         lymphoma in dogs           123.1295   \n",
       "20  121                       sit and reach test           121.9814   \n",
       "18   69                      sewing instructions           114.4124   \n",
       "33   34                              cell phones           114.1950   \n",
       "11  162                                      dnr           112.1133   \n",
       "17  118                  poem in your pocket day           107.4421   \n",
       "18  119                      interview thank you            91.8932   \n",
       "4    55                                     iron            88.1704   \n",
       "7   158                     septic system design            87.3629   \n",
       "26  177            best long term care insurance            86.1771   \n",
       "3   104                    indiana child support            84.3150   \n",
       "42  193                        dog clean up bags            83.9026   \n",
       "24  175                   signs of a heartattack            78.6511   \n",
       "1   152                        angular cheilitis            76.9094   \n",
       "2     3                        getting organized            74.7692   \n",
       "19   70  to be or not to be that is the question            68.8508   \n",
       "27  178                          pork tenderloin            66.2645   \n",
       "32   83                                   memory            64.2118   \n",
       "16  117                      dangers of asbestos            63.6375   \n",
       "31  132                        mothers day songs            60.7894   \n",
       "42  143                  arkadelphia health club            60.6634   \n",
       "8     9                           used car parts            57.5809   \n",
       "6     7                   air travel information            56.5040   \n",
       "10  161               furniture for small spaces            55.1753   \n",
       "38  189                              gs pay rate            54.8410   \n",
       "14  115             pacific northwest laboratory            54.6976   \n",
       "8    59                     how to build a fence            54.1619   \n",
       "41   92                                 the wall            51.8888   \n",
       "33  184                     civil right movement            50.0213   \n",
       "25  176                            weather strip            48.1532   \n",
       "11  112                 kenmore gas water heater            46.8892   \n",
       "\n",
       "    median_health_score  mean_medical_score  median_medical_score  \n",
       "45             411.7695             99.3674               99.3674  \n",
       "37             354.0813            310.8832              310.8832  \n",
       "31             331.5549            240.9360              240.9360  \n",
       "27             277.4749            452.8779              399.4028  \n",
       "16             267.6323            434.0579              434.0579  \n",
       "22             233.0622            308.1734              308.1734  \n",
       "3              226.5945             24.5363               24.5363  \n",
       "27             220.2939            245.9618              245.9618  \n",
       "38             218.6667            328.5885              328.5885  \n",
       "29             203.9913            243.6227              243.6227  \n",
       "12             201.4836             94.9256               94.9256  \n",
       "22             226.8965            279.1960              335.1087  \n",
       "17             181.9301            496.2743              496.2743  \n",
       "31             177.5335            190.5365              190.5365  \n",
       "25             130.3137            198.0859              184.8707  \n",
       "6              167.2068            179.6955              179.6955  \n",
       "24             125.0510             52.6185               54.7226  \n",
       "44             129.7544            126.1504              126.1504  \n",
       "10             129.3127             99.3838               99.3838  \n",
       "10              39.4860            228.4353              198.1618  \n",
       "20              78.1871             80.8871               54.8896  \n",
       "18             114.4124             38.1157               38.1157  \n",
       "33             114.1950            194.8462              194.8462  \n",
       "11             112.1133            193.3705              193.3705  \n",
       "17              34.8868             40.2603               48.4656  \n",
       "18               1.8060             19.2758                6.6938  \n",
       "4               88.1704            106.5617              106.5617  \n",
       "7               74.0658            199.0964              117.9455  \n",
       "26              47.4360             83.9122               44.2507  \n",
       "3               57.3328             56.8610               75.9244  \n",
       "42              81.7193             43.2959               40.5533  \n",
       "24              33.5019             66.7424               49.2636  \n",
       "1               76.9094            201.6837              201.6837  \n",
       "2               74.7692             22.8042               22.8042  \n",
       "19              59.1396             70.9016               80.7897  \n",
       "27              66.2645             41.0316               41.0316  \n",
       "32              64.2118            145.4128              145.4128  \n",
       "16              48.6819             89.2323               53.1681  \n",
       "31              57.7526             98.9898               52.5622  \n",
       "42               0.4916             62.4479                0.5497  \n",
       "8               75.1587             49.4786               32.9048  \n",
       "6               44.4720             55.5602               23.8385  \n",
       "10              39.9681             52.5583               59.6951  \n",
       "38              23.4032            123.0190              101.9324  \n",
       "14               2.7565             70.4378               11.0099  \n",
       "8               38.2836             45.7611               45.3592  \n",
       "41              51.8888             64.8467               64.8467  \n",
       "33              73.1061             40.7264               53.3459  \n",
       "25              48.1532             18.7130               18.7130  \n",
       "11              54.1085             54.4762               61.0496  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# all topics with mean health score > 150 are health-related\n",
    "tira.pt.transform_queries('ir-benchmarks/fschlatt/query-health-classification', all_datasets[0])\n",
    "pd.concat(\n",
    "    [query_health_classification(topics) for query_health_classification, topics in zip(all_query_health_classifications, all_topics)]\n",
    ").sort_values(\"mean_health_score\", ascending=False).head(50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "approaches = {\n",
    "    \"BM25\": \"ir-benchmarks/tira-ir-starter/BM25 Re-Rank (tira-ir-starter-pyterrier)\",\n",
    "    \"Dirichlet\": \"ir-benchmarks/tira-ir-starter/DirichletLM Re-Rank (tira-ir-starter-pyterrier)\",\n",
    "    \"MonoT5 Base\": \"ir-benchmarks/tira-ir-starter/MonoT5 Base (tira-ir-starter-gygaggle)\",\n",
    "    \"Splade\": \"workshop-on-open-web-search/naverlabseurope/Splade (re-ranker)\",\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are multiple qrel fields available: ['relevance', 'method', 'iprob']. Defaulting to \"relevance\", but to use a different one, supply variant\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 842kiB [00:00, 8.80MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb09-en-trec-web-2009-20230107-training/tira-ir-starter\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 57.4MiB [00:00, 94.9MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/workshop-on-open-web-search/clueweb09-en-trec-web-2009-20230107-training/naverlabseurope\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[INFO] [starting] https://trec.nist.gov/data/web/10/10.adhoc-qrels.final\n",
      "[INFO] [finished] https://trec.nist.gov/data/web/10/10.adhoc-qrels.final: [00:00] [837kB] [2.05MB/s]\n",
      "Download: 842kiB [00:00, 6.95MiB/s]                                               \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb09-en-trec-web-2010-20230107-training/tira-ir-starter\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 844kiB [00:00, 7.41MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb09-en-trec-web-2010-20230107-training/tira-ir-starter\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 855kiB [00:00, 7.66MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb09-en-trec-web-2010-20230107-training/tira-ir-starter\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 59.8MiB [00:00, 109MiB/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/workshop-on-open-web-search/clueweb09-en-trec-web-2010-20230107-training/naverlabseurope\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/pyterrier/pipelines.py:129: UserWarning: 2 topic(s) not found in qrels. Scores for these topics are given as NaN and should not contribute to averages.\n",
      "  warn(f'{backfill_count} topic(s) not found in qrels. Scores for these topics are given as NaN and should not contribute to averages.')\n",
      "/usr/local/lib/python3.10/dist-packages/pyterrier/pipelines.py:129: UserWarning: 2 topic(s) not found in qrels. Scores for these topics are given as NaN and should not contribute to averages.\n",
      "  warn(f'{backfill_count} topic(s) not found in qrels. Scores for these topics are given as NaN and should not contribute to averages.')\n",
      "/usr/local/lib/python3.10/dist-packages/pyterrier/pipelines.py:129: UserWarning: 2 topic(s) not found in qrels. Scores for these topics are given as NaN and should not contribute to averages.\n",
      "  warn(f'{backfill_count} topic(s) not found in qrels. Scores for these topics are given as NaN and should not contribute to averages.')\n",
      "/usr/local/lib/python3.10/dist-packages/pyterrier/pipelines.py:129: UserWarning: 2 topic(s) not found in qrels. Scores for these topics are given as NaN and should not contribute to averages.\n",
      "  warn(f'{backfill_count} topic(s) not found in qrels. Scores for these topics are given as NaN and should not contribute to averages.')\n",
      "[INFO] [starting] https://trec.nist.gov/data/web/11/qrels.adhoc\n",
      "[INFO] [finished] https://trec.nist.gov/data/web/11/qrels.adhoc: [00:01] [660kB] [627kB/s]\n",
      "Download: 807kiB [00:00, 6.74MiB/s]                                     \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb09-en-trec-web-2011-20230107-training/tira-ir-starter\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 811kiB [00:00, 8.56MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb09-en-trec-web-2011-20230107-training/tira-ir-starter\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 777kiB [00:00, 7.61MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb09-en-trec-web-2011-20230107-training/tira-ir-starter\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 53.6MiB [00:00, 103MiB/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/workshop-on-open-web-search/clueweb09-en-trec-web-2011-20230107-training/naverlabseurope\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[INFO] [starting] https://trec.nist.gov/data/web/12/qrels.adhoc\n",
      "[INFO] [finished] https://trec.nist.gov/data/web/12/qrels.adhoc: [00:00] [611kB] [628kB/s]\n",
      "Download: 835kiB [00:00, 8.01MiB/s]                                     \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb09-en-trec-web-2012-20230107-training/tira-ir-starter\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 835kiB [00:00, 6.20MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb09-en-trec-web-2012-20230107-training/tira-ir-starter\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 842kiB [00:00, 8.24MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb09-en-trec-web-2012-20230107-training/tira-ir-starter\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 56.5MiB [00:00, 100MiB/s] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/workshop-on-open-web-search/clueweb09-en-trec-web-2012-20230107-training/naverlabseurope\n"
     ]
    }
   ],
   "source": [
    "non_health_results = []\n",
    "health_results = []\n",
    "iterator = zip(all_topics, all_datasets, all_query_health_classifications)\n",
    "for topics, dataset, query_health_classification in iterator:\n",
    "    topics = query_health_classification(topics)\n",
    "    qrels = dataset.get_qrels(variant=\"relevance\")\n",
    "    non_health_topics = topics.loc[topics[\"mean_health_score\"] <= 150]\n",
    "    health_topics = topics.loc[topics[\"mean_health_score\"] >= 150]\n",
    "    rankings = [\n",
    "        tira.pt.from_submission(approach, dataset) for approach in approaches.values()\n",
    "    ]\n",
    "    non_health_results.append(\n",
    "        pt.Experiment(\n",
    "            rankings,\n",
    "            non_health_topics,\n",
    "            qrels,\n",
    "            ['ndcg_cut.10', 'recip_rank'],\n",
    "            names=list(approaches.keys()),\n",
    "            perquery=True\n",
    "        )\n",
    "    )\n",
    "    health_results.append(\n",
    "        pt.Experiment(\n",
    "            rankings,\n",
    "            health_topics,\n",
    "            qrels,\n",
    "            ['ndcg_cut.10', 'recip_rank'],\n",
    "            names=list(approaches.keys()),\n",
    "            perquery=True\n",
    "        )\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(184, 16)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat(non_health_results)[\"qid\"].unique().shape[0], pd.concat(health_results)[\"qid\"].unique().shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>measure</th>\n",
       "      <th>ndcg_cut.10</th>\n",
       "      <th>recip_rank</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>name</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Dirichlet</th>\n",
       "      <td>0.249843</td>\n",
       "      <td>0.485740</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Splade</th>\n",
       "      <td>0.192459</td>\n",
       "      <td>0.418589</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BM25</th>\n",
       "      <td>0.191653</td>\n",
       "      <td>0.438125</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>MonoT5 Base</th>\n",
       "      <td>0.184061</td>\n",
       "      <td>0.384390</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "measure      ndcg_cut.10  recip_rank\n",
       "name                                \n",
       "Dirichlet       0.249843    0.485740\n",
       "Splade          0.192459    0.418589\n",
       "BM25            0.191653    0.438125\n",
       "MonoT5 Base     0.184061    0.384390"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat(non_health_results).groupby([\"name\", \"measure\"])[\"value\"].mean().unstack(\"measure\").sort_values(\"ndcg_cut.10\", ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>measure</th>\n",
       "      <th>ndcg_cut.10</th>\n",
       "      <th>recip_rank</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>name</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Dirichlet</th>\n",
       "      <td>0.275507</td>\n",
       "      <td>0.542262</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BM25</th>\n",
       "      <td>0.240164</td>\n",
       "      <td>0.523327</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>MonoT5 Base</th>\n",
       "      <td>0.197945</td>\n",
       "      <td>0.452517</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Splade</th>\n",
       "      <td>0.166980</td>\n",
       "      <td>0.457813</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "measure      ndcg_cut.10  recip_rank\n",
       "name                                \n",
       "Dirichlet       0.275507    0.542262\n",
       "BM25            0.240164    0.523327\n",
       "MonoT5 Base     0.197945    0.452517\n",
       "Splade          0.166980    0.457813"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat(health_results).groupby([\"name\", \"measure\"])[\"value\"].mean().unstack(\"measure\").sort_values(\"ndcg_cut.10\", ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# TREC Health Misinformation Results\n",
    "\n",
    "Filtering documents by median health relatedness score can improve the effectiveness of BM25 on the TREC Health Misinformation task."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[INFO] [starting] https://trec.nist.gov/data/misinfo/2019topics.xml\n",
      "[INFO] [finished] https://trec.nist.gov/data/misinfo/2019topics.xml: [00:00] [30.0kB] [228kB/s]\n",
      "[INFO] [starting] https://trec.nist.gov/data/misinfo/2019qrels_raw.txt       \n",
      "[INFO] [finished] https://trec.nist.gov/data/misinfo/2019qrels_raw.txt: [00:00] [879kB] [1.33MB/s]\n",
      "                                                                                \r"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are multiple qrel fields available: ['relevance', 'effectiveness', 'redibility']. Defaulting to \"relevance\", but to use a different one, supply variant\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 1.14MiB [00:00, 10.1MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb12-trec-misinfo-2019-20240214-training/fschlatt\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 0.99MiB [00:00, 8.33MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb12-trec-misinfo-2019-20240214-training/tira-ir-starter\n"
     ]
    }
   ],
   "source": [
    "dataset = \"clueweb12-trec-misinfo-2019-20240214-training\"\n",
    "topics = pt.get_dataset(\"irds:clueweb12/b13/trec-misinfo-2019\").get_topics(variant=\"title\")\n",
    "qrels = pt.get_dataset(\"irds:clueweb12/b13/trec-misinfo-2019\").get_qrels(variant=\"relevance\")\n",
    "document_health_classification = tira.pt.transform_documents(\"ir-benchmarks/fschlatt/document-health-classification\", dataset)\n",
    "bm25 = tira.pt.from_submission(\"ir-benchmarks/tira-ir-starter/BM25 Re-Rank (tira-ir-starter-pyterrier)\", dataset)\n",
    "bm25_health = bm25 % 1000 >> document_health_classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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>qid</th>\n",
       "      <th>query</th>\n",
       "      <th>q0</th>\n",
       "      <th>rank</th>\n",
       "      <th>score</th>\n",
       "      <th>system</th>\n",
       "      <th>docno</th>\n",
       "      <th>tira_task</th>\n",
       "      <th>tira_dataset</th>\n",
       "      <th>tira_first_stage_run_id</th>\n",
       "      <th>mean_health_score</th>\n",
       "      <th>median_health_score</th>\n",
       "      <th>mean_medical_score</th>\n",
       "      <th>median_medical_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>Q0</td>\n",
       "      <td>1</td>\n",
       "      <td>23.569565</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1712wb-84-02961</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-18-10-10-30</td>\n",
       "      <td>108.1876</td>\n",
       "      <td>67.2081</td>\n",
       "      <td>114.9722</td>\n",
       "      <td>74.7917</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>Q0</td>\n",
       "      <td>2</td>\n",
       "      <td>23.476046</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1304wb-51-24177</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-18-10-10-30</td>\n",
       "      <td>100.8895</td>\n",
       "      <td>57.2659</td>\n",
       "      <td>111.6515</td>\n",
       "      <td>76.4557</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>Q0</td>\n",
       "      <td>3</td>\n",
       "      <td>23.404021</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1304wb-88-12518</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-18-10-10-30</td>\n",
       "      <td>108.0479</td>\n",
       "      <td>66.5542</td>\n",
       "      <td>113.0515</td>\n",
       "      <td>74.5736</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>Q0</td>\n",
       "      <td>4</td>\n",
       "      <td>23.380401</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-0109wb-83-09551</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-18-10-10-30</td>\n",
       "      <td>107.0973</td>\n",
       "      <td>55.9035</td>\n",
       "      <td>124.5779</td>\n",
       "      <td>84.6786</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>Q0</td>\n",
       "      <td>5</td>\n",
       "      <td>23.370749</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1912wb-62-07565</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-18-10-10-30</td>\n",
       "      <td>92.6918</td>\n",
       "      <td>46.1817</td>\n",
       "      <td>100.2450</td>\n",
       "      <td>60.0157</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50994</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>Q0</td>\n",
       "      <td>995</td>\n",
       "      <td>10.490712</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-0500wb-06-07863</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-18-10-10-30</td>\n",
       "      <td>68.3214</td>\n",
       "      <td>35.4775</td>\n",
       "      <td>93.5654</td>\n",
       "      <td>59.5807</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50995</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>Q0</td>\n",
       "      <td>996</td>\n",
       "      <td>10.487029</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1307wb-71-08120</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-18-10-10-30</td>\n",
       "      <td>79.5897</td>\n",
       "      <td>42.9657</td>\n",
       "      <td>102.9494</td>\n",
       "      <td>61.0742</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50996</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>Q0</td>\n",
       "      <td>997</td>\n",
       "      <td>10.462628</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-0100wb-15-07532</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-18-10-10-30</td>\n",
       "      <td>101.0621</td>\n",
       "      <td>69.3955</td>\n",
       "      <td>127.3782</td>\n",
       "      <td>91.2131</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50997</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>Q0</td>\n",
       "      <td>998</td>\n",
       "      <td>10.440274</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-0600tw-25-01456</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-18-10-10-30</td>\n",
       "      <td>99.0062</td>\n",
       "      <td>72.4115</td>\n",
       "      <td>93.6922</td>\n",
       "      <td>60.6180</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50998</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>Q0</td>\n",
       "      <td>999</td>\n",
       "      <td>9.650459</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1611wb-80-00880</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-18-10-10-30</td>\n",
       "      <td>61.0268</td>\n",
       "      <td>38.0222</td>\n",
       "      <td>81.9152</td>\n",
       "      <td>64.4986</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>50949 rows × 14 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      qid                                 query  q0  rank      score  \\\n",
       "0       1  cranberries urinary tract infections  Q0     1  23.569565   \n",
       "1       1  cranberries urinary tract infections  Q0     2  23.476046   \n",
       "2       1  cranberries urinary tract infections  Q0     3  23.404021   \n",
       "3       1  cranberries urinary tract infections  Q0     4  23.380401   \n",
       "4       1  cranberries urinary tract infections  Q0     5  23.370749   \n",
       "...    ..                                   ...  ..   ...        ...   \n",
       "50994  51                  dehumidifiers asthma  Q0   995  10.490712   \n",
       "50995  51                  dehumidifiers asthma  Q0   996  10.487029   \n",
       "50996  51                  dehumidifiers asthma  Q0   997  10.462628   \n",
       "50997  51                  dehumidifiers asthma  Q0   998  10.440274   \n",
       "50998  51                  dehumidifiers asthma  Q0   999   9.650459   \n",
       "\n",
       "                                               system  \\\n",
       "0      pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "1      pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "2      pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "3      pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "4      pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "...                                               ...   \n",
       "50994  pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "50995  pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "50996  pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "50997  pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "50998  pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "\n",
       "                           docno      tira_task  \\\n",
       "0      clueweb12-1712wb-84-02961  ir-benchmarks   \n",
       "1      clueweb12-1304wb-51-24177  ir-benchmarks   \n",
       "2      clueweb12-1304wb-88-12518  ir-benchmarks   \n",
       "3      clueweb12-0109wb-83-09551  ir-benchmarks   \n",
       "4      clueweb12-1912wb-62-07565  ir-benchmarks   \n",
       "...                          ...            ...   \n",
       "50994  clueweb12-0500wb-06-07863  ir-benchmarks   \n",
       "50995  clueweb12-1307wb-71-08120  ir-benchmarks   \n",
       "50996  clueweb12-0100wb-15-07532  ir-benchmarks   \n",
       "50997  clueweb12-0600tw-25-01456  ir-benchmarks   \n",
       "50998  clueweb12-1611wb-80-00880  ir-benchmarks   \n",
       "\n",
       "                                        tira_dataset tira_first_stage_run_id  \\\n",
       "0      clueweb12-trec-misinfo-2019-20240214-training     2024-03-18-10-10-30   \n",
       "1      clueweb12-trec-misinfo-2019-20240214-training     2024-03-18-10-10-30   \n",
       "2      clueweb12-trec-misinfo-2019-20240214-training     2024-03-18-10-10-30   \n",
       "3      clueweb12-trec-misinfo-2019-20240214-training     2024-03-18-10-10-30   \n",
       "4      clueweb12-trec-misinfo-2019-20240214-training     2024-03-18-10-10-30   \n",
       "...                                              ...                     ...   \n",
       "50994  clueweb12-trec-misinfo-2019-20240214-training     2024-03-18-10-10-30   \n",
       "50995  clueweb12-trec-misinfo-2019-20240214-training     2024-03-18-10-10-30   \n",
       "50996  clueweb12-trec-misinfo-2019-20240214-training     2024-03-18-10-10-30   \n",
       "50997  clueweb12-trec-misinfo-2019-20240214-training     2024-03-18-10-10-30   \n",
       "50998  clueweb12-trec-misinfo-2019-20240214-training     2024-03-18-10-10-30   \n",
       "\n",
       "       mean_health_score  median_health_score  mean_medical_score  \\\n",
       "0               108.1876              67.2081            114.9722   \n",
       "1               100.8895              57.2659            111.6515   \n",
       "2               108.0479              66.5542            113.0515   \n",
       "3               107.0973              55.9035            124.5779   \n",
       "4                92.6918              46.1817            100.2450   \n",
       "...                  ...                  ...                 ...   \n",
       "50994            68.3214              35.4775             93.5654   \n",
       "50995            79.5897              42.9657            102.9494   \n",
       "50996           101.0621              69.3955            127.3782   \n",
       "50997            99.0062              72.4115             93.6922   \n",
       "50998            61.0268              38.0222             81.9152   \n",
       "\n",
       "       median_medical_score  \n",
       "0                   74.7917  \n",
       "1                   76.4557  \n",
       "2                   74.5736  \n",
       "3                   84.6786  \n",
       "4                   60.0157  \n",
       "...                     ...  \n",
       "50994               59.5807  \n",
       "50995               61.0742  \n",
       "50996               91.2131  \n",
       "50997               60.6180  \n",
       "50998               64.4986  \n",
       "\n",
       "[50949 rows x 14 columns]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bm25_ranking = bm25_health(topics)\n",
    "bm25_ranking"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "rankings = []\n",
    "thresholds = list(range(0, 60, 5))\n",
    "for threshold in thresholds:\n",
    "    rankings.append(bm25_ranking.loc[bm25_ranking[\"median_health_score\"] >= threshold])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>nDCG@10</th>\n",
       "      <th>MRR</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Health Score Threshold</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.467323</td>\n",
       "      <td>0.720945</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.467232</td>\n",
       "      <td>0.723088</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.473522</td>\n",
       "      <td>0.728755</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.475634</td>\n",
       "      <td>0.738773</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.485593</td>\n",
       "      <td>0.743662</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.496041</td>\n",
       "      <td>0.757014</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.490500</td>\n",
       "      <td>0.770645</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>0.479722</td>\n",
       "      <td>0.767978</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.479449</td>\n",
       "      <td>0.765714</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>0.482691</td>\n",
       "      <td>0.816157</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50</th>\n",
       "      <td>0.455525</td>\n",
       "      <td>0.783242</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>55</th>\n",
       "      <td>0.407818</td>\n",
       "      <td>0.716444</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         nDCG@10       MRR\n",
       "Health Score Threshold                    \n",
       "0                       0.467323  0.720945\n",
       "5                       0.467232  0.723088\n",
       "10                      0.473522  0.728755\n",
       "15                      0.475634  0.738773\n",
       "20                      0.485593  0.743662\n",
       "25                      0.496041  0.757014\n",
       "30                      0.490500  0.770645\n",
       "35                      0.479722  0.767978\n",
       "40                      0.479449  0.765714\n",
       "45                      0.482691  0.816157\n",
       "50                      0.455525  0.783242\n",
       "55                      0.407818  0.716444"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results = pt.Experiment(\n",
    "    rankings,\n",
    "    topics,\n",
    "    qrels.rename({\"label\": \"relevance\"}, axis=1),\n",
    "    ['ndcg_cut.10', 'recip_rank'],\n",
    "    names=thresholds,\n",
    ").rename({\"name\": \"Health Score Threshold\", \"ndcg_cut.10\": \"nDCG@10\", \"recip_rank\": \"MRR\"}, axis=1)\n",
    "results = results.set_index(\"Health Score Threshold\")\n",
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 560x280 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "scale = 0.7\n",
    "fig, axes = plt.subplots(1, 2, figsize=(8 * scale, 4 * scale))\n",
    "for idx, metric in enumerate((\"nDCG@10\", \"MRR\")):\n",
    "    ax = axes[idx]\n",
    "    ax.plot(results.index, results.loc[:, metric], marker=\"o\")\n",
    "    ax.set_ylabel(metric)\n",
    "    ax.set_xlabel(\"Health Score Threshold\")\n",
    "fig.suptitle(\"TREC Health Misinformation\")\n",
    "fig.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# DirichletLM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are multiple qrel fields available: ['relevance', 'effectiveness', 'redibility']. Defaulting to \"relevance\", but to use a different one, supply variant\n"
     ]
    }
   ],
   "source": [
    "dataset = \"clueweb12-trec-misinfo-2019-20240214-training\"\n",
    "topics = pt.get_dataset(\"irds:clueweb12/b13/trec-misinfo-2019\").get_topics(variant=\"title\")\n",
    "qrels = pt.get_dataset(\"irds:clueweb12/b13/trec-misinfo-2019\").get_qrels(variant=\"relevance\")\n",
    "document_health_classification = tira.pt.transform_documents(\"ir-benchmarks/fschlatt/document-health-classification\", dataset)\n",
    "bm25 = tira.pt.from_submission(\"ir-benchmarks/tira-ir-starter/DirichletLM Re-Rank (tira-ir-starter-pyterrier)\", dataset)\n",
    "bm25_health = bm25 % 1000 >> document_health_classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "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>qid</th>\n",
       "      <th>query</th>\n",
       "      <th>q0</th>\n",
       "      <th>rank</th>\n",
       "      <th>score</th>\n",
       "      <th>system</th>\n",
       "      <th>docno</th>\n",
       "      <th>tira_task</th>\n",
       "      <th>tira_dataset</th>\n",
       "      <th>tira_first_stage_run_id</th>\n",
       "      <th>mean_health_score</th>\n",
       "      <th>median_health_score</th>\n",
       "      <th>mean_medical_score</th>\n",
       "      <th>median_medical_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>Q0</td>\n",
       "      <td>1</td>\n",
       "      <td>16.953256</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1514wb-75-03723</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-56-54</td>\n",
       "      <td>60.0634</td>\n",
       "      <td>0.1428</td>\n",
       "      <td>85.1522</td>\n",
       "      <td>8.0731</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>Q0</td>\n",
       "      <td>2</td>\n",
       "      <td>16.794832</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1712wb-84-02961</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-56-54</td>\n",
       "      <td>108.1876</td>\n",
       "      <td>67.2081</td>\n",
       "      <td>114.9722</td>\n",
       "      <td>74.7917</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>Q0</td>\n",
       "      <td>3</td>\n",
       "      <td>16.564677</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1514wb-50-05021</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-56-54</td>\n",
       "      <td>97.8244</td>\n",
       "      <td>52.5293</td>\n",
       "      <td>106.5242</td>\n",
       "      <td>71.3221</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>Q0</td>\n",
       "      <td>4</td>\n",
       "      <td>16.520074</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1515wb-33-25929</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-56-54</td>\n",
       "      <td>101.3218</td>\n",
       "      <td>57.5093</td>\n",
       "      <td>119.0994</td>\n",
       "      <td>79.4458</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>Q0</td>\n",
       "      <td>5</td>\n",
       "      <td>16.486682</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1304wb-88-12518</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-56-54</td>\n",
       "      <td>108.0479</td>\n",
       "      <td>66.5542</td>\n",
       "      <td>113.0515</td>\n",
       "      <td>74.5736</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50994</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>Q0</td>\n",
       "      <td>995</td>\n",
       "      <td>1.385201</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1007wb-24-17071</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-56-54</td>\n",
       "      <td>52.0621</td>\n",
       "      <td>19.0757</td>\n",
       "      <td>64.5174</td>\n",
       "      <td>26.3843</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50995</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>Q0</td>\n",
       "      <td>996</td>\n",
       "      <td>1.316988</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1411wb-44-17605</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-56-54</td>\n",
       "      <td>62.2313</td>\n",
       "      <td>36.9000</td>\n",
       "      <td>66.3142</td>\n",
       "      <td>49.2285</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50996</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>Q0</td>\n",
       "      <td>997</td>\n",
       "      <td>1.307131</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-0507wb-68-20656</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-56-54</td>\n",
       "      <td>62.0919</td>\n",
       "      <td>34.9866</td>\n",
       "      <td>86.9873</td>\n",
       "      <td>54.7226</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50997</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>Q0</td>\n",
       "      <td>998</td>\n",
       "      <td>1.259419</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-1614wb-02-17964</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-56-54</td>\n",
       "      <td>93.1677</td>\n",
       "      <td>48.3995</td>\n",
       "      <td>94.6239</td>\n",
       "      <td>65.8722</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50998</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>Q0</td>\n",
       "      <td>999</td>\n",
       "      <td>1.233724</td>\n",
       "      <td>pyterrier.default_pipelines.wmodel_text_scorer</td>\n",
       "      <td>clueweb12-0200wb-21-19579</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-56-54</td>\n",
       "      <td>69.4880</td>\n",
       "      <td>35.2037</td>\n",
       "      <td>80.4149</td>\n",
       "      <td>44.5226</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>50949 rows × 14 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      qid                                 query  q0  rank      score  \\\n",
       "0       1  cranberries urinary tract infections  Q0     1  16.953256   \n",
       "1       1  cranberries urinary tract infections  Q0     2  16.794832   \n",
       "2       1  cranberries urinary tract infections  Q0     3  16.564677   \n",
       "3       1  cranberries urinary tract infections  Q0     4  16.520074   \n",
       "4       1  cranberries urinary tract infections  Q0     5  16.486682   \n",
       "...    ..                                   ...  ..   ...        ...   \n",
       "50994  51                  dehumidifiers asthma  Q0   995   1.385201   \n",
       "50995  51                  dehumidifiers asthma  Q0   996   1.316988   \n",
       "50996  51                  dehumidifiers asthma  Q0   997   1.307131   \n",
       "50997  51                  dehumidifiers asthma  Q0   998   1.259419   \n",
       "50998  51                  dehumidifiers asthma  Q0   999   1.233724   \n",
       "\n",
       "                                               system  \\\n",
       "0      pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "1      pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "2      pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "3      pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "4      pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "...                                               ...   \n",
       "50994  pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "50995  pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "50996  pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "50997  pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "50998  pyterrier.default_pipelines.wmodel_text_scorer   \n",
       "\n",
       "                           docno      tira_task  \\\n",
       "0      clueweb12-1514wb-75-03723  ir-benchmarks   \n",
       "1      clueweb12-1712wb-84-02961  ir-benchmarks   \n",
       "2      clueweb12-1514wb-50-05021  ir-benchmarks   \n",
       "3      clueweb12-1515wb-33-25929  ir-benchmarks   \n",
       "4      clueweb12-1304wb-88-12518  ir-benchmarks   \n",
       "...                          ...            ...   \n",
       "50994  clueweb12-1007wb-24-17071  ir-benchmarks   \n",
       "50995  clueweb12-1411wb-44-17605  ir-benchmarks   \n",
       "50996  clueweb12-0507wb-68-20656  ir-benchmarks   \n",
       "50997  clueweb12-1614wb-02-17964  ir-benchmarks   \n",
       "50998  clueweb12-0200wb-21-19579  ir-benchmarks   \n",
       "\n",
       "                                        tira_dataset tira_first_stage_run_id  \\\n",
       "0      clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-14-56-54   \n",
       "1      clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-14-56-54   \n",
       "2      clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-14-56-54   \n",
       "3      clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-14-56-54   \n",
       "4      clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-14-56-54   \n",
       "...                                              ...                     ...   \n",
       "50994  clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-14-56-54   \n",
       "50995  clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-14-56-54   \n",
       "50996  clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-14-56-54   \n",
       "50997  clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-14-56-54   \n",
       "50998  clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-14-56-54   \n",
       "\n",
       "       mean_health_score  median_health_score  mean_medical_score  \\\n",
       "0                60.0634               0.1428             85.1522   \n",
       "1               108.1876              67.2081            114.9722   \n",
       "2                97.8244              52.5293            106.5242   \n",
       "3               101.3218              57.5093            119.0994   \n",
       "4               108.0479              66.5542            113.0515   \n",
       "...                  ...                  ...                 ...   \n",
       "50994            52.0621              19.0757             64.5174   \n",
       "50995            62.2313              36.9000             66.3142   \n",
       "50996            62.0919              34.9866             86.9873   \n",
       "50997            93.1677              48.3995             94.6239   \n",
       "50998            69.4880              35.2037             80.4149   \n",
       "\n",
       "       median_medical_score  \n",
       "0                    8.0731  \n",
       "1                   74.7917  \n",
       "2                   71.3221  \n",
       "3                   79.4458  \n",
       "4                   74.5736  \n",
       "...                     ...  \n",
       "50994               26.3843  \n",
       "50995               49.2285  \n",
       "50996               54.7226  \n",
       "50997               65.8722  \n",
       "50998               44.5226  \n",
       "\n",
       "[50949 rows x 14 columns]"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bm25_ranking = bm25_health(topics)\n",
    "bm25_ranking"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "rankings = []\n",
    "thresholds = list(range(0, 60, 5))\n",
    "for threshold in thresholds:\n",
    "    rankings.append(bm25_ranking.loc[bm25_ranking[\"median_health_score\"] >= threshold])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "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>nDCG@10</th>\n",
       "      <th>MRR</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Health Score Threshold</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.395890</td>\n",
       "      <td>0.639551</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.392827</td>\n",
       "      <td>0.650913</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.396698</td>\n",
       "      <td>0.659913</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.400936</td>\n",
       "      <td>0.656579</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.408964</td>\n",
       "      <td>0.660270</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.420976</td>\n",
       "      <td>0.667413</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.413129</td>\n",
       "      <td>0.674285</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>0.395621</td>\n",
       "      <td>0.684285</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.396495</td>\n",
       "      <td>0.705022</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>0.400268</td>\n",
       "      <td>0.728804</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50</th>\n",
       "      <td>0.386856</td>\n",
       "      <td>0.722764</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>55</th>\n",
       "      <td>0.355001</td>\n",
       "      <td>0.639244</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         nDCG@10       MRR\n",
       "Health Score Threshold                    \n",
       "0                       0.395890  0.639551\n",
       "5                       0.392827  0.650913\n",
       "10                      0.396698  0.659913\n",
       "15                      0.400936  0.656579\n",
       "20                      0.408964  0.660270\n",
       "25                      0.420976  0.667413\n",
       "30                      0.413129  0.674285\n",
       "35                      0.395621  0.684285\n",
       "40                      0.396495  0.705022\n",
       "45                      0.400268  0.728804\n",
       "50                      0.386856  0.722764\n",
       "55                      0.355001  0.639244"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results = pt.Experiment(\n",
    "    rankings,\n",
    "    topics,\n",
    "    qrels.rename({\"label\": \"relevance\"}, axis=1),\n",
    "    ['ndcg_cut.10', 'recip_rank'],\n",
    "    names=thresholds,\n",
    ").rename({\"name\": \"Health Score Threshold\", \"ndcg_cut.10\": \"nDCG@10\", \"recip_rank\": \"MRR\"}, axis=1)\n",
    "results = results.set_index(\"Health Score Threshold\")\n",
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 560x280 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "scale = 0.7\n",
    "fig, axes = plt.subplots(1, 2, figsize=(8 * scale, 4 * scale))\n",
    "for idx, metric in enumerate((\"nDCG@10\", \"MRR\")):\n",
    "    ax = axes[idx]\n",
    "    ax.plot(results.index, results.loc[:, metric], marker=\"o\")\n",
    "    ax.set_ylabel(metric)\n",
    "    ax.set_xlabel(\"Health Score Threshold\")\n",
    "fig.suptitle(\"TREC Health Misinformation\")\n",
    "fig.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# MonoT5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are multiple qrel fields available: ['relevance', 'effectiveness', 'redibility']. Defaulting to \"relevance\", but to use a different one, supply variant\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Download: 964kiB [00:00, 8.07MiB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Download finished. Extract...\n",
      "Extraction finished:  /root/.tira/extracted_runs/ir-benchmarks/clueweb12-trec-misinfo-2019-20240214-training/tira-ir-starter\n"
     ]
    }
   ],
   "source": [
    "dataset = \"clueweb12-trec-misinfo-2019-20240214-training\"\n",
    "topics = pt.get_dataset(\"irds:clueweb12/b13/trec-misinfo-2019\").get_topics(variant=\"title\")\n",
    "qrels = pt.get_dataset(\"irds:clueweb12/b13/trec-misinfo-2019\").get_qrels(variant=\"relevance\")\n",
    "document_health_classification = tira.pt.transform_documents(\"ir-benchmarks/fschlatt/document-health-classification\", dataset)\n",
    "bm25 = tira.pt.from_submission(\"ir-benchmarks/tira-ir-starter/MonoT5 Base (tira-ir-starter-gygaggle)\", dataset)\n",
    "bm25_health = bm25 % 1000 >> document_health_classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "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>qid</th>\n",
       "      <th>query</th>\n",
       "      <th>q0</th>\n",
       "      <th>rank</th>\n",
       "      <th>score</th>\n",
       "      <th>system</th>\n",
       "      <th>docno</th>\n",
       "      <th>tira_task</th>\n",
       "      <th>tira_dataset</th>\n",
       "      <th>tira_first_stage_run_id</th>\n",
       "      <th>mean_health_score</th>\n",
       "      <th>median_health_score</th>\n",
       "      <th>mean_medical_score</th>\n",
       "      <th>median_medical_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-0.011960</td>\n",
       "      <td>castorini/monot5-base-msmarco-10k</td>\n",
       "      <td>clueweb12-0612wb-48-06128</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-55-57</td>\n",
       "      <td>52.5139</td>\n",
       "      <td>12.2976</td>\n",
       "      <td>84.1604</td>\n",
       "      <td>44.9997</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>-0.012179</td>\n",
       "      <td>castorini/monot5-base-msmarco-10k</td>\n",
       "      <td>clueweb12-1100wb-42-24270</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-55-57</td>\n",
       "      <td>73.6753</td>\n",
       "      <td>28.6258</td>\n",
       "      <td>123.6957</td>\n",
       "      <td>60.0157</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>-0.012684</td>\n",
       "      <td>castorini/monot5-base-msmarco-10k</td>\n",
       "      <td>clueweb12-1309wb-66-04887</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-55-57</td>\n",
       "      <td>85.1328</td>\n",
       "      <td>43.6221</td>\n",
       "      <td>133.2357</td>\n",
       "      <td>87.1570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>-0.012884</td>\n",
       "      <td>castorini/monot5-base-msmarco-10k</td>\n",
       "      <td>clueweb12-0908wb-85-24723</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-55-57</td>\n",
       "      <td>65.9521</td>\n",
       "      <td>30.0989</td>\n",
       "      <td>86.9107</td>\n",
       "      <td>52.9474</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>-0.014070</td>\n",
       "      <td>castorini/monot5-base-msmarco-10k</td>\n",
       "      <td>clueweb12-1207wb-55-13945</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-55-57</td>\n",
       "      <td>78.8672</td>\n",
       "      <td>42.2202</td>\n",
       "      <td>148.7198</td>\n",
       "      <td>99.2418</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50994</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>0</td>\n",
       "      <td>995</td>\n",
       "      <td>-11.716431</td>\n",
       "      <td>castorini/monot5-base-msmarco-10k</td>\n",
       "      <td>clueweb12-1516wb-85-16736</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-55-57</td>\n",
       "      <td>61.8350</td>\n",
       "      <td>38.3390</td>\n",
       "      <td>78.2221</td>\n",
       "      <td>51.9701</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50995</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>0</td>\n",
       "      <td>996</td>\n",
       "      <td>-11.828877</td>\n",
       "      <td>castorini/monot5-base-msmarco-10k</td>\n",
       "      <td>clueweb12-0300wb-97-10173</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-55-57</td>\n",
       "      <td>48.7505</td>\n",
       "      <td>14.5529</td>\n",
       "      <td>80.8725</td>\n",
       "      <td>45.4343</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50996</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>0</td>\n",
       "      <td>997</td>\n",
       "      <td>-11.857443</td>\n",
       "      <td>castorini/monot5-base-msmarco-10k</td>\n",
       "      <td>clueweb12-1703wb-89-09044</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-55-57</td>\n",
       "      <td>54.0652</td>\n",
       "      <td>18.8561</td>\n",
       "      <td>63.2402</td>\n",
       "      <td>26.7763</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50997</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>0</td>\n",
       "      <td>998</td>\n",
       "      <td>-11.907468</td>\n",
       "      <td>castorini/monot5-base-msmarco-10k</td>\n",
       "      <td>clueweb12-0210wb-53-21439</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-55-57</td>\n",
       "      <td>54.2535</td>\n",
       "      <td>15.9042</td>\n",
       "      <td>73.5243</td>\n",
       "      <td>29.0075</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50998</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>0</td>\n",
       "      <td>999</td>\n",
       "      <td>-12.072540</td>\n",
       "      <td>castorini/monot5-base-msmarco-10k</td>\n",
       "      <td>clueweb12-1702wb-63-15696</td>\n",
       "      <td>ir-benchmarks</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-14-55-57</td>\n",
       "      <td>16.1865</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>27.6521</td>\n",
       "      <td>0.1161</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>50949 rows × 14 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      qid                                 query  q0  rank      score  \\\n",
       "0       1  cranberries urinary tract infections   0     1  -0.011960   \n",
       "1       1  cranberries urinary tract infections   0     2  -0.012179   \n",
       "2       1  cranberries urinary tract infections   0     3  -0.012684   \n",
       "3       1  cranberries urinary tract infections   0     4  -0.012884   \n",
       "4       1  cranberries urinary tract infections   0     5  -0.014070   \n",
       "...    ..                                   ...  ..   ...        ...   \n",
       "50994  51                  dehumidifiers asthma   0   995 -11.716431   \n",
       "50995  51                  dehumidifiers asthma   0   996 -11.828877   \n",
       "50996  51                  dehumidifiers asthma   0   997 -11.857443   \n",
       "50997  51                  dehumidifiers asthma   0   998 -11.907468   \n",
       "50998  51                  dehumidifiers asthma   0   999 -12.072540   \n",
       "\n",
       "                                  system                      docno  \\\n",
       "0      castorini/monot5-base-msmarco-10k  clueweb12-0612wb-48-06128   \n",
       "1      castorini/monot5-base-msmarco-10k  clueweb12-1100wb-42-24270   \n",
       "2      castorini/monot5-base-msmarco-10k  clueweb12-1309wb-66-04887   \n",
       "3      castorini/monot5-base-msmarco-10k  clueweb12-0908wb-85-24723   \n",
       "4      castorini/monot5-base-msmarco-10k  clueweb12-1207wb-55-13945   \n",
       "...                                  ...                        ...   \n",
       "50994  castorini/monot5-base-msmarco-10k  clueweb12-1516wb-85-16736   \n",
       "50995  castorini/monot5-base-msmarco-10k  clueweb12-0300wb-97-10173   \n",
       "50996  castorini/monot5-base-msmarco-10k  clueweb12-1703wb-89-09044   \n",
       "50997  castorini/monot5-base-msmarco-10k  clueweb12-0210wb-53-21439   \n",
       "50998  castorini/monot5-base-msmarco-10k  clueweb12-1702wb-63-15696   \n",
       "\n",
       "           tira_task                                   tira_dataset  \\\n",
       "0      ir-benchmarks  clueweb12-trec-misinfo-2019-20240214-training   \n",
       "1      ir-benchmarks  clueweb12-trec-misinfo-2019-20240214-training   \n",
       "2      ir-benchmarks  clueweb12-trec-misinfo-2019-20240214-training   \n",
       "3      ir-benchmarks  clueweb12-trec-misinfo-2019-20240214-training   \n",
       "4      ir-benchmarks  clueweb12-trec-misinfo-2019-20240214-training   \n",
       "...              ...                                            ...   \n",
       "50994  ir-benchmarks  clueweb12-trec-misinfo-2019-20240214-training   \n",
       "50995  ir-benchmarks  clueweb12-trec-misinfo-2019-20240214-training   \n",
       "50996  ir-benchmarks  clueweb12-trec-misinfo-2019-20240214-training   \n",
       "50997  ir-benchmarks  clueweb12-trec-misinfo-2019-20240214-training   \n",
       "50998  ir-benchmarks  clueweb12-trec-misinfo-2019-20240214-training   \n",
       "\n",
       "      tira_first_stage_run_id  mean_health_score  median_health_score  \\\n",
       "0         2024-03-19-14-55-57            52.5139              12.2976   \n",
       "1         2024-03-19-14-55-57            73.6753              28.6258   \n",
       "2         2024-03-19-14-55-57            85.1328              43.6221   \n",
       "3         2024-03-19-14-55-57            65.9521              30.0989   \n",
       "4         2024-03-19-14-55-57            78.8672              42.2202   \n",
       "...                       ...                ...                  ...   \n",
       "50994     2024-03-19-14-55-57            61.8350              38.3390   \n",
       "50995     2024-03-19-14-55-57            48.7505              14.5529   \n",
       "50996     2024-03-19-14-55-57            54.0652              18.8561   \n",
       "50997     2024-03-19-14-55-57            54.2535              15.9042   \n",
       "50998     2024-03-19-14-55-57            16.1865               0.0000   \n",
       "\n",
       "       mean_medical_score  median_medical_score  \n",
       "0                 84.1604               44.9997  \n",
       "1                123.6957               60.0157  \n",
       "2                133.2357               87.1570  \n",
       "3                 86.9107               52.9474  \n",
       "4                148.7198               99.2418  \n",
       "...                   ...                   ...  \n",
       "50994             78.2221               51.9701  \n",
       "50995             80.8725               45.4343  \n",
       "50996             63.2402               26.7763  \n",
       "50997             73.5243               29.0075  \n",
       "50998             27.6521                0.1161  \n",
       "\n",
       "[50949 rows x 14 columns]"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bm25_ranking = bm25_health(topics)\n",
    "bm25_ranking"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "rankings = []\n",
    "thresholds = list(range(0, 60, 5))\n",
    "for threshold in thresholds:\n",
    "    rankings.append(bm25_ranking.loc[bm25_ranking[\"median_health_score\"] >= threshold])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "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>nDCG@10</th>\n",
       "      <th>MRR</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Health Score Threshold</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.486835</td>\n",
       "      <td>0.782772</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.490378</td>\n",
       "      <td>0.784442</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.492162</td>\n",
       "      <td>0.798803</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.497177</td>\n",
       "      <td>0.802237</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.497976</td>\n",
       "      <td>0.802262</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.485814</td>\n",
       "      <td>0.796602</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.482651</td>\n",
       "      <td>0.810317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>0.470366</td>\n",
       "      <td>0.772951</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.444610</td>\n",
       "      <td>0.751048</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>0.443756</td>\n",
       "      <td>0.778813</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50</th>\n",
       "      <td>0.402525</td>\n",
       "      <td>0.789967</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>55</th>\n",
       "      <td>0.355586</td>\n",
       "      <td>0.727647</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         nDCG@10       MRR\n",
       "Health Score Threshold                    \n",
       "0                       0.486835  0.782772\n",
       "5                       0.490378  0.784442\n",
       "10                      0.492162  0.798803\n",
       "15                      0.497177  0.802237\n",
       "20                      0.497976  0.802262\n",
       "25                      0.485814  0.796602\n",
       "30                      0.482651  0.810317\n",
       "35                      0.470366  0.772951\n",
       "40                      0.444610  0.751048\n",
       "45                      0.443756  0.778813\n",
       "50                      0.402525  0.789967\n",
       "55                      0.355586  0.727647"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results = pt.Experiment(\n",
    "    rankings,\n",
    "    topics,\n",
    "    qrels.rename({\"label\": \"relevance\"}, axis=1),\n",
    "    ['ndcg_cut.10', 'recip_rank'],\n",
    "    names=thresholds,\n",
    ").rename({\"name\": \"Health Score Threshold\", \"ndcg_cut.10\": \"nDCG@10\", \"recip_rank\": \"MRR\"}, axis=1)\n",
    "results = results.set_index(\"Health Score Threshold\")\n",
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 560x280 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "scale = 0.7\n",
    "fig, axes = plt.subplots(1, 2, figsize=(8 * scale, 4 * scale))\n",
    "for idx, metric in enumerate((\"nDCG@10\", \"MRR\")):\n",
    "    ax = axes[idx]\n",
    "    ax.plot(results.index, results.loc[:, metric], marker=\"o\")\n",
    "    ax.set_ylabel(metric)\n",
    "    ax.set_xlabel(\"Health Score Threshold\")\n",
    "fig.suptitle(\"TREC Health Misinformation\")\n",
    "fig.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Splade"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are multiple qrel fields available: ['relevance', 'effectiveness', 'redibility']. Defaulting to \"relevance\", but to use a different one, supply variant\n"
     ]
    }
   ],
   "source": [
    "dataset = \"clueweb12-trec-misinfo-2019-20240214-training\"\n",
    "topics = pt.get_dataset(\"irds:clueweb12/b13/trec-misinfo-2019\").get_topics(variant=\"title\")\n",
    "qrels = pt.get_dataset(\"irds:clueweb12/b13/trec-misinfo-2019\").get_qrels(variant=\"relevance\")\n",
    "document_health_classification = tira.pt.transform_documents(\"ir-benchmarks/fschlatt/document-health-classification\", dataset)\n",
    "bm25 = tira.pt.from_submission(\"workshop-on-open-web-search/naverlabseurope/Splade (re-ranker)\", dataset)\n",
    "bm25_health = bm25 % 1000 >> document_health_classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "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>qid</th>\n",
       "      <th>query</th>\n",
       "      <th>q0</th>\n",
       "      <th>rank</th>\n",
       "      <th>score</th>\n",
       "      <th>system</th>\n",
       "      <th>docno</th>\n",
       "      <th>tira_task</th>\n",
       "      <th>tira_dataset</th>\n",
       "      <th>tira_first_stage_run_id</th>\n",
       "      <th>mean_health_score</th>\n",
       "      <th>median_health_score</th>\n",
       "      <th>mean_medical_score</th>\n",
       "      <th>median_medical_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>24.694701</td>\n",
       "      <td>my-system-name</td>\n",
       "      <td>clueweb12-1309wb-66-04887</td>\n",
       "      <td>workshop-on-open-web-search</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-15-03-57</td>\n",
       "      <td>85.1328</td>\n",
       "      <td>43.6221</td>\n",
       "      <td>133.2357</td>\n",
       "      <td>87.1570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>23.784552</td>\n",
       "      <td>my-system-name</td>\n",
       "      <td>clueweb12-1100wb-42-24270</td>\n",
       "      <td>workshop-on-open-web-search</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-15-03-57</td>\n",
       "      <td>73.6753</td>\n",
       "      <td>28.6258</td>\n",
       "      <td>123.6957</td>\n",
       "      <td>60.0157</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>23.752923</td>\n",
       "      <td>my-system-name</td>\n",
       "      <td>clueweb12-0109wb-83-09551</td>\n",
       "      <td>workshop-on-open-web-search</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-15-03-57</td>\n",
       "      <td>107.0973</td>\n",
       "      <td>55.9035</td>\n",
       "      <td>124.5779</td>\n",
       "      <td>84.6786</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>23.736344</td>\n",
       "      <td>my-system-name</td>\n",
       "      <td>clueweb12-1809wb-97-14794</td>\n",
       "      <td>workshop-on-open-web-search</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-15-03-57</td>\n",
       "      <td>72.5543</td>\n",
       "      <td>40.2659</td>\n",
       "      <td>98.6875</td>\n",
       "      <td>59.9471</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>cranberries urinary tract infections</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>23.512882</td>\n",
       "      <td>my-system-name</td>\n",
       "      <td>clueweb12-0104wb-93-33234</td>\n",
       "      <td>workshop-on-open-web-search</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-15-03-57</td>\n",
       "      <td>125.4036</td>\n",
       "      <td>70.0141</td>\n",
       "      <td>124.0446</td>\n",
       "      <td>74.0077</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50994</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>0</td>\n",
       "      <td>995</td>\n",
       "      <td>0.041757</td>\n",
       "      <td>my-system-name</td>\n",
       "      <td>clueweb12-1703wb-61-06943</td>\n",
       "      <td>workshop-on-open-web-search</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-15-03-57</td>\n",
       "      <td>49.7276</td>\n",
       "      <td>28.1963</td>\n",
       "      <td>67.4862</td>\n",
       "      <td>40.9963</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50995</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>0</td>\n",
       "      <td>996</td>\n",
       "      <td>0.033382</td>\n",
       "      <td>my-system-name</td>\n",
       "      <td>clueweb12-0101wb-99-20260</td>\n",
       "      <td>workshop-on-open-web-search</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-15-03-57</td>\n",
       "      <td>62.8626</td>\n",
       "      <td>36.7756</td>\n",
       "      <td>81.7350</td>\n",
       "      <td>51.5897</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50996</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>0</td>\n",
       "      <td>997</td>\n",
       "      <td>0.019441</td>\n",
       "      <td>my-system-name</td>\n",
       "      <td>clueweb12-0305wb-27-19517</td>\n",
       "      <td>workshop-on-open-web-search</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-15-03-57</td>\n",
       "      <td>55.2072</td>\n",
       "      <td>23.8824</td>\n",
       "      <td>60.7514</td>\n",
       "      <td>31.5140</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50997</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>0</td>\n",
       "      <td>998</td>\n",
       "      <td>0.008731</td>\n",
       "      <td>my-system-name</td>\n",
       "      <td>clueweb12-0300wb-97-10173</td>\n",
       "      <td>workshop-on-open-web-search</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-15-03-57</td>\n",
       "      <td>48.7505</td>\n",
       "      <td>14.5529</td>\n",
       "      <td>80.8725</td>\n",
       "      <td>45.4343</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50998</th>\n",
       "      <td>51</td>\n",
       "      <td>dehumidifiers asthma</td>\n",
       "      <td>0</td>\n",
       "      <td>999</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>my-system-name</td>\n",
       "      <td>clueweb12-0407wb-99-23432</td>\n",
       "      <td>workshop-on-open-web-search</td>\n",
       "      <td>clueweb12-trec-misinfo-2019-20240214-training</td>\n",
       "      <td>2024-03-19-15-03-57</td>\n",
       "      <td>38.2860</td>\n",
       "      <td>16.8330</td>\n",
       "      <td>63.8298</td>\n",
       "      <td>35.3084</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>50949 rows × 14 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      qid                                 query  q0  rank      score  \\\n",
       "0       1  cranberries urinary tract infections   0     1  24.694701   \n",
       "1       1  cranberries urinary tract infections   0     2  23.784552   \n",
       "2       1  cranberries urinary tract infections   0     3  23.752923   \n",
       "3       1  cranberries urinary tract infections   0     4  23.736344   \n",
       "4       1  cranberries urinary tract infections   0     5  23.512882   \n",
       "...    ..                                   ...  ..   ...        ...   \n",
       "50994  51                  dehumidifiers asthma   0   995   0.041757   \n",
       "50995  51                  dehumidifiers asthma   0   996   0.033382   \n",
       "50996  51                  dehumidifiers asthma   0   997   0.019441   \n",
       "50997  51                  dehumidifiers asthma   0   998   0.008731   \n",
       "50998  51                  dehumidifiers asthma   0   999   0.000000   \n",
       "\n",
       "               system                      docno                    tira_task  \\\n",
       "0      my-system-name  clueweb12-1309wb-66-04887  workshop-on-open-web-search   \n",
       "1      my-system-name  clueweb12-1100wb-42-24270  workshop-on-open-web-search   \n",
       "2      my-system-name  clueweb12-0109wb-83-09551  workshop-on-open-web-search   \n",
       "3      my-system-name  clueweb12-1809wb-97-14794  workshop-on-open-web-search   \n",
       "4      my-system-name  clueweb12-0104wb-93-33234  workshop-on-open-web-search   \n",
       "...               ...                        ...                          ...   \n",
       "50994  my-system-name  clueweb12-1703wb-61-06943  workshop-on-open-web-search   \n",
       "50995  my-system-name  clueweb12-0101wb-99-20260  workshop-on-open-web-search   \n",
       "50996  my-system-name  clueweb12-0305wb-27-19517  workshop-on-open-web-search   \n",
       "50997  my-system-name  clueweb12-0300wb-97-10173  workshop-on-open-web-search   \n",
       "50998  my-system-name  clueweb12-0407wb-99-23432  workshop-on-open-web-search   \n",
       "\n",
       "                                        tira_dataset tira_first_stage_run_id  \\\n",
       "0      clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-15-03-57   \n",
       "1      clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-15-03-57   \n",
       "2      clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-15-03-57   \n",
       "3      clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-15-03-57   \n",
       "4      clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-15-03-57   \n",
       "...                                              ...                     ...   \n",
       "50994  clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-15-03-57   \n",
       "50995  clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-15-03-57   \n",
       "50996  clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-15-03-57   \n",
       "50997  clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-15-03-57   \n",
       "50998  clueweb12-trec-misinfo-2019-20240214-training     2024-03-19-15-03-57   \n",
       "\n",
       "       mean_health_score  median_health_score  mean_medical_score  \\\n",
       "0                85.1328              43.6221            133.2357   \n",
       "1                73.6753              28.6258            123.6957   \n",
       "2               107.0973              55.9035            124.5779   \n",
       "3                72.5543              40.2659             98.6875   \n",
       "4               125.4036              70.0141            124.0446   \n",
       "...                  ...                  ...                 ...   \n",
       "50994            49.7276              28.1963             67.4862   \n",
       "50995            62.8626              36.7756             81.7350   \n",
       "50996            55.2072              23.8824             60.7514   \n",
       "50997            48.7505              14.5529             80.8725   \n",
       "50998            38.2860              16.8330             63.8298   \n",
       "\n",
       "       median_medical_score  \n",
       "0                   87.1570  \n",
       "1                   60.0157  \n",
       "2                   84.6786  \n",
       "3                   59.9471  \n",
       "4                   74.0077  \n",
       "...                     ...  \n",
       "50994               40.9963  \n",
       "50995               51.5897  \n",
       "50996               31.5140  \n",
       "50997               45.4343  \n",
       "50998               35.3084  \n",
       "\n",
       "[50949 rows x 14 columns]"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bm25_ranking = bm25_health(topics)\n",
    "bm25_ranking"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "rankings = []\n",
    "thresholds = list(range(0, 60, 5))\n",
    "for threshold in thresholds:\n",
    "    rankings.append(bm25_ranking.loc[bm25_ranking[\"median_health_score\"] >= threshold])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "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>nDCG@10</th>\n",
       "      <th>MRR</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Health Score Threshold</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.452698</td>\n",
       "      <td>0.808510</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.455017</td>\n",
       "      <td>0.808534</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.455339</td>\n",
       "      <td>0.809381</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.456980</td>\n",
       "      <td>0.810169</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.454947</td>\n",
       "      <td>0.811246</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.443900</td>\n",
       "      <td>0.812434</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.439326</td>\n",
       "      <td>0.812818</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>0.439902</td>\n",
       "      <td>0.804937</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>0.413496</td>\n",
       "      <td>0.726326</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>0.387697</td>\n",
       "      <td>0.707630</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50</th>\n",
       "      <td>0.364366</td>\n",
       "      <td>0.695315</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>55</th>\n",
       "      <td>0.341249</td>\n",
       "      <td>0.699365</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         nDCG@10       MRR\n",
       "Health Score Threshold                    \n",
       "0                       0.452698  0.808510\n",
       "5                       0.455017  0.808534\n",
       "10                      0.455339  0.809381\n",
       "15                      0.456980  0.810169\n",
       "20                      0.454947  0.811246\n",
       "25                      0.443900  0.812434\n",
       "30                      0.439326  0.812818\n",
       "35                      0.439902  0.804937\n",
       "40                      0.413496  0.726326\n",
       "45                      0.387697  0.707630\n",
       "50                      0.364366  0.695315\n",
       "55                      0.341249  0.699365"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results = pt.Experiment(\n",
    "    rankings,\n",
    "    topics,\n",
    "    qrels.rename({\"label\": \"relevance\"}, axis=1),\n",
    "    ['ndcg_cut.10', 'recip_rank'],\n",
    "    names=thresholds,\n",
    ").rename({\"name\": \"Health Score Threshold\", \"ndcg_cut.10\": \"nDCG@10\", \"recip_rank\": \"MRR\"}, axis=1)\n",
    "results = results.set_index(\"Health Score Threshold\")\n",
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 560x280 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "scale = 0.7\n",
    "fig, axes = plt.subplots(1, 2, figsize=(8 * scale, 4 * scale))\n",
    "for idx, metric in enumerate((\"nDCG@10\", \"MRR\")):\n",
    "    ax = axes[idx]\n",
    "    ax.plot(results.index, results.loc[:, metric], marker=\"o\")\n",
    "    ax.set_ylabel(metric)\n",
    "    ax.set_xlabel(\"Health Score Threshold\")\n",
    "fig.suptitle(\"TREC Health Misinformation\")\n",
    "fig.tight_layout()"
   ]
  }
 ],
 "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": "3.10.12"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}