Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,167 @@
|
|
1 |
---
|
2 |
-
license:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
license: cc-by-4.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
- es
|
6 |
+
- fr
|
7 |
+
- it
|
8 |
+
tags:
|
9 |
+
- casimedicos
|
10 |
+
- explainability
|
11 |
+
- medical exams
|
12 |
+
- medical question answering
|
13 |
+
- multilinguality
|
14 |
+
- LLMs
|
15 |
+
- LLM
|
16 |
+
pretty_name: MedExpQA
|
17 |
+
configs:
|
18 |
+
- config_name: en
|
19 |
+
data_files:
|
20 |
+
- split: train
|
21 |
+
path:
|
22 |
+
- data/en/dev.en.train.casimedicos.rag.jsonl
|
23 |
+
- split: validation
|
24 |
+
path:
|
25 |
+
- data/en/dev.en.casimedicos.rag.jsonl
|
26 |
+
- split: test
|
27 |
+
path:
|
28 |
+
- data/en/test.en.casimedicos.rag.jsonl
|
29 |
+
- config_name: es
|
30 |
+
data_files:
|
31 |
+
- split: train
|
32 |
+
path:
|
33 |
+
- data/es/dev.es.train.casimedicos.rag.jsonl
|
34 |
+
- split: validation
|
35 |
+
path:
|
36 |
+
- data/es/dev.es.casimedicos.rag.jsonl
|
37 |
+
- split: test
|
38 |
+
path:
|
39 |
+
- data/es/test.es.casimedicos.rag.jsonl
|
40 |
+
- config_name: fr
|
41 |
+
data_files:
|
42 |
+
- split: train
|
43 |
+
path:
|
44 |
+
- data/fr/dev.fr.train.casimedicos.rag.jsonl
|
45 |
+
- split: validation
|
46 |
+
path:
|
47 |
+
- data/fr/dev.fr.casimedicos.rag.jsonl
|
48 |
+
- split: test
|
49 |
+
path:
|
50 |
+
- data/fr/test.fr.casimedicos.rag.jsonl
|
51 |
+
- config_name: it
|
52 |
+
data_files:
|
53 |
+
- split: train
|
54 |
+
path:
|
55 |
+
- data/it/dev.it.train.casimedicos.rag.jsonl
|
56 |
+
- split: validation
|
57 |
+
path:
|
58 |
+
- data/it/dev.it.casimedicos.rag.jsonl
|
59 |
+
- split: test
|
60 |
+
path:
|
61 |
+
- data/it/test.it.casimedicos.rag.jsonl
|
62 |
+
task_categories:
|
63 |
+
- text-generation
|
64 |
+
- question-answering
|
65 |
+
size_categories:
|
66 |
+
- 1K<n<10K
|
67 |
---
|
68 |
+
|
69 |
+
<p align="center">
|
70 |
+
<br>
|
71 |
+
<img src="http://www.ixa.eus/sites/default/files/anitdote.png" style="height: 200px;">
|
72 |
+
<br>
|
73 |
+
|
74 |
+
# MexExpQA: Benchmarking Medical QA with reference gold explanations and Retrieval Augmented Generation Methods (RAG)
|
75 |
+
|
76 |
+
We present a new multilingual parallel medical benchmark, MedExpQA, for the evaluation of LLMs on Medical Question Answering.
|
77 |
+
This dataset can be used for various NLP tasks including: **Medical Question Answering**, **Explanatory Argument Extraction** or **Explanation Generation**.
|
78 |
+
|
79 |
+
Although the design of MedExpQA is independent of the specific dataset, for the first version of the MedExpQA benchmark we leverage the commented MIR exams,
|
80 |
+
from the [Antidote CasiMedicos dataset which includes gold reference explanations](https://huggingface.co/datasets/HiTZ/casimedicos-exp), which is currently
|
81 |
+
available for 4 languages: **English, French, Italian and Spanish**.
|
82 |
+
|
83 |
+
<table style="width:33%">
|
84 |
+
<tr>
|
85 |
+
<th>Antidote CasiMedicos splits</th>
|
86 |
+
<tr>
|
87 |
+
<td>train</td>
|
88 |
+
<td>434</td>
|
89 |
+
</tr>
|
90 |
+
<tr>
|
91 |
+
<td>validation</td>
|
92 |
+
<td>63</td>
|
93 |
+
</tr>
|
94 |
+
<tr>
|
95 |
+
<td>test</td>
|
96 |
+
<td>125</td>
|
97 |
+
</tr>
|
98 |
+
</table>
|
99 |
+
|
100 |
+
- 📖 Paper:[MedExpQA: Multilingual Benchmarking of Large Language Models for Medical Question Answering](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4780937)
|
101 |
+
- 💻 Github Repo (Data and Code): [https://github.com/hitz-zentroa/MedExpQA](https://github.com/hitz-zentroa/MedExpQA)
|
102 |
+
- 🌐 Project Website: [https://univ-cotedazur.eu/antidote](https://univ-cotedazur.eu/antidote)
|
103 |
+
- Funding: CHIST-ERA XAI 2019 call. Antidote (PCI2020-120717-2) funded by MCIN/AEI /10.13039/501100011033 and by European Union NextGenerationEU/PRTR
|
104 |
+
|
105 |
+
|
106 |
+
## Example
|
107 |
+
|
108 |
+
<p align="center">
|
109 |
+
<img src="https://github.com/ixa-ehu/antidote-casimedicos/blob/main/casimedicos-exp.png?raw=true" style="height: 650px;">
|
110 |
+
</p>
|
111 |
+
|
112 |
+
In this repository you can find the following data:
|
113 |
+
|
114 |
+
- **casimedicos-raw**: The textual content including Clinical Case (C), Question (Q), Possible Answers (P), and Explanation (E) as shown in the example above.
|
115 |
+
- **casimedicos-exp**: The manual annotations linking the explanations of the correct and incorrect possible answers.
|
116 |
+
- **MedExpQA**: benchmark for Medical QA based on gold reference explanations from casimedicos-exp and knowledge automatically extracted using RAG methods.
|
117 |
+
|
118 |
+
## Data Explanation
|
119 |
+
|
120 |
+
The following attributes composed **casimedicos-raw**:
|
121 |
+
|
122 |
+
- **id**: unique doc identifier.
|
123 |
+
- **year**: year in which the exam was published by the Spanish Ministry of Health.
|
124 |
+
- **question_id_specific**: id given to the original exam published by the Spanish Ministry of Health.
|
125 |
+
- **full_question**: Clinical Case (C) and Question (Q) as illustrated in the example document above.
|
126 |
+
- **full answer**: Full commented explanation (E) as illustrated in the example document above.
|
127 |
+
- **type**: medical speciality.
|
128 |
+
- **options**: Possible Answers (P) as illustrated in the example document above.
|
129 |
+
- **correct option**: solution to the exam question.
|
130 |
+
|
131 |
+
Additionally, the following jsonl attribute was added to create **casimedicos-exp**:
|
132 |
+
|
133 |
+
- **explanations**: for each possible answer above, manual annotation states whether:
|
134 |
+
1. the explanation for each possible answer exists in the full comment (E) and
|
135 |
+
2. if present, then we provide character and token offsets plus the text corresponding to the explanation for each possible answer.
|
136 |
+
|
137 |
+
For **MedExpQA** benchmarking we have added the following elements in the data:
|
138 |
+
|
139 |
+
- **rag**
|
140 |
+
1. **clinical_case_options**: etc.
|
141 |
+
|
142 |
+
|
143 |
+
|
144 |
+
|
145 |
+
## Citation
|
146 |
+
|
147 |
+
If you use Antidote CasiMedicos dataset then please **cite the following paper**:
|
148 |
+
|
149 |
+
```bibtex
|
150 |
+
@inproceedings{Agerri2023HiTZAntidoteAE,
|
151 |
+
title={HiTZ@Antidote: Argumentation-driven Explainable Artificial Intelligence for Digital Medicine},
|
152 |
+
author={Rodrigo Agerri and I{\~n}igo Alonso and Aitziber Atutxa and Ander Berrondo and Ainara Estarrona and Iker Garc{\'i}a-Ferrero and Iakes Goenaga and Koldo Gojenola and Maite Oronoz and Igor Perez-Tejedor and German Rigau and Anar Yeginbergenova},
|
153 |
+
booktitle={SEPLN 2023: 39th International Conference of the Spanish Society for Natural Language Processing.},
|
154 |
+
year={2023}
|
155 |
+
}
|
156 |
+
|
157 |
+
@misc{goenaga2023explanatory,
|
158 |
+
title={Explanatory Argument Extraction of Correct Answers in Resident Medical Exams},
|
159 |
+
author={Iakes Goenaga and Aitziber Atutxa and Koldo Gojenola and Maite Oronoz and Rodrigo Agerri},
|
160 |
+
year={2023},
|
161 |
+
eprint={2312.00567},
|
162 |
+
archivePrefix={arXiv}
|
163 |
+
}
|
164 |
+
```
|
165 |
+
|
166 |
+
**Contact**: [Iñigo Alonso](https://hitz.ehu.eus/en/node/282) and [Rodrigo Agerri](https://ragerri.github.io/)
|
167 |
+
HiTZ Center - Ixa, University of the Basque Country UPV/EHU
|