Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -27,3 +27,57 @@ configs:
|
|
27 |
path: "custom_test.jsonl"
|
28 |
|
29 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
path: "custom_test.jsonl"
|
28 |
|
29 |
---
|
30 |
+
|
31 |
+
|
32 |
+
# Model Card for Raidium ECN-QA
|
33 |
+
|
34 |
+
|
35 |
+
The dataset is introduced in the paper "Efficient Medical Question Answering with Knowledge-Augmented Question Generation".
|
36 |
+
|
37 |
+
Paper: [https://arxiv.org/abs/2405.14654](https://arxiv.org/abs/2405.14654)
|
38 |
+
|
39 |
+
## Dataset Details
|
40 |
+
|
41 |
+
In the expanding field of language model applications, medical knowledge representation remains a significant challenge due to the specialized nature of the domain.
|
42 |
+
Large language models, such as GPT-4, obtain reasonable scores on medical question-answering tasks, but smaller models are far behind.
|
43 |
+
In this work, we introduce a method to improve the proficiency of a small language model in the medical domain by employing a two-fold approach.
|
44 |
+
We first fine-tune the model on a corpus of medical textbooks. Then, we use GPT-4 to generate questions similar to the downstream task, prompted with textbook knowledge, and use them to fine-tune the model.
|
45 |
+
We show the benefits of our training strategy on a medical answering question dataset.
|
46 |
+
The study's findings highlight the potential of small language models in the medical domain when appropriately fine-tuned.
|
47 |
+
|
48 |
+
### Dataset Description
|
49 |
+
|
50 |
+
The dataset contains medical questions of different types. It was built from passed ECN exams (french medical examination) and questions created by [FreeCN](https://www.freecn.io/).
|
51 |
+
The questions can be:
|
52 |
+
- IQ (individual question) containing a question and several propositions that can be right or wrong
|
53 |
+
- Custom which are IQ created by FreeCN
|
54 |
+
- PQ (progressive questions) containing a case with an introduction and several following questions with multiple propositions
|
55 |
+
|
56 |
+
|
57 |
+
- **Developed by:** Raidium
|
58 |
+
- **License:** Apache 2.0
|
59 |
+
|
60 |
+
### Dataset Sources [optional]
|
61 |
+
|
62 |
+
<!-- Provide the basic links for the model. -->
|
63 |
+
|
64 |
+
- **Repository:** [https://github.com/raidium-med/MQG]
|
65 |
+
- **Paper:** [https://arxiv.org/abs/2405.14654](https://arxiv.org/abs/2405.14654)
|
66 |
+
|
67 |
+
|
68 |
+
## Citation
|
69 |
+
|
70 |
+
|
71 |
+
**BibTeX:**
|
72 |
+
```
|
73 |
+
@article{khlaut2024efficient,
|
74 |
+
title={Efficient Medical Question Answering with Knowledge-Augmented Question Generation},
|
75 |
+
author={Khlaut, Julien and Dancette, Corentin and Ferreres, Elodie and Bennani, Alaedine and H{\'e}rent, Paul and Manceron, Pierre},
|
76 |
+
journal={Clinical NLP Workshop, NAACL 2024},
|
77 |
+
year={2024}
|
78 |
+
}
|
79 |
+
```
|
80 |
+
|
81 |
+
## Dataset Card Contact
|
82 |
+
|
83 |
+
julien.khlaut at raidium.fr
|