Datasets:

Modalities:
Text
Formats:
json
ArXiv:
Libraries:
Datasets
pandas
License:
cdancette commited on
Commit
f6d5c55
·
verified ·
1 Parent(s): 5453887

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +33 -25
README.md CHANGED
@@ -4,52 +4,59 @@ configs:
4
  - config_name: english-full
5
  data_files:
6
  - split: train
7
- path: "english/train.jsonl"
8
  - split: test
9
- path: "english/test.jsonl"
10
  - config_name: english-qi
11
  data_files:
12
  - split: train
13
- path: "english/qi_train.jsonl"
14
  - split: test
15
- path: "english/qi_test.jsonl"
16
  - config_name: english-dp
17
  data_files:
18
  - split: train
19
- path: "english/dp_train.jsonl"
20
  - split: test
21
- path: "english/dp_test.jsonl"
22
  - config_name: english-custom
23
  data_files:
24
  - split: train
25
- path: "english/custom_train.jsonl"
26
  - split: test
27
- path: "english/custom_test.jsonl"
28
  - config_name: french-full
29
  data_files:
30
  - split: train
31
- path: "french/train.jsonl"
32
  - split: test
33
- path: "french/test.jsonl"
34
  - config_name: french-qi
35
  data_files:
36
  - split: train
37
- path: "french/qi_train.jsonl"
38
  - split: test
39
- path: "french/qi_test.jsonl"
40
  - config_name: french-dp
41
  data_files:
42
  - split: train
43
- path: "french/dp_train.jsonl"
44
  - split: test
45
- path: "french/dp_test.jsonl"
46
  - config_name: french-custom
47
  data_files:
48
  - split: train
49
- path: "french/custom_train.jsonl"
50
  - split: test
51
- path: "french/custom_test.jsonl"
52
-
 
 
 
 
 
 
 
53
  ---
54
 
55
 
@@ -61,12 +68,6 @@ Paper: [https://arxiv.org/abs/2405.14654](https://arxiv.org/abs/2405.14654)
61
 
62
  ## Dataset Details
63
 
64
- In the expanding field of language model applications, medical knowledge representation remains a significant challenge due to the specialized nature of the domain.
65
- Large language models, such as GPT-4, obtain reasonable scores on medical question-answering tasks, but smaller models are far behind.
66
- 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.
67
- 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.
68
- We show the benefits of our training strategy on a medical answering question dataset.
69
- The study's findings highlight the potential of small language models in the medical domain when appropriately fine-tuned.
70
 
71
  ### Dataset Description
72
 
@@ -78,7 +79,7 @@ The questions can be:
78
 
79
  There are two versions of this dataset: the **french** and the **english** versions.
80
  The **french** split is the original dataset version.
81
- The english version is an automated translation (using Azure Translation api), hence, it might contain translation errors.
82
 
83
  ### Use this dataset
84
 
@@ -107,7 +108,14 @@ ds_english_custom = load_dataset("raidium/ECN-QA", "english-custom")
107
  - **Developed by:** Raidium
108
  - **License:** Apache 2.0
109
 
110
- ### Dataset Sources [optional]
 
 
 
 
 
 
 
111
 
112
  <!-- Provide the basic links for the model. -->
113
 
 
4
  - config_name: english-full
5
  data_files:
6
  - split: train
7
+ path: english/train.jsonl
8
  - split: test
9
+ path: english/test.jsonl
10
  - config_name: english-qi
11
  data_files:
12
  - split: train
13
+ path: english/qi_train.jsonl
14
  - split: test
15
+ path: english/qi_test.jsonl
16
  - config_name: english-dp
17
  data_files:
18
  - split: train
19
+ path: english/dp_train.jsonl
20
  - split: test
21
+ path: english/dp_test.jsonl
22
  - config_name: english-custom
23
  data_files:
24
  - split: train
25
+ path: english/custom_train.jsonl
26
  - split: test
27
+ path: english/custom_test.jsonl
28
  - config_name: french-full
29
  data_files:
30
  - split: train
31
+ path: french/train.jsonl
32
  - split: test
33
+ path: french/test.jsonl
34
  - config_name: french-qi
35
  data_files:
36
  - split: train
37
+ path: french/qi_train.jsonl
38
  - split: test
39
+ path: french/qi_test.jsonl
40
  - config_name: french-dp
41
  data_files:
42
  - split: train
43
+ path: french/dp_train.jsonl
44
  - split: test
45
+ path: french/dp_test.jsonl
46
  - config_name: french-custom
47
  data_files:
48
  - split: train
49
+ path: french/custom_train.jsonl
50
  - split: test
51
+ path: french/custom_test.jsonl
52
+ task_categories:
53
+ - multiple-choice
54
+ language:
55
+ - fr
56
+ - en
57
+ tags:
58
+ - medical question answering
59
+ pretty_name: ecn-qa
60
  ---
61
 
62
 
 
68
 
69
  ## Dataset Details
70
 
 
 
 
 
 
 
71
 
72
  ### Dataset Description
73
 
 
79
 
80
  There are two versions of this dataset: the **french** and the **english** versions.
81
  The **french** split is the original dataset version.
82
+
83
 
84
  ### Use this dataset
85
 
 
108
  - **Developed by:** Raidium
109
  - **License:** Apache 2.0
110
 
111
+
112
+ ### Warnings
113
+
114
+ - Some questions require images to be answered. They have not been filtered out in this dataset, to it is impossible to get 100% accuracy on this dataset.
115
+ - The english version is an automated translation (using Azure Translation api), hence, it might contain translation errors.
116
+
117
+
118
+ ### Dataset Sources
119
 
120
  <!-- Provide the basic links for the model. -->
121