whaleloops commited on
Commit
6805f21
1 Parent(s): 88fd15a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +14 -6
README.md CHANGED
@@ -1,16 +1,25 @@
1
  ---
2
  license: apache-2.0
 
 
 
 
 
3
  ---
4
 
5
- KEPTlongfomer pretrained using [contrastive learning](https://arxiv.org/pdf/2210.03304.pdf).
 
 
6
 
 
 
7
 
8
- First, The model was first inited from [clinical longformer](https://huggingface.co/yikuan8/Clinical-Longformer).
9
-
10
- And then pretrained with Hierarchical Self-Alignment Pretrainumls (HSAP) using Knowledge Graph UMLS.
11
  This includes (a) Hierarchy, (b) Synonym, (c) Abbreviation. For more info, see section 3.3 in [paper](https://arxiv.org/pdf/2210.03304.pdf).
 
12
 
13
- See [here](https://github.com/whaleloops/KEPT/tree/rerank300) for how to use this on auto ICD coding.
 
14
 
15
  With the following result:
16
  | Metric | Score |
@@ -42,5 +51,4 @@ With the following result:
42
  |acc_macro| =0.08639139221100567|
43
 
44
 
45
- A sister model is available [here](https://huggingface.co/whaleloops/KEPTlongformer-PMM3/).
46
 
 
1
  ---
2
  license: apache-2.0
3
+ language: "en"
4
+ tags:
5
+ - longformer
6
+ - clinical
7
+ - biomedical
8
  ---
9
 
10
+ <span style="font-size:larger;">**KEPTlongfomer**</span> is a medical knowledge enhanced version of Longformer that was further pre-trained using [contrastive learning](https://arxiv.org/pdf/2210.03304.pdf).
11
+ The model achieves SOTA performance on auto ICD coding on MIMIC-III as of 11/12/2022.
12
+ A sister model for better performance is available [here](https://huggingface.co/whaleloops/KEPTlongformer-PMM3/).
13
 
14
+ ### Pre-training
15
+ We initialized this model from [clinical longformer](https://huggingface.co/yikuan8/Clinical-Longformer).
16
 
17
+ And then pretrained with Hierarchical Self-Alignment Pretrain (HSAP) using Knowledge Graph UMLS.
 
 
18
  This includes (a) Hierarchy, (b) Synonym, (c) Abbreviation. For more info, see section 3.3 in [paper](https://arxiv.org/pdf/2210.03304.pdf).
19
+ The learning rate was 5e-5, weight decay was 0.01, adam epsilon was 1e-5.
20
 
21
+ ### Usage
22
+ See our [github](https://github.com/whaleloops/KEPT/tree/rerank300) for how to use this with prompts on auto ICD coding.
23
 
24
  With the following result:
25
  | Metric | Score |
 
51
  |acc_macro| =0.08639139221100567|
52
 
53
 
 
54