Merge branch 'main' of https://huggingface.co/nsi319/legal-led-base-16384 into main
Browse files
README.md
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
|
| 3 |
+
## LED for legal domain
|
| 4 |
+
This is Longformer Encoder Decoder [led-base-16384] model for the legal domain, trained for long document abstractive summarization task.
|
| 5 |
+
|
| 6 |
+
## Training data
|
| 7 |
+
|
| 8 |
+
The legal-led-base-16384 model was trained on [sec-litigation-releases](https://www.sec.gov/litigation/litreleases.htm) dataset consisting more than 4000 litigation releases and complaints.
|
| 9 |
+
|
| 10 |
+
## Evaluation results
|
| 11 |
+
|
| 12 |
+
When the model is used for summarizing legal documents, achieves the following results:
|
| 13 |
+
|
| 14 |
+
Test results :
|
| 15 |
+
|
| 16 |
+
| Model | Rouge1-Precision | Rouge1 |
|
| 17 |
+
|:-----:|:-----:|:-----:|
|
| 18 |
+
| legal-led-base-16384 | 58.42|54.89 |
|