|
--- |
|
language: en |
|
datasets: |
|
- cuad |
|
--- |
|
|
|
# Finetuned legal contract review QA model based 👩⚖️ 📑 |
|
|
|
Best model presented in the master thesis [*Exploring CUAD using RoBERTa span-selection QA models for legal contract review*](https://github.com/gustavhartz/transformers-legal-tasks) for QA on the Contract Understanding Atticus Dataset. Full training logic and associated thesis available through link. |
|
|
|
Outperform the most popular HF cuad model [Rakib/roberta-base-on-cuad](https://huggingface.co/Rakib/roberta-base-on-cuad) and is the best model for CUAD on Hugging Face 26/06/2022 |
|
|
|
| **Model name** | **Top 1 Has Ans F1** | **Top 3 Has Ans F1** | |
|
|-----------------------------------------|----------------------|----------------------| |
|
| gustavhartz/roberta-base-cuad-finetuned | 85.68 | 94.06 | |
|
| Rakib/roberta-base-on-cuad | 81.26 | 92.48 | |
|
|
|
|
|
For questions etc. go through the Github repo :) |
|
|
|
### Citation |
|
|
|
If you found the code of thesis helpful you can please cite it :) |
|
``` |
|
@thesis{ha2022, |
|
author = {Hartz, Gustav Selfort}, |
|
title = {Exploring CUAD using RoBERTa span-selection QA models for legal contract review}, |
|
language = {English}, |
|
format = {thesis}, |
|
year = {2022}, |
|
publisher = {DTU Department of Applied Mathematics and Computer Science} |
|
} |
|
``` |
