RoBERTa Fine-tuned for Legal Contract Clause Extraction

Model Description

This model is a fine-tuned version of RoBERTa on the CUAD dataset for extracting 6 key clause types from legal contracts:

• Governing Law
• Expiration Date
• Effective Date
• Anti-Assignment
• Cap On Liability
• License Grant

Training Details

• Base Model: roberta-base
• Training Data: 30% of CUAD dataset (357 contracts)
• Epochs: 2
• Final Training Loss: 0.348

Usage

from transformers import pipeline

qa_pipeline = pipeline("question-answering", model="srraghuram/roberta-cuad-clause-extraction")

result = qa_pipeline( question="Highlight the parts related to Governing Law", context="Your contract text here..." ) print(result)

Limitations

• Trained on commercial contracts, may not generalize to other legal document types
• Performance varies by clause type