Text Simplification System for Legal Contract Review

People tend to avoid reading contracts due to their complexity. As such, this research tackles the challenge of improving the accessibility and readability of contracts, which are often lengthy and difficult to understand. This study proposes a system that integrates automated contract review with text simplification. The system leverages a language model fine-tuned on legal data to extract salient clauses from contracts. Complex words are then replaced with simpler alternatives generated by language models trained on legal documents. Moreover, the simplified output is further refined by breaking down the text into shorter sentences based on their semantic hierarchy. Initial results show that the readability of the simplified contracts improved, making them understandable for 10th graders instead of requiring a postgraduate level of education. Human evaluations were generally positive, although the observed improvements were relatively minor. The research concludes that integrating text simplification with automated contract review has the potential to enhance contract readability, but more research is necessary to improve the quality of simplification further.