A Knowledge Graph Modeling Approach for Augmenting Language Model-Based Contract Risk Identification

Abstract: Contract risk identification is essential for preventing disputes and losses in construction industry. Large language models (LLMs) have achieved substantial impact across a wide range of natural language processing tasks, which presents an opportunity for automating contract review without heavy data processing and feature engineering. However, LLMs still has difficulty in recalling facts while generating knowledge-grounded analysis, especially when related to complex domain knowledge. This paper introduces a Knowledge Graph (KG) modeling approach to enhance the LLM-based automated contract risk identification. A case study demonstrates that our approach exhibits enhanced performance on risk identification tasks compared to non-augmentation scenario.