DR-LACR: A Dual-Role Framework for Legal-Article-Based Contract Risk Review
Authors:
- Xiaorui LiXLXiaorui Li
- Siwen ZhaoSZSiwen Zhao
- Zhe ChenZCZhe Chen
- Sen LinSLSen Lin
- Feng QiaoFQFeng Qiao
- Hailong ChenHCHailong Chen
- Ziheng ZhuZZZiheng Zhu
- Yujun LiYLYujun Li
Preprint
- 18 April 2026
- preprint
- Published by Elsevier in SSRN Electronic Journal
Abstract
Contract risk review is a critical task in legal practice, as contractual risks may lead to compliance violations, financial losses, and legal disputes. However, it is challenging because it requires reliably pairing contractual clauses with relevant legal articles while distinguishing role-specific risk. Most existing contract review approaches often fail to accurately pair clauses with the appropriate legal articles and to provide reliable role-specific risk review.To address this gap, we propose DR-LACR, a Dual-Role Framework for Legal-Article-Based Contract Risk Review. DR-LACR consists of two coupled modules: a legal article retrieval module that performs structured legal information extraction and enhances both contractual clauses and legal articles for accurate article retrieval, and a dual-role risk review module that performs role-specific risk review based on the retrieved articles.We evaluate DR-LACR on CLaD, a newly constructed clause-level contract risk review dataset with explicit legal article alignment and dual-role annotations. Experimental results show that DR-LACR consistently outperforms strong baselines. In legal article retrieval, DR-LACR achieves a relative improvement of over 15% in R@1 compared with the best baseline. In contract risk review, DR-LACR reduces MAE by approximately 12% compared with retrieval-based models, and by more than 60% relative to DeepSeek-v3.Keywords
This publication has 11 references indexed in Scilit:
- ACORD: An Expert-Annotated Retrieval Dataset for Legal Contract Drafting2025
- Design of Legal Judgment Prediction on Knowledge Graph and Deep Learning2024
- LA-MGFM: A legal judgment prediction method via sememe-enhanced graph neural networks and multi-graph fusion mechanismInformation Processing & Management, 2023
- Query2doc: Query Expansion with Large Language Models2023
- Precise Zero-Shot Dense Retrieval without Relevance Labels2023
- LeXFiles and LegalLAMA: Facilitating English Multinational Legal Language Model Development2023
- Lawformer: A pre-trained language model for Chinese legal long documentsAI Open, 2021
- How Does NLP Benefit Legal System: A Summary of Legal Artificial Intelligence2020
- Neural Legal Judgment Prediction in English2019
- Deep learning in law: early adaptation and legal word embeddings trained on large corporaArtificial Intelligence and Law, 2018