《计算机应用》唯一官方网站 ›› 2026, Vol. 46 ›› Issue (9): 2820-2826.DOI: 10.11772/j.issn.1001-9081.2025091093

• 人工智能 • 上一篇    

基于LLM的动态规则生成与神经符号推理合同审查框架

霍耀冉1(), 李会芝2, 曾大和1, 唐震宇1, 刘辰梅3, 吴岩1, 张啸1, 陈思芹4, 杨厚晖4   

  1. 1.四川科锐得电力通信技术有限公司，成都 610041
    2.国网四川省电力公司，成都 610041
    3.国网四川综合服务中心，成都 610041
    4.北京英孚泰克信息技术股份有限公司，北京 100089
  • 收稿日期:2025-09-19 修回日期:2025-11-14 接受日期:2025-11-16 发布日期:2025-12-09 出版日期:2026-09-10
  • 通讯作者: 霍耀冉
  • 作者简介:霍耀冉（1987—），男，河南商丘人，高级工程师，博士，主要研究方向：电力人工智能、合同合规智能、大语言模型应用
    李会芝（1986—），女，四川什邡人，高级经济师，硕士，主要研究方向：电力法治智能、合规管理
    曾大和（1973—），男，四川中江人，高级工程师，主要研究方向：知识工程、电力数字化
    唐震宇（1989—），男，四川自贡人，高级工程师，硕士，主要研究方向：数据治理、自动化数据处理、审查自动化
    刘辰梅（1994—），女，四川仁寿人，经济师，硕士，主要研究方向：电力法治智能、合规管理
    吴岩（1982—），男，陕西汉中人，高级工程师，主要研究方向：机器学习、自然语言处理、电力文本挖掘
    张啸（1988—），男，四川成都人，工程师，主要研究方向：合同法务智能、合规知识图谱
    陈思芹（1999—），女，四川内江人，工程师，硕士，主要研究方向：人工智能应用、法务智能
    杨厚晖（1974—），女，四川都江堰人，中级经济师，主要研究方向：经济学。
  • 基金资助:
    国网四川省电力公司科技项目(52199925000N)

Contract review framework based on dynamic rule generation and neuro-symbolic reasoning using large language models

Yaoran HUO1(), Huizhi LI2, Dahe ZENG1, Zhenyu TANG1, Chenmei LIU3, Yan WU1, Xiao ZHANG1, Siqin CHEN4, Houhui YANG4   

  1. 1.Sichuan Keruide Power Communication Technology Company Limited，Chengdu Sichuan 610041，China
    2.State Grid Sichuan Electric Power Company，Chengdu Sichuan 610041，China
    3.State Grid Sichuan Electric Power Company Service Center，Chengdu Sichuan 610041，China
    4.Information Technology （Beijing） Company Limited，Beijing 100089，China
  • Received:2025-09-19 Revised:2025-11-14 Accepted:2025-11-16 Online:2025-12-09 Published:2026-09-10
  • Contact: Yaoran HUO
  • About author:HUO Yaoran， born in 1987， Ph. D.， senior engineer. His research interests include power AI， contract-compliance intelligence， LLM applications.
    LI Huizhi， born in 1986， M. S.， senior economist. Her research interests include electricity legal intelligence， compliance management.
    ZENG Dahe， born in 1973， senior engineer. His research interests include knowledge engineering， power digitalization.
    TANG Zhenyu， born in 1989， M. S.， senior engineer. His research interests include data governance， automated data processing， review automation.
    LIU Chenmei， born in 1994， M. S.， economist. Her research interests include electricity legal intelligence， compliance management.
    WU Yan， born in 1982， senior engineer. His research interests include machine learning， natural language processing， power-text mining.
    ZHANG Xiao， born in 1988， engineer. His research interests include contract legal intelligence， compliance knowledge graphs.
    CHEN Siqin， born in 1999， M. S.， engineer. Her research interests include AI applications， legal intelligence.
    YANG Houhui， born in 1974， intermediate economist. Her research interests include economics.
  • Supported by:
    Science and Technology Project of State Grid Sichuan Electric Power Company(52199925000N)

RichHTML

PDF

摘要：

自动化合同审查在动态适配多样化条款与准确执行复杂审查规则方面仍存在局限。因此，提出一种基于大语言模型（LLM）的动态规则生成与神经符号推理合同审查框架（DRNS-Review）。首先，设计动态规则生成方法，结合领域知识与合同上下文，利用LLM即时生成自然语言与符号化审查规则；其次，构建神经符号推理机制自动提取局部事实，通过符号匹配进行规则定位，并结合符号归约和LLM语义执行实现可解释推理；最后，整合多段推理结果，形成结构化与自然语言双轨的审查报告。实验结果表明，相较于基线方法LLM-Direct和ManualRule-LLM，DRNS-Review在风险识别F1值上分别提升了13.0和8.3个百分点，在风险等级准确率（RLA）上分别提升了13.3和8.7个百分点。可见，动态规则生成与神经符号推理的结合能有效提升合同审查的准确性与适应性，为智能合规类应用提供了可推广的技术路径。

关键词: 合同审查, 大语言模型, 动态规则生成, 神经符号推理, 智能合规

Abstract:

Automated contract review has limitations in adapting to diverse clauses dynamically and executing complex review rules accurately. Therefore， a Dynamic Rule generation and Neuro-Symbolic reasoning framework for contract Review （DRNS-Review） using Large Language Models （LLMs） was proposed. First， a dynamic rule generation method was designed， domain knowledge and contract context were integrated， and LLMs were used to generate natural language and symbolic review rules in real time. Then， a contract review mechanism based on neuro-symbolic reasoning was constructed to extract local facts automatically， rules were localized through symbolic matching， and explainable reasoning was completed by combining symbolic reduction with LLM semantic execution. Finally， reasoning outputs from multiple segments were aggregated， and a dual?track review report， comprising both structured and natural?language content， was generated. Experimental results demonstrate that DRNS-Review improves the risk identification F1 value by 13.0 and 8.3 percentage points， respectively， and improves the Risk Level Accuracy （RLA） by 13.3 and 8.7 percentage points， respectively， compared to baseline methods LLM-Direct and ManualRule-LLM. It can be seen that integrating dynamic rule generation with neuro-symbolic reasoning improves the accuracy and adaptability of contract review effectively， providing a generalizable technical pathway for intelligent compliance applications.

Key words: contract review, Large Language Model (LLM), dynamic rule generation, neuro-symbolic reasoning, intelligent compliance

中图分类号: 

版权所有 © 2021 四川计算机应用杂志社有限公司
蜀ICP备 05010208 号-3　新出网证(川)字026号
技术支持：北京玛格泰克科技发展有限公司
访问总数：
今日访问：