基于司法案例知识图谱的类案推荐  

Case recommendation based on knowledge graph of judicial cases

作　　者：黄治纲 谢新强[1,2] 邢铁军 葛东[1] 蔡晨秋 窦丽莉 王天翊 Huang Zhigang;Xie Xinqiang;Ge Dong;Cai Chenqiu;Dou Lili;Wang Tianyi(Neusoft Corporation,Shenyang,110179,China;School of Computer Science and Engineering,Northeastern University,Shenyang,110169,China)

机构地区：[1]东软集团股份有限公司,沈阳110179 [2]计算机科学与工程学院,东北大学,沈阳110169

出　　处：《南京大学学报（自然科学版）》2021年第6期1053-1063,共11页Journal of Nanjing University（Natural Science）

基　　金：国家重点研发计划(2018YFC0831200)。

摘　　要：近年来,伴随着人工智能的发展及法院裁判文书的公开化,"智慧司法"、案例推荐成为热点问题.针对案例推荐中存在的推荐准确性差、传统知识图谱向量化表示精度不高等问题,提出基于知识图谱的案件推荐(Knowledge Graph based Case Recommendation,KGCR)模型.该模型以知识图谱为辅助信息,利用文本分类和信息抽取技术构建面向刑事案例的知识图谱,针对当事人的陈词供述,利用知识表示学习求解相似的案件,进一步实现法条推荐.针对TransH算法的负采样问题进行改进,提出FU-TransH算法模型.以公开的刑事判决书为数据集进行实验,实验结果表明,与相关的具有代表性的算法相比,该算法的推荐准确率更高.In recent years,with the development of artificial intelligence and the openness of court judgment documents,intelligent justice and case recommendation have become hot issues. Aiming at the problems of poor recommendation accuracy,low accuracy of vectorized representation of traditional knowledge graphs in case recommendation. This paper proposes Knowledge Graph based Case Recommendation(KGCR) model,which uses knowledge map as auxiliary information to solve similar cases. It uses text classification and information extraction technology to build a knowledge map for criminal cases,and uses knowledge representation learning to solve similar cases according to the confession of the parties.Aiming at the negative sampling problem of TransH algorithm,the Fu-TransH algorithm is proposed. The experimental results show that,compared with the representative algorithms,the proposed algorithm can improve the recommendation accuracy and recall rate.

关 键 词：知识图谱 信息抽取 表示学习 法条推荐 类案推荐 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]