基于BERT与要素提取的相似案例匹配  

Similar case matching based on BERT and feature extraction

作　　者：焦宇超 阎刚 JIAO Yuchao;YAN Gang(School of Artificial Intelligence,Hebei University of Technology,Tianjin 300401,China)

机构地区：[1]河北工业大学人工智能与数据科学学院,天津300401

出　　处：《智能计算机与应用》2025年第1期130-135,共6页Intelligent Computer and Applications

摘　　要：相似法律案件检索是一项特殊的检索任务,对于给定的查询案例,需要从给定的候选案例中搜索相似的案例。与传统的文本匹配不同,法律案件匹配具有文本较长、主题性强的特点。针对上述问题,本文提出了一种基于案件要素的相似案例检索方法。首先对BERT模型使用通用语料进行微调;然后采用段落聚合方法,对案件文书上下文语义信息进行编码,同时将法律文书数据融入模型。本文在LeCaRD数据集上进行了广泛的实验,实验结果表明,本文提出的模型优于现有模型。Similar legal case retrieval is a special retrieval task in which similar cases need to be searched from given candidate cases for a given query case.Unlike traditional text matching,legal case matching has the characteristics of long text and strong subjectivity.To address these issues in similar case matching in legal cases,this thesis proposes a similar case retrieval method based on case elements.This thesis first uses general corpora to fine-tune the BERT model,then encodes the context-specific semantic information of case documents using the paragraph aggregation method,and integrates legal document data into the model.Extensive experiments on the LeCaRD dataset were conducted in this paper,and the results show that the proposed model is superior to existing models.

关 键 词：相似案例匹配 BERT 长文本 法律要素 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]