基于词性和语序分析的法律知识图谱自动构建方法  

Automatic Construction Method of Legal Knowledge Graph Based on Part of Speech and Word Order Analysis

作　　者：唐明伟[1] 赵煌涛 李文雯 TANG Mingwei;ZHAO Huangtao;LI Wenwen(School of Computer Science,Nanjing Audit University,Nanjing 211815,China)

机构地区：[1]南京审计大学计算机学院,江苏南京211815

出　　处：《现代信息科技》2024年第22期85-91,共7页Modern Information Technology

基　　金：教育部人文社会科学研究规划基金项目(23YJA870009);江苏省高校哲学社会科学研究重大项目(2021SJZDA153);江苏高校“青蓝工程”项目。

摘　　要：文章挖掘法律文本中的实体和关系,构建法律知识图谱,为实现智能判案提供支持,完善法律知识图谱的构建方法。应用基于LexNLP的自然语言处理方法,分析法律文本,以句子为单位进行词性分析,标注出名词且为主语或宾语时作为实体,动词且为谓语时标注为关系。在这一基础上,将同一个句子中的实体和关系按照<实体1,关系,实体2>进行排列组合,生成不重复的知识三元组,以生成高质量的法律知识图谱。提出了一种基于词性和语序分析的法律知识图谱自动构建方法,并基于美国Caselaw Access Project项目所含的法律判例为原始数据,并对生成三元组进行质量评估,最后生成了关于法律的知识图谱。The paper constructed a legal Knowledge Graph through mining the entities and their relationships in the legal text,aiming to facilitate intelligent judgment and enhance the methodology of legal Knowledge Graph construction.This paper employs Natural Language Processing techniques rooted in LexNLP,analyzes the legal texts,and conducts sentence-level partof-speech analysis,wherein nouns functioning as subjects or objects are labeled as entities,while verbs serving as predicates are labeled as relationships.Based on this framework,the entities and relationships within each sentence are permuted and combined according to the format<entity 1,relationship,entity 2>,resulting in the generation of non-repetitive knowledge triplets so as to generate a high-caliber legal knowledge graph.The paper proposes an automated construction approach for the legal Knowledge Graph based on part-of-speech and word order analyses,takes the legal precedents contained in the Caselaw Access Project in the US as the raw data,and assesses the quality of the generated triplets and presents a legal Knowledge Graph.

关 键 词：知识图谱构建 实体识别 关系抽取 自然语言处理 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]