基于司法判决书的知识图谱构建与知识服务应用分析

Construction of Knowledge Graph and Analysis of Knowledge Service Application Based On Judicial Decision Documents

【目的/意义】构建基于司法判决书的案件知识图谱是对司法数字资源的有效利用,有助于提升司法智能化水平,积极响应国家"智慧法院"建设发展战略。【方法/过程】以"网络诈骗"领域为例,用"自顶向下"的方式构建知识图谱。首先,结合文书内容与专家意见构建案件领域本体;接着,通过知识抽取、知识表示、知识融合等环节获取实体、属性及关系;再利用Neo4j生成案件知识图谱。最后,提出了基于知识图谱的智慧司法知识服务框架。【结果/结论】基于2015年-2020年的"网络诈骗"领域司法判决书,构建了含有约3万个实体和18万条关系的案件知识图谱,并详细阐述了具备基础资源层、知识图谱层、服务应用层的智慧司法知识服务框架设计。【创新/局限】实现了案件知识图谱的实体类型扩充,以丰富图谱应用场景,并将知识图谱技术与智慧司法知识服务框架进行融合;局限在于仅使用网络诈骗领域判决书数据进行实证研究。

【Purpose/significance】The construction of case knowledge graph based on judicial decision documents is an effective use of judicial digital resources,which helps to improve the level of judicial intelligence and actively respond to the national "smart court" construction and development strategy.【Method/process】Take the field of cyber fraud as an example to build the knowledge graph in a top-down way.Firstly,the case domain ontology is constructed by combining the contents of documents and expert opinions.Then,entities, attributes and relationships are acquired through knowledge extraction. knowledge representation and knowledge fusion. Then Neo4 j was used to generate knowledge graph of crime domain.Finally,an intelligent judicial knowledge service framework in crime domain based on knowledge graph is proposed.【Result/conclusion】Based on the judicial decision documents in the field of cyber fraud from 2015 to 2020,a knowledge graph containing about 30,000 entities and 180,000 relationships is constructed.and the framework design of intelligent judicial knowledge service with basic resource layer,knowledge graph layer and service application layer is elaborated in detail.【Innovation/limitation】The entity types of the case knowledge graph were expanded to enrich the application scenarios of the case knowledge graph,and the knowledge graph technology was integrated with the intelligent judicial knowledge service framework.The limitation lies in the empirical research only using the data of judgment in the field of network fraud.