人工智能产业政策知识图谱构建研究  

Research on the Knowledge Graph Construction of Industrial Policies for Artificial Intelligence

作　　者：赵晋世 沈永珞 ZHAO Jinshi;SHEN Yonguo(School of Information,Guangdong University of Finance and Economics,Guangzhou 510320,China)

机构地区：[1]广东财经大学信息学院,广东广州510320

出　　处：《现代信息科技》2024年第23期93-99,共7页Modern Information Technology

基　　金：教育部人文社科研究项目(22YJAZH091)。

摘　　要：随着人工智能产业的高速发展,国家政策层面出台了大量的人工智能产业政策,对此,文章提出了一种基于语义分析的人工智能政策知识图谱构建方法。首先,根据语义角色标注和依存句法关系提取实体关系三元组;然后运用ChatGLM-6B大语言模型识别政策中的政策客体,用于知识筛选;再计算关系词汇的余弦相似度对关系进行融合对齐,实现知识融合;最后,对实体关系查询和政策信息服务两个应用领域进行了探究。所构建的知识图谱可以精准地向企业等用户提供所需的政策信息,提高政策信息的利用效率。With the rapid development of the Artificial Intelligence industry,a large number of industrial policies for Artificial Intelligence have been introduced at the national policy level.Therefore,this paper proposes a Knowledge Graph construction method of Artificial Intelligence policies based on semantic analysis.Firstly,it extracts entity relationship triples according to semantic role labeling and dependency syntactic relationship.Secondly,it uses the ChatGLM-6B Large Language Model to identify the policy objects in the policies for knowledge screening.Thirdly,it calculates the cosine similarity of the relationship vocabulary to fuse and align the relationship to achieve knowledge fusion.Finally,the two application fields of entity relationship query and policy information service are explored.The constructed Knowledge Graph can accurately provide the required policy information to enterprises and other users and improve the utilization efficiency of policy information.

关 键 词：产业政策 知识图谱 知识抽取 大语言模型 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]