知识图谱构建研究综述  

Research Review of Knowledge Graph Construction

作　　者：杨延云 胡军 YANG Yanyun;HU Jun(Jiangxi Open University,Nanchang 330046,China;Jiangxi Flight University,Nanchang 330088,China)

机构地区：[1]江西开放大学,江西南昌330046 [2]江西飞行学院,江西南昌330088

出　　处：《现代信息科技》2025年第8期117-125,131,共10页Modern Information Technology

基　　金：江西省教育厅科技项目(GJJ215702);江西飞行学院校级系列科学研究项目(2024YB01)。

摘　　要：知识图谱作为结构化的语义知识库,在信息检索、智能问答和推荐系统等多个领域发挥着关键作用。文章综述了知识图谱构建的三个核心环节:信息抽取、知识融合和知识推理。信息抽取技术从基于规则的方法发展到机器学习模型,再到深度学习模型,目前正朝着减少误差传播和提高准确性的实体关系联合抽取模型方向演进。在知识融合部分,探讨了实体链接和知识合并策略,以及通过实体消歧和实体对齐解决实体识别问题。知识推理部分则分析了基于规则、表示学习和深度学习的推理方法,及其在新知识发现和错误信息纠正中的应用。最后,指出了构建过程中的挑战,并对未来研究方向提出了建议,旨在促进知识图谱研究和应用的发展。As a structured semantic knowledge base,the Knowledge Graph plays a key role in many fields such as information retrieval,intelligent question answering,and recommendation systems.This paper reviews the three core components of Knowledge Graph construction,information extraction,knowledge fusion,and knowledge reasoning.Information extraction technology has developed from rule-based methods to Machine Learning model,and then to Deep Learning model.It is currently evolving towards a joint Entity Relationship Extraction model that reduces error propagation and improves accuracy.In the part of knowledge fusion,the strategies of entity linking and knowledge merging are discussed,and the problem of entity recognition is solved by entity disambiguation and entity alignment.The section on knowledge reasoning analyzes the reasoning methods based on rules,representation learning and Deep Learning,and its application in new knowledge discovery and error information correction.Finally,the challenges in the construction process are pointed out,and suggestions for future research directions are proposed to promote the development of knowledge graph research and application.

关 键 词：知识图谱 信息抽取 知识融合 知识推理 深度学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]