基于统计的知识推理研究进展  

Survey of Knowledge Reasoning Based on Statistic

作　　者：赵国清 何佳洲[1] 李永盛 王景石 ZHAO Guo-qing;HE Jia-zhou;LI Yong-sheng;WANG Jing-shi(Jiangsu Automation Research Institute, Lianyungang 222061, China)

机构地区：[1]江苏自动化研究所,江苏连云港222061

出　　处：《指挥控制与仿真》2020年第4期8-12,共5页Command Control & Simulation

摘　　要：知识推理作为知识图谱构建的关键技术之一,旨在补全知识图谱,纠正知识图谱中的错误关系。利用知识图谱遵循的统计规律,结合机器学习方法的知识推理技术,在对大型知识图谱的补全方面有着巨大的优势。针对基于统计的知识推理技术做了较为系统性的整理,首先阐释了面向知识图谱的知识推理技术,然后从基于隐特征的关系学习、基于图特征的关系学习和基于实体分布规律的类型推理三个方面对已有的知识推理技术进行了介绍,最后总结并展望了知识推理方法未来的发展。As a key technology of Knowledge Graph(KG)construction,Knowledge Reasoning Technology aims to make up KG and correct wrong relationships in KG.Knowledge reasoning based on statistics has great advantages in the completion of the large knowledge graph,while can utilize the statistical laws followed by KG.This paper presents a systematic survey to knowledge reasoning technology based on statistics.It briefly introduces the concept of knowledge reasoning for KG.Then,it introduces existing knowledge reasoning technology from three aspects:relationship learning based on hidden features,relationship learning based on graph features and type reasoning based on entity distribution law.This paper also discusses the challenges and the future research direction of knowledge reasoning based on statistics.

关 键 词：知识图谱 知识推理 统计关系学习 知识图谱补全 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]