一种采用贝叶斯推理的知识图谱补全方法  

Method for Knowledge Graph Completion Based on Bayesian Reasoning

作　　者：翟社平 郭琳 高山 段宏宇 李兆兆 马越 ZHAI She-ping;GUO Lin;GAO Shan;DUAN Hong-yu;LI Zhao-zhao;MA Yue(College of Computer Science and Technology, Xi＇an University of Posts and Telecommunications, Xi＇an 710121, China;College of Economics and Business Administration,Xi＇an University of Posts and Telecommunications,Xi＇an 710121 ,China})

机构地区：[1]西安邮电大学计算机学院,西安710121 [2]西安邮电大学经济与管理学院,西安710121

出　　处：《小型微型计算机系统》2018年第5期995-999,共5页Journal of Chinese Computer Systems

基　　金：工业和信息化部通信软科学项目(2017-R-22)资助;陕西省教育厅科学研究计划项目(12JK0733)资助;陕西省社会科学基金项目(2016N008)资助;西安市社会科学规划基金项目(17X63)资助;西安邮电大学研究生创(114-602080105)资助

摘　　要：为了适应Web数据信息高度动态化的发展趋势,针对知识图谱的补全更新实时性欠佳的问题,提出一种知识图谱补全的链接预测方法.该方法运用贝叶斯网络概率推理图模型优势并结合本体推理规则,对知识图谱节点间依赖程度进行定量分析,旨在充分挖掘模型潜在因素,实现正确预测.实验结果表明,该方法能有效提升知识图谱的链接预测效率,保证较高的预测准确度,并及时更新知识图谱.In order to adapt to the highly dynamic trend of Web data development,and to solve the problem of the real-time performance of knowledge graph updating,a method of knowledge graph link prediction is proposed. This method combined the advantages of Bayesian network probabilistic inference model and the ontology reasoning rules,the confidence among entities is analyzed quantitatively which is aiming at exploiting the potential factors of the model to attain the correct prediction. The results of analysis and experiment show that this method can improve the link prediction efficiency of knowledge graph,ensure the high prediction accuracy,and update the knowledge graph in time.

关 键 词：知识图谱补全 链接预测 贝叶斯推理 RDFS蕴含模式 图数据库 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]