基于概率推理的知识图谱链接预测方法  

Approach for link prediction of knowledge graph based on probabilistic inferences

作　　者：姚建军 李剑宇 岳昆[1,2] 段亮[1,2] 付晓东[3] YAO Jianjun;LI Jianyu;YUE Kun;DUAN Liang;FU Xiaodong(School of Information Science and Engineering,Yunnan University,Kunming 650500,China;Key Lab of Intelligent System and Application of Yunnan Province,Yunnan University,Kunming 650500,China;Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650504,China)

机构地区：[1]云南大学信息学院,云南昆明650500 [2]云南大学云南省智能系统与计算重点实验室,云南昆明650500 [3]昆明理工大学信息工程与自动化学院,云南昆明650504

出　　处：《计算机集成制造系统》2023年第10期3483-3495,共13页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(62002311);云南省重点实验室建设资助项目(202205AG07003);云南省重大科技专项计划资助项目(202202AD080001);云南省基础研究资助项目(202201AT070394);云南大学“东陆学者”培育计划资助项目。

摘　　要：为了有效发现实体间隐含的关联关系并对其进行量化,以全面准确地进行知识图谱(KG)链接预测,提出基于概率推理KG链接预测方法。该方法以描述实体间隐含关联关系并度量链接存在的可能性为目标,基于AMIE算法挖掘KG中的规则并将其转换为Horn子句,进一步构建描述不同实体依赖关系的规则链接贝叶斯网(RLBN),将KG的链接预测任务转换为RLBN的概率推理任务来计算实体间的关联度,从而预测实体间的链接关系。实验结果表明,基于RLBN的KG链接预测精确率和召回率优于其他方法,验证了所提模型的有效性与高效性。The knowledge in Knowledge Graph(KG)is incomplete and there are missing links between some entities.To effectively discover the implicit association between entities and quantify them for fulfilling KG link prediction comprehensively and accurately,a prediction method based on probabilistic inference was proposed,which aimed to describe the implicit association relationships between entities and measure the possibility of links.Based on AMIE algorithm,rules in KG were obtained and transformed into Horn clauses to further build a Rule-Linked Bayesian Network(RLBN)describing different entity dependencies.Link prediction of KG was transformed into the probabilistic inference over the RLBN to calculate the correlation degree between entities,so as to predict the link relationship between entities.Experimental results showed that the accuracy and recall of the RLBN based link prediction results was better than other competitors,which verified the effectiveness and efficiency of the proposed model.

关 键 词：知识图谱 链接预测 贝叶斯网 HORN子句 概率推理 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]