结合静态事实和重复历史事实的动态知识图谱推理方法  

Dynamic knowledge graph inference method combining static facts and repeated historical facts

作　　者：林栋 李永强[1] 仇翔[1] 冯远静[1] 谢碧峰 LIN Dong;LI Yong-qiang;QIU Xiang;FENG Yuan-jing;XIE Bi-feng(School of Information Engineering,Zhejiang University of Technology,Hangzhou 310012,China;Hangzhou Fengjing Technology Company,Hangzhou 310000,China)

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310012 [2]杭州峰景科技有限公司,浙江杭州310000

出　　处：《浙江大学学报（工学版）》2023年第10期1915-1922,共8页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(62073294);浙江省自然科学基金资助项目(LZ21F030003).

摘　　要：针对现有的动态知识图谱推理方法容易忽略动态知识图谱中存在着大量静态信息和重复历史事实的问题,提出结合静态事实和重复历史事实的动态知识图谱网络方法.该方法利用动态知识图谱中实体间隐藏的静态联系来构成静态事实,并协助动态知识图谱推理;利用历史事实构建历史词表,在预测未来时对历史词表进行查询;对历史中未发生的事实进行惩罚,提高重复历史事实的预测概率.在2个公开的数据集上进行动态知识图谱推理实验,对比实验时选用目前主流的5个模型作为基线.在实体预测实验中,平均倒数排名(MRR)达到0.4891和0.5303,Hits@10达到0.5887和0.6165,证明了所提方法的有效性.A static-historical network(Sta-HisNet)method combining static facts and repeating historical facts was proposed,aiming at the problem that existing dynamic knowledge graph reasoning methods tend to overlook the vast amount of static information and repeating historical facts present in the dynamic knowledge graphs.The hidden static connections between entities in the dynamic knowledge graph were used to form static facts,assisting in the inference of the dynamic knowledge graph.Historical facts were employed to construct a historical vocabulary,and the historical vocabulary was queried when predicting the future.Facts that had not occurred in history were punished,and the probability of predicting duplicate historical facts was increased.Experiments were conducted on two public datasets for dynamic knowledge graph reasoning.Comparative experiments were performed using five mainstream models as baselines.In entity prediction experiments,the mean reciprocal rank(MRR)was 0.4891 and 0.5303,and Hits@10 reached 0.5887 and 0.6165 respectively,demonstrating the effectiveness of the proposed method.

关 键 词：动态知识图谱 静态事实 重复历史事实 历史惩罚 实体预测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]