摘要
知识图谱实体对齐的目的是找到两个或两个以上知识图谱中指向现实世界中同一对象的过程。目前的传统实体对齐方法和基于表示学习的实体对齐方法主要关注实体本身或者关系信息,然而,知识图谱中的属性信息在构建知识图谱的过程中也至关重要,能够提高实体对齐的准确率。为此,文中提出一种融合高速路门机制的联合实体关系和属性信息的实体对齐模型,利用带有高速路门机制(Highway Gates)图卷积的表示方法学习关系三元组和属性信息的嵌入表示。在大型跨语言数据集上进行实验,结果表明:相比于MTransE、IPTransE、JAPE、GCN-Align和AlignEA等方法,所提方法在Hits@1上分别高出1.29%、4.7%、1.88%、1.35%、0.66%;在Hits@10上分别高出2.49%、3.81%、2.28%、1.39%、2.23%。
Knowledge graph entity alignment is the process of finding two or more knowledge graphs that point to the same object in the real world.The current traditional entity alignment methods and representation learning based entity alignment methods mainly focus on the entity itself or relationship information.However,the attribute information in the knowledge graph is also crucial in the process of constructing the knowledge graph,which can improve the accuracy of entity alignment.An entity alignment model integrating highway gate mechanism and joint entity relationships and attribute information is proposed,and the convolutional representation method with highway gates is used to learn the embedding representation of relationship triplets and attribute information.Experiments were conducted on large cross language datasets,and the results show that in comparsion with methods such as MTransE,IPTransE,JAPE,GCN⁃Align,and AlignEA,the proposed method are 1.29%,4.7%,1.88%,1.35%and 0.66%higher in Hits@1,and 2.49%,3.81%,2.28%,1.39%and 2.23%higher in Hits@10,respectively.
作者
时慧芳
SHI Huifang(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650000,China)
出处
《现代电子技术》
2023年第20期167-172,共6页
Modern Electronics Technique
关键词
知识图谱
实体对齐
跨语言
实体关系
属性信息
高速路门机制
图卷积网络
knowledge graph
entity alignment
cross⁃language
entity relationships
attribute information
highway gate mechanism
graph convolutional network
