摘要
现有的基于图卷积的实体对齐算法大多基于实体之间的关系结构构建,没有有效利用实体的属性结构信息,为此提出一种结合实体属性结构信息的图卷积实体对齐方法。该方法在实体以属性连接起来的知识图上进行卷积,学习实体基于属性结构的嵌入,再结合实体基于关系结构的嵌入来比较实体的相似性。在真实数据集上的实验结果表明提出的方法优于基准方法,从而为实体对齐提供了一种新的可能。
Most of the existing GCN-based methods are based on the relationship structure between entities and don’t utilize the attribute structure information effectively.This paper proposed a GCN-based method combined with attribute structure.This method convolved entities on the knowledge graph connected by attributes to learn the embedding of entities based on attribute structure,and calculated the similarity between entities combining with the embedding of entities based on relationship structure.Experimental results on real-world datasets show that the proposed method is better than baselines,thus providing a new possible method for entity alignment.
作者
田江伟
李俊锋
柳青
Tian Jiangwei;Li Junfeng;Liu Qing(School of Software,Yunnan University,Kunming 650504,China)
出处
《计算机应用研究》
CSCD
北大核心
2021年第7期1979-1982,1992,共5页
Application Research of Computers
基金
国家自然科学基金资助项目(61866139)。
关键词
实体对齐
知识库
图卷积
属性结构
表示学习
entity alignment
knowledge base
GCN
attribute structure
representation learning
