基于图卷积神经网络的石油数据资产知识图谱实体对齐方法

Entity alignment method of petroleum data assets knowledge graph based on GCN

在石油数据资产知识图谱融合过程中存在命名规则差异性大、专业性强和特殊语义问题。提出基于图卷积神经网络的石油数据资产知识图谱(KG)实体对齐方法,给定一组预先对齐的实体种子,采用Graph Convolutional Networks(GCNS)网络学习实体结构和属性信息嵌入统一向量空间,计算空间中实体之间距离;在石油数据资产数据集中对两个KGs进行实体对齐实验。结果表明:基于GCN融合实体关系和属性的嵌入模型优于基于实体关系的TransE实体对齐模型,Hits@1最高为16.96%,比TransE实体对齐模型平均提升6.18%。基于图卷积神经网络的融合实体关系、属性和属性值的实体对齐方法适用于石油数据资产知识管理。

To deal with the problems of large differences in naming rules,strong professionalism and special semantics in the process of knowledge graph fusion of data assets in the oil field,we proposed a novel approach for KG alignment of data assets in the oil field based on graph convolutional networks(GCNs).Given a set of pre-aligned entity seeds,the GCN network is used to learn the entity structure and attribute information embedded in the unified vector space,and the distance between entities in the space is calculated to obtain the alignment result.The entity alignment experiment is carried out on two knowledge graphs in the petroleum field.The experimental results show that the GCN-based model is superior to the TransE entity alignment model based on the relationship among entities,with a Hits@1 value of 16.96%,which is 6.18%higher than TransE.And the method of entity alignment fusing attributes and attribute values based on graph convolutional neural network is suitable for petroleum data asset knowledge management.