一种联合指导嵌入的实体对齐方法

Joint Guidance Embedding Method for Entity Alignment

实体对齐是知识图谱完善过程中必不可少的环节,当前实体对齐工作中利用了许多实体属性的信息,忽略了实体和本体之间的联系.针对实体对齐问题中缺少本体概念知识信息的问题,提出一种由实体属性和本体自下而上共同指导及联合嵌入的实体对齐方法.该方法首先对本体信息、实体信息、实体属性信息进行嵌入,之后通过本体类别信息和种子词进行本体类冲突构建、再经过本体和实体属性进行嵌入指导,最终将得到的实体对加入到种子词中并通过全连接网络进行迭代的训练.在公共数据集上进行了大量实验,使用Hits@1,Hits@5,MRR值作为模型的评价指标,并通过对比实验和消融实验验证了该模型的有效性.

Entity alignment is an essential part in the process of knowledge graphs improvement.In the current entity alignment work,a lot of entity attribute information is used,but the relationship between entity and ontology is ignored.In order to solve the problem of lack of ontology conceptual knowledge information in entity alignment,this paper proposes a bottom-up entity alignment method based on guidance and joint embedding by entity attributes and ontology.The method first embeds ontology information,entity information,and entity attribute information,then constructs ontology class conflict through ontology category information and seed words,and then embeds the entity pair into seed words through ontology and entity attribute guidance,and finally adds the entity pair to seed words and conducts iterative training through the fully connected network.Experiments were carried out on the public data set using Hits@1,Hits@5,MRR value was used as the evaluation index of the model,and the validity of the model was verified through comparative experiments and ablation experiments.