Study On Embedding-based Entity Alignment For Knowledge Graphs

In recent years,with the development of various knowledge graph technologies,the scale of knowledge graphs has been expanding,and various fields have begun to build new knowledge graphs according to their own needs.Today,knowledge graphs are widely used for tasks such as retrieval,question answering,and reasoning,supporting applications in many industries.However,a single knowledge graph can no longer meet the actual application needs,and there are problems such as information redundancy and heterogeneity among various knowledge graphs,so knowledge fusion has become a topic of concern.Entity alignment is an important part of knowledge graph fusion,and the purpose is to find nodes between different knowledge graphs that point to the same entity in the real world.With the in-depth research of knowledge graph embedding methods,the entity alignment model based on knowledge graph embedding has received extensive attention.However,these models still have imperfections: most models only consider the influence of the structural information of the knowledge graph on the effect of entity alignment,and some models use attribute information to improve the alignment effect,while entity names are often ignored,resulting in existing models.The use of information on the knowledge graph is not comprehensive enough;for models that use structural information,most models currently do not consider the distant neighbor entities of the central entity,and usually model the first-order neighbor structure of the central entity,resulting in the embedded representation of the central entity.Information is limited.In response to these problems,this paper designs an entity alignment model based on embedding,which improves the effect of entity alignment by utilizing the semantic information and structural information on the knowledge graph.The model in this paper represents the entity name,attribute information and entity category of the knowledge graph through a large-scale pretrained language model,and fully utilizes the semantic information on the knowledge graph.In addition,the model in this paper uses the graph attention network to model the one-hop neighbors and two-hop neighbors of the entities respectively,which realizes the utilization of more distant entity information,so that the model can capture more complex neighbor structures.In this paper,the method of feature linear modulation is used to effectively combine word embedding and structure embedding.The experimental results on different datasets show that the model in this paper is superior to the existing methods in the task of entity alignment,and also reflects the strong robustness of the model.At the same time,this paper sets up ablation experiments to verify the effectiveness of each module.In practice,this paper performs entity alignment of three published biomedical knowledge graphs and fuses them into a larger-scale knowledge graph.By developing a knowledge graph application platform,the knowledge graph is provided to other researchers for use.