Entity Alignment of Knowledge Graph by Joint Graph Attention and Translation Representation

Entity alignment is a crucial and challenging research task in the fields of knowledge graph and natural language processing. It aims to integrate knowledge graph information from different languages or different sources and apply it to subsequent knowledge graph construction and other downstream tasks. Currently, most entity alignment solutions are based on knowledge graph embedding, which align entities by mapping entities into low dimensional spaces. But it also face problems such as fail to make full use of the neighbor nodes and relation information of the graph, and the heterogeneity between the knowledge graphs. This paper proposes a method of combining graph attention and translation models. We use graph attention mechanism to convey information about neighboring nodes, and merge relation information into the entity representation, make full use of the internal information of the knowledge graph to learn a better entity representation. Then we combine the graph attention mechanism with knowledge translation representation model to constrain the consistency between different knowledge graphs and guarantee the accuracy of embedding vectors in low-dimensional space. The results of comparative experiments with other entity alignment methods show that the performance of our method is better than others, and the alignment effect is significantly improved.