Type-Based Neighborhood Aggregation for Knowledge Graph Alignment

Yaqi Cheng,Bohan Li,Zhengming Qian,Hailian Yin

DOI: https://doi.org/10.1007/978-981-97-7235-3_22

2024-01-01

Abstract:Entity alignment (EA) is a critical step in integrating multiple knowledge graphs (KGs), which aims at identifying entities referring to the same real-world object from different KGs. The majority of existing embedding-based methods usually enhance the contextual representation of entities by performing neighborhood aggregation to support more accurate entity alignment. However, the vanilla aggregation process simply considers each neighbor to be equally important without distinguishing the difference between neighbors, which is much limited. To address the problem, we propose TNAEA, a Type-based Neighborhood Aggregation for Entity Alignment. The core idea of TNAEA is to introduce a selective neighborhood aggregation mechanism to enhance entity representation by classifying neighbors from different angles. Specifically, we classify each neighbor node for each central entity from three different perspectives, namely semantic similarity, spatiality and topological structure, and add corresponding type information to distinguish each other. Then, we incorporate the type information into the neighborhood aggregation process and hence better represent each central entity. Extensive experiments are conducted to study TNAEA across three real-world cross-lingual datasets, revealing that our method outperforms existing methods in performance.