Entity Alignment Across Knowledge Graphs Based on Representative Relations Selection

摘要

Entity alignment across knowledge graphs is an important task in web mining. The aligned entities can be used for transferring knowledge across knowledge graphs and benefit several tasks such as cross-lingual knowledge graph construction and knowledge reasoning. This paper propose a representation learning based algorithm for embedding knowledge graph and aligning entities. In particular, considering the multi-type relations in knowledge graph, we select the alignment-task driven representative relations based on the pre-aligned entity pairs. With the help of selected relations, we embed the entities across networks into a common space by modeling entities' head/tail are with corresponding context vectors. For entity alignment task, pre-aligned entities are adopted to facilitate the transfer of context information across the knowledges graphs. Through this way, the problem of entity embedding and alignment can be solved simultaneously under a unified framework. Extensive experiments on two multi-lingual knowledge graphs demonstrate the effectiveness of the proposed model comparing with several state-of-the-art models.