Learning graph attention-aware knowledge graph embedding

The knowledge graph, which utilizes graph structure to represent multi-relational data, has been widely used in the reasoning and prediction tasks, attracting considerable research efforts recently. However, most existing works still concentrate on learning embeddings straightforwardly intuitively without subtly considering context of knowledge. Specifically, recent models deal with each single triple independently or consider contexts indiscriminately, is one-sided as unit (i.e., triple) can be derived from its partial surrounding triples. In this paper, we propose a graph-attention-based model encode entities, formulates an irregular explores number concrete interpretable compositions by integrating graph-structured information via multiple independent channels. To measure correlation between entities different angles entity pair, relation, structure), respectively develop three attention metrics. By making use our enhanced embeddings, further introduce several improved factorization functions for updating relation evaluating candidate We conduct extensive experiments downstream tasks including classification, typing, link validate methods. Empirical results importance introduced metrics demonstrate that proposed method improve performance large-scale graphs.