TAGAT: Type-Aware Graph Attention neTworks for reasoning over knowledge graphs
作者：
Highlights：
• A reasoning model over KG based on graph attention networks is proposed.
• The model jointly embeds semantic, neighborhood and entity type information.
• Multi-layer attention mechanism considering type is applied to learn neighborhood.
• Type is integrated into embedding as constraints conforming to human intuition.
• The embedding space has type-related interpretation by means of embedding types.
摘要
•A reasoning model over KG based on graph attention networks is proposed.•The model jointly embeds semantic, neighborhood and entity type information.•Multi-layer attention mechanism considering type is applied to learn neighborhood.•Type is integrated into embedding as constraints conforming to human intuition.•The embedding space has type-related interpretation by means of embedding types.
论文关键词：Knowledge reasoning,Knowledge graph embedding,Type-aware,Graph attention network,Hierarchical attention mechanism
论文评审过程：Received 13 May 2021, Revised 17 August 2021, Accepted 14 September 2021, Available online 28 September 2021, Version of Record 5 October 2021.
论文官网地址：https://doi.org/10.1016/j.knosys.2021.107500
