METransE: Manifold-like mechanism enhanced embedding for reasoning over knowledge graphs

作者：

Highlights：

• A type-aware reasoning framework over KGs is proposed.

• Embedding model learns KG in semantic-view and entity-type-view jointly.

• Manifold-like mechanism is proposed to effectively embed type constraints.

• Type acquisition is proposed to overcome type-aware models’ generality defects.

• The embedding space is equipped with better type-related interpretability.

摘要

•A type-aware reasoning framework over KGs is proposed.•Embedding model learns KG in semantic-view and entity-type-view jointly.•Manifold-like mechanism is proposed to effectively embed type constraints.•Type acquisition is proposed to overcome type-aware models’ generality defects.•The embedding space is equipped with better type-related interpretability.

论文关键词：Knowledge graph embedding,Knowledge graph reasoning,Type-constrained embedding,Manifold-like mechanism

论文评审过程：Received 2 April 2022, Revised 28 June 2022, Accepted 24 July 2022, Available online 30 July 2022, Version of Record 5 August 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118288