KGEL: A novel end-to-end embedding learning framework for knowledge graph completion

作者：

Highlights：

• Embedding models perform the link prediction task on the basis of individual triples.

• Gathering neighborhood information leads to richer node embeddings.

• A GCN based model is used to produce two embedding vectors of each node.

• A novel tensor factorization model predicts missing entities to derive new triples.

• The results prove that richer node embeddings lead to significant performance gain.

摘要

•Embedding models perform the link prediction task on the basis of individual triples.•Gathering neighborhood information leads to richer node embeddings.•A GCN based model is used to produce two embedding vectors of each node.•A novel tensor factorization model predicts missing entities to derive new triples.•The results prove that richer node embeddings lead to significant performance gain.

论文关键词：Knowledge graph,Link prediction,Weighted graph convolutional network,Tensor train decomposition,Tensor factorization

论文评审过程：Received 2 June 2020, Revised 25 August 2020, Accepted 24 October 2020, Available online 27 October 2020, Version of Record 10 February 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.114164