Concept-driven representation learning model for knowledge graph completion

Yan Xiang, Hongguang He, Zhengtao Yu, Yuxin Huang

Abstract: Highlights•We propose a Concept-driven Representation Learning Model (CDRL) for knowledge graph completion, which is the first attempt to utilize the concept entity sets to drive the model to obtain better entity representations.•CDRL employs homo- and hetero-concept entity sets for contrastive learning, optimizing representations by utilizing intra-and inter-correlations within and between homo-concept sets.•Extensive experiments on three benchmark datasets demonstrate that CDRL significantly outperforms other competitive methods, with in-depth analysis validating the effectiveness of its various modules.