GLSEC: Global and local semantic-enhanced contrastive framework for knowledge graph completion

Ruixin Ma, Xiaoru Wang, Cunxi Cao, Xiya Bu, Hao Wu, Liang Zhao

Abstract: Highlights•We consider both global attribute and local topology information of entities.•We construct a global attribute interaction graph based on entity attribute pairs.•We design a light graph aggregation network for efficient vector updates.•We develop a novel global-local cross-view contrastive learning method.