over-generalization

This paper proposes PogRE, a method to address pattern over-generalization in knowledge graph embedding by using dense linear transformations, improving link prediction performance on standard benchmarks.