GEnI: A framework for the generation of explanations and insights of knowledge graph embedding predictions

Elvira Amador-Domínguez, Emilio Serrano, Daniel Manrique

Abstract: Highlights•An overview on the interaction proposals between rule-learning and KGE models is presented.•Most of the studied proposals do not consider the use of rule-learning approaches to explain KGE predictions.•GEnI, a novel three-phase framework for the explanation of KGE predictions, is presented.•GEnI supports bilinear and translational models and has been successfully tested on different KGE models and datasets.•GEnI is evaluated from three perspectives: coherence, meaningfulness, and reliability.