KG-XQRec: Knowledge Graph-based Explainable Question Recommendation

In order to cope with the challenges of information overload and loss of learning direction caused by the lack of interpretation mechanisms in the online education field recommendation system, this paper proposes a knowledge-based interpretable question recommendation (KG-XQRec) method. This method represents the sequence of answers into a directed graph structure. The purpose is to extract learning behavior trajectories from the sequence, and then carefully build a knowledge graph to represent the learner's knowledge network and model its cognitive state. Finally, by comprehensively considering the three types of characteristics of behavior, knowledge-based and cognitive, we recommend target questions to learners, and clarify the recommendation results through interpretation at three different levels: behavioral path, knowledge path and cognitive path. Experimental evaluation on the MoocRadar dataset confirmed that the model improves on all evaluation measures and provides richer explanations than the baseline method.