ProPRL Improves Prerequisite Relation Learning in EdTech

A novel framework named ProPRL (Property-aware Prerequisite Relation Learning) has been developed to enhance the accuracy of identifying prerequisite relationships within educational knowledge graphs. Traditional methods often treat this as a generic link prediction problem, which limits their ability to integrate diverse educational evidence and prevent contradictory reverse predictions. ProPRL overcomes these limitations by first learning complementary concept representations from both concept-resource hypergraphs and directed learning-behavior graphs, employing personalized propagation to aggregate multi-hop behavioral evidence. The framework then utilizes a Pair-conditioned Gate to adaptively weight and fuse these two distinct views for each potential concept pair. A crucial innovation is the introduction of an Irreversibility Constraint, an anti-symmetry regularizer that penalizes scenarios where high confidence is simultaneously assigned to both directions of a prerequisite relationship. Experiments on several real-world educational datasets confirm that ProPRL achieves state-of-the-art performance in prerequisite relation learning.