Knowledge-Aware Self-supervised Educational Resources Recommendation

In the evolving landscape of educational technology, personalized recommendation systems play a pivotal role in delivering tailored educational content to learners. The knowledge-aware recommendation has emerged as a new trend in this field. However, knowledge graphs often suffer from sparsity and noise, characterized by long-tail entity distributions. We propose a novel framework, Knowledge-Aware Self-Supervised Educational Resources Recommendation (KASERRec), which leverages an integrated approach using knowledge graph embeddings, Light Graph Convolution Network(LightGCN), and cross-view knowledge contrastive learning to address the challenges inherent in educational resources recommendation. Focused on enhancing the discoverability of long-tail educational content, which often contains valuable but overlooked knowledge points, KASERRec introduces innovative modules designed to optimize recommendation accuracy and diversity. Employing a real-world educational knowledge graph, the framework demonstrates significant improvements in personalized learning experiences by effectively recommending educational resources that cater to the specific needs and preferences of learners. Our evaluations on the MOOCCube dataset highlight the framework's superiority over existing methods in terms of recommendation relevance and user satisfaction.