Profiling temporal learning interests with time-aware transformers and knowledge graph for online course recommendation

Abstract

Profiling users' temporal learning interests is key to online course recommendation. Previous studies mainly profile users' learning interests by aggregating their historical behaviors with simple fusing strategies, which fails to capture their temporal interest patterns underlying the sequential user behaviors. To fill the gap, we devise a recommender that incorporates time-aware Transformers and a knowledge graph to better capture users' temporal learning interests. First, we introduce stacked Transformers to extract users' temporal learning interests underlying users' course enrollment sequences. In addition, we design a time-aware positional encoding module to consider the enrollment time intervals between courses. Third, we incorporate a knowledge graph to utilize the latent knowledge connections between courses. The proposed method outperforms state-of-the-art baselines for course recommendation. Furthermore, findings in the ablation study offers several insights for future research. The proposed model can be implemented in online learning platforms to increase user engagement and reduce dropout rate.

English Education Tutoring Teaching System Based on MOOC

“…Students should choose resources for learning that are reasonable for them based on their own preferences and learning needs. This is more favorable to the realization of students' individualized learning and can boost students' learning interest and efficiency [ 14 – 18 ].…”

Supporting Curriculum Designers in Developing Balanced Outcome-based Programs using Knowledge Graphs

“…Knowledge graphs can create explicit and irrational knowledge relationships between courses, but a recommender system using knowledge graphs can better reflect and intelligently extend the semantic relationships between courses that motivate long-term learning interests. The proposed solution by Zhou et al [11] can be used in online platforms to recommend suitable courses based on learners' dynamic interests, which will increase user engagement and reduce dropout rates. Furthermore, the proposed model can be used in online platforms to increase user engagement and reduce dropout rates.…”

Learning Path Optimization based on Multi-Attribute Matching and Variable Length Continuous Representation

“…The sequence of learning materials is constrained by the relation between the materials. The relation can be captured by a knowledge model, such as knowledge map [13,14], knowledge graph [15,16], or concept map [17], etc. The model can either be built from the experts' experience [18] or via educational data mining technology [19].…”