A Smart MOOC Learning Assistant

Semantic Course Hybrid Recommendations Based on Multi-Platform Data Integration

Abstract

With the rapid expansion of MOOCs, learners face increasing challenges due to information overload and data fragmentation across multiple platforms. Traditional recommendation systems rely on keyword matching, failing to capture complex semantic relations between courses, skills, and learner goals. This paper details the development and use of a Smart MOOC Learning Assistant, a hybrid semantic recommender system developed with data coming from four different Learning Management Systems: Coursera.com, Udemy.com, Edx.org and the Kurdish Education Platform, comprising over 5000 MOOCs. The OWL ontology (designed using Protégé) serves as the foundation for the smart assistant and defines how all entities within the domain relate to each other. A combination of the Flask framework and the RDFLib library provides the architecture for executing SPARQL queries to provide context-sensitive results. To provide for more flexibility in matching natural language query variations, Levenshtein-based fuzzy matching has been used. The Smart MOOC Learning Assistant also implements a hybrid approach to recommendation, by incorporating a collaborative feedback loop (liking/disliking) which refines the ranking mechanism and removes the "cold-start" effect. The system architecture is characterized by a clear distinction between retrieval (filtering relevant courses via SPARQL), recommendation (semantic planning based on educational relationships), and ranking (ranking the results using hybrid social-semantic scores).Evaluation of the Smart MOOC Learning Assistant shows outstanding results with accuracy at 0.984, precision at 0.984, recall at 1.0, and F1-score at 0.99. This demonstrates that combining semantic-driven approaches with datasets from multiple sources can significantly reduce retrieval noise by providing highly relevant personalized digital learning.