Ontology-Based Recommender Systems for Online Learning Platforms: A Review

Abstract

The rapid growth of massive open online courses, learning management systems, and open educational repositories has produced an abundance of learning resources that overwhelms learners and complicates the selection of suitable content. Recommender systems have become a standard response to this information overload, yet the classical paradigms of collaborative filtering, content-based filtering, knowledge-based filtering, and their hybrids struggle with cold start, data sparsity, and a persistent semantic gap between raw interaction data and pedagogical meaning. This paper reviews how ontologies, formal and shareable specifications of a domain expressed in languages such as RDF and OWL, have been used to address these limitations in educational recommenders. It surveys the background of recommender paradigms and their weaknesses, explains what an ontology is and how domain, learner, and pedagogical ontologies encode knowledge, and analyses the mechanisms through which ontologies support semantic similarity, logical reasoning, learner modelling, prerequisite and zone of proximal development reasoning, and learning path sequencing. A taxonomy of ontology-based approaches is proposed and representative systems are compared, followed by a discussion of evaluation methods and datasets. The review then examines open challenges, including ontology construction cost, scalability, dynamic knowledge, cold start, evaluation validity, and explainability. It closes by outlining future directions that combine large language models with knowledge graphs, hybrid neuro-symbolic reasoning, and federated privacy-preserving personalization. Throughout, the aim is an honest synthesis of what prior work reports rather than any new empirical claim.