Heterogeneous Ontology Repository for Intelligent E-Learning

A large number of ontologies have been developed over the past two decades for the education domain. Some of these ontologies are available in public repositories published within the Linked Open Data cloud. However, a significant portion of educational ontologies remain distributed across project-specific websites, making their discovery and access challenging. Ontologies designed for education often have domain- or task-specific characteristics and conceptual structures. To facilitate their discovery, interoperability, actualization and reuse, it is essential to annotate them with rich, standardized metadata, such as domain coverage, pedagogical objectives, target learner groups, and technical specifications, to enable effective search and support integration within educational systems. Other components, such as knowledge graphs, rules, learning analytics, and machine learning-based models also play an important role. In this research, a conceptual model of a heterogeneous educational ontology repository for storing and reusing ontologies, knowledge graphs, and other objects and tools needed for the development of knowledge bases for intelligent education systems is proposed. An OWL ontology modeling the needed metadata for the description of repository objects and supporting semantic search and recommendations to support the development of knowledge bases for intelligent educational systems is also developed. The proposed heterogeneous ontology repository can help in solving many of the challenges related to hallucinations, transparency, personalization, privacy, and pedagogical alignment that arise when integrating large language models into educational systems by proposing or recommending easy-to-use ontologies for the development of intelligent educational systems, integrating generative AI, symbolic AI, machine learning and statistical techniques. It also integrates LLMs to ensure effective and easy search, recommendation of stored objects, and ontology management. The proposed LLM-powered ontology extraction use case demonstrates an encouraging ontology metadata extraction quality (a precision of about 0.7 and a recall of about 0.9) combined with an ontology development strategy that is easy for education professionals to use.