Enriching the EuSO ontology through GenAI-based terminological selection and conceptualisation

Abstract

In academic and institutional contexts, the management and structuring of specialised knowledge are essential to support effective communication and internationalisation. Traditionally, ontology development and ontology-driven terminology work have relied on expert-based manual processes and corpus analysis techniques. However, recent advances in generative artificial intelligence (GenAI) and large language models (LLMs) have opened new possibilities for supporting knowledge representation tasks. In this context, this study explores the use of a customised GenAI-based chatbot for assisting ontology development in the domain of European higher education. Specifically, The Ontology Assistant, built on ChatGPT-5.2 and enriched with a domain-specific corpus, was designed to support the identification and structuring of ontological categories within the EuroScholar Ontology (EuSO). The model’s performance was evaluated through six experimental runs using a gold standard derived from a selected ontological category. Precision, recall, and F1-measure were used to assess category retrieval. The results show consistent behaviour across runs, characterised by high recall and low precision, reflecting the exploratory capacity of the chatbot to identify candidate ontological categories. These findings suggest that GenAI chatbots can function as useful support tools within ontology development workflows, provided that expert intervention is applied to verify and maintain alignment with the EuSO.