Towards Explainable Agentic Ontology Engineering

Large Language Models have revolutionized the research landscape in many domains, including knowledge engineering.
However, while several approaches have been proposed to aid ontology engineers, ontology development remains a complex and labor-intensive task that often requires substantial expertise to ensure competency question coverage, adherence to modeling best practices, vocabulary reuse and documentation.
In this talk I will describe how multi-agent systems may be used to enable end-to-end generation of ontologies to satisfy a set of competency questions.
Here, the role of accountability and provenance become essential to help determine why a design decision has been included in the final ontology model. The talk will review early experimentation at the Ontology Engineering Group, evaluation challenges and future lines of work for enabling explainable ontology generation.