AutOnto: Towards A Semi-Automated Ontology Engineering Methodology

This paper addresses the challenge of efficiently constructing domain ontologies for large, rapidly evolving domains, where manual approaches often struggle to overcome knowledge acquisition bottlenecks. To overcome these limitations, we developed an automated framework, AutOnto, for knowledge extraction and ontology conceptualization that leverages Large Language Models (LLMs) and natural language processing (NLP) techniques. AutOnto integrates BERT-based topic modeling with LLMs to automate the extraction of concepts and relationships from text corpora, facilitating the construction of taxonomies and the generation of domain ontologies. We applied AutOnto to a dataset of NLP-specific articles from OpenAlex and compared the resulting ontology generated by our automated process against a well-established gold-standard ontology. The results indicate that AutOnto achieves comparable levels of quality and correctness while significantly reducing the amount of data required and the dependence on domain-specific expertise. These findings highlight AutOnto’s efficiency and effectiveness in knowledge extraction and ontology generation. This work has significant implications for rapid ontology development in large, evolving domains, potentially mitigating the knowledge acquisition bottleneck in ontology engineering.