Assessing the Capability of Large Language Models for Domain-Specific Ontology Generation