Integrating Ontology Rules with Large Language Models for Enhanced Reasoning
Incheon Paik · 2025
Ontology-based reasoning enables structured inference and logical consistency in specialized domains. However, traditional ontology reasoning frameworks rely heavily on predefined inference engines and XML-based structured data, limiting their integration with Large Language Models (LLMs). Despite their strong general reasoning capabilities, LLMs struggle with structured ontological knowledge due to their reliance on unstructured text.This study proposes a novel approach to integrating ontology-based reasoning into LLMs by transforming ontological concepts and rules into structured natural language. By systematically defining ontology components—such as classes, properties, and rules—using a natural language-friendly format, we enable LLMs to process and infer knowledge from domain-specific ontologies without dedicated inference engines. The feasibility of this transformation is demonstrated through experiments on a domain-specific cocktail ontology. Our findings highlight the potential of natural language-driven ontology representation in bridging the gap between rule-based inference and flexible LLM reasoning.