Scholarlink

This study introduces an automated framework for constructing subsumption hierarchies in alignment with OntoClean principles, utilizing Large Language Models (LLMs), specifically GPT-4. The framework utilizes LLMs for concept grouping, intermediate layer generation, and validation. Specifically, the framework focuses on constructing hierarchical structures that maintain structural and semantic alignment with OntoClean principles. Using the FoodOn Ontology as a benchmark, the evaluation revealed a Normalized Graph Edit Distance score of 0.8, and Jaccard Index of 0.76 for edge sets indicating a high degree of structural similarity between the ground truth and the generated graphs. The automated nature of the framework offers significant efficiency gains compared to manual or semi-automated methods.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automating OntoClean - Subsumption Hierarchy Construction

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Shathika Kularatne,
  • Wolfgang Mayer,
  • Markus Stumptner

摘要

This study introduces an automated framework for constructing subsumption hierarchies in alignment with OntoClean principles, utilizing Large Language Models (LLMs), specifically GPT-4. The framework utilizes LLMs for concept grouping, intermediate layer generation, and validation. Specifically, the framework focuses on constructing hierarchical structures that maintain structural and semantic alignment with OntoClean principles. Using the FoodOn Ontology as a benchmark, the evaluation revealed a Normalized Graph Edit Distance score of 0.8, and Jaccard Index of 0.76 for edge sets indicating a high degree of structural similarity between the ground truth and the generated graphs. The automated nature of the framework offers significant efficiency gains compared to manual or semi-automated methods.