Scholarlink

Ontologies are known to improve the ability of a Large Language Model (LLM) to translate natural language queries into formal query languages. One approach provides the ontology as part of the input question, but enterprise ontologies are often too large due to token size limitations. Our solution incrementally reveals “just enough” of an ontology to answer a given question.

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

Using Off-the-Shelf LLMs to Query Enterprise Data by Progressively Revealing Ontologies

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • C. Civili,
  • E. Sherkhonov,
  • R. E. K. Stirewalt

摘要

Ontologies are known to improve the ability of a Large Language Model (LLM) to translate natural language queries into formal query languages. One approach provides the ontology as part of the input question, but enterprise ontologies are often too large due to token size limitations. Our solution incrementally reveals “just enough” of an ontology to answer a given question.