Scholarlink

This paper presents insights into a real-world railway ontology and a corresponding Knowledge Graph Question Answering (KGQA) approach within the domain of train and track operations, developed in collaboration with DB Systel GmbH (subsidiary of the main German railway company). Our solution enables seamless access to infrastructure elements, train schedules, and operational data via a question-answering system powered by Large Language Models (LLMs). To ensure high quality, cost efficiency, and runtime performance, LLMs are utilized exclusively for SPARQL query generation, striking a balance between flexibility and control. This approach significantly enhances the precision and efficiency of information retrieval, making railway data more accessible and actionable. We discuss the implementation, evaluation results, and key trade-offs of our approach.

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

A Domain-Specific Question-Answering System for Railway Data: A Hybrid Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Andreas Both,
  • Kai Herbst,
  • Rene Krieg,
  • Magdalena Landmann,
  • Martin Raphael Alef,
  • Anton Diettrich

摘要

This paper presents insights into a real-world railway ontology and a corresponding Knowledge Graph Question Answering (KGQA) approach within the domain of train and track operations, developed in collaboration with DB Systel GmbH (subsidiary of the main German railway company). Our solution enables seamless access to infrastructure elements, train schedules, and operational data via a question-answering system powered by Large Language Models (LLMs). To ensure high quality, cost efficiency, and runtime performance, LLMs are utilized exclusively for SPARQL query generation, striking a balance between flexibility and control. This approach significantly enhances the precision and efficiency of information retrieval, making railway data more accessible and actionable. We discuss the implementation, evaluation results, and key trade-offs of our approach.