Enhancing sustainability assessment of highway Infrastructure through ontology-based knowledge representation and large language models

Zenodo (CERN European Organization for Nuclear Research)

Abstract

Sustainability assessment of highway infrastructure is hindered by fragmented, inconsistent, and incomplete information, limiting transparency and comparability. To overcome these limitations, an ontology-based knowledge representation was developed to formally structure sustainability-related data, complemented by large language models (LLMs) for natural-language querying, semantic linking, and quality control. LLMs assist in detecting inconsistencies, identifying missing information, and aligning content with expert knowledge. Using competency questions and expert review, the ontology was iteratively refined. Results show a transparent, structured, and scalable knowledge base. The approach's contribution lies in combining formal ontologies with LLMs to enhance accessibility, interpretability, and reliability for future assessments.

Authors 4

1. Michelle Kaus Aachen ICoM

   RWTH Aachen University

   Affiliation as printed

   ICoM RWTH Aachen University, Germany

2. Joy Sengupta Aachen ICoM

   RWTH Aachen University

   Affiliation as printed

   ICoM RWTH Aachen University, Germany

3. Sven Mackenbach Aachen ICoM

   RWTH Aachen University

   Affiliation as printed

   ICoM RWTH Aachen University, Germany

4. Katharina Klemt-Albert Aachen ICoM

   RWTH Aachen University

   Affiliation as printed

   ICoM RWTH Aachen University, Germany