Enhancing Construction Requirements Documents Analytics through Integrating Ontology and Large Language Model

Abstract Ineffective management of construction requirements results in delays, cost overruns, and compliance challenges due to extensive documentation and manual processes. While digital engineering improves automation, its effectiveness depends on accurate information extraction. Existing large language model (LLM)-based approaches have improved construction text analytics, and construction ontologies have supported structured knowledge representation. However, ontology-guided LLM workflows for practical sentence-level extraction and structuring of construction requirements remain underexplored. This study addresses these challenges by integrating a new ontology with an LLM, using the ontology for structured knowledge representation and the LLM for advanced language processing. The proposed ontology systematically captures the foundational aspects of construction requirements across diverse document types, improving organization, retrieval, and interpretation. Then this study integrates ontology with a fine-tuned LLM to better understand construction-specific language. The results demonstrate the feasibility of the proposed framework for producing consistent sentence-level semantic extraction from quality-related construction requirement text. This study contributes a reusable ontology for modeling construction requirement sentences and demonstrates how ontology-guided fine-tuning can provide explicit semantic grounding for LLM-based information extraction. The proposed framework transforms unstructured requirement text into structured semantic entities, supporting downstream tasks such as requirement retrieval, traceability analysis, and automated compliance checking.