| Title |
Development of an LLM-based Automated Model for Improved Construction Constraint Recognition and Classification |
| Authors |
이슬비(Lee, Seulbi) ; 고태우(Ko, Taewoo) |
| DOI |
https://doi.org/10.5659/JAIK.2026.42.1.391 |
| Keywords |
Schedule Plan; Construction Constraints; Large Language Model; Automation; Information Management |
| Abstract |
The management of requirements and constraints is critical in construction project management, as these factors fundamentally influence
project performance. However, construction projects frequently faces issues due to the late or inaccurate identification of project constraints,
leading to inefficiencies, delays, and increased costs. Traditional constraint management approaches rely heavily on manual reviews, which are
unable to keep pace with the growing complexity and volume of documents. To address this challenge, this study proposes a novel model
for constraint recognition and classification utilizing large language models (LLMs), specifically applying both encoder-based and
decoder-based transformer architectures. The methodology involves collecting construction project documents, labeling constraints based on
their types, and training fine-tuned LLMs for automatic identification and classification of legal, technical, economic, and environmental
constraints. The results indicate that all models delivered satisfactory performance in the initial phase of constraint recognition, validating their
general applicability. The main contributions of this research include (1) the development of an automated and scalable LLM-based framework
for constraint management in construction projects, (2) comprehensive benchmarking of encoder and decoder transformer models in this
domain, and (3) practical insights for optimized deployment of AI models for construction process improvement. |