Integration of a Natural Language Processing and Project Scheduling Tool for Contractual Delay Risk Identification in a Highway Construction Project Using LLMs

Construction contracts contain essential information regarding delay clauses. These clauses are associated with improper execution of work in accordance with contractual requirements. Contract documents require thorough review. Additionally, the limited time provided during bidding, affects manual reviews, and contractual delay terms may be overlooked and later arise during the construction stage. This research aims to develop a framework for automating identification of contractual delay terms associated with schedule delays through integration of natural language processing (NLP), project scheduling tools, and large language models (LLMs). The developed framework was applied in a case study and its applicability and functionality were evaluated. Evaluations focused on the correctness of the contractual delay terms generated by the model. The evaluation results indicate that the developed framework supports identification of contractual delay terms associated with schedule delays in the construction phase of the projects. The model performance is validated leveraging K-fold cross validation technique, evaluating training loss, training token-level accuracy, validation loss, validation token-level accuracy. It was found that the model displayed a strong performance, represented by the preceding model performance metrics with training loss and validation loss of 0.34 and 0.359, respectively and achieving training token-level accuracy and validation token-level accuracy of 92.69% and 91.87%, respectively. The framework supports a promising approach of integrating contract analysis with dynamic project schedule in a single environment and effectively enhance automation in contract analysis and identification of contractual delay risks, eventually reducing time and efforts of project stakeholders during manual contract analysis and project tracking.