Multifunctional Analysis of Construction Contracts Using a Machine Learning Approach

Abstract

In the intricate domain of construction contracts, precise descriptions and measurements of contract structures are crucial. This study provides an objective analysis of the structure of construction contracts from a multifunctional perspective. A deep learning–based machine coding model was trained using 17 standard contracts and 35 actual contracts. The model was then used to code an additional 117 actual contracts. Statistical analysis was conducted to compare the distribution of the three functions (i.e., control, coordination, and adaptation) between standard and actual contracts. The results revealed that coordination has the highest contribution among the three functions. Moreover, actual contracts exhibit increased complexity compared with standard contracts, often containing additional control and coordination provisions related to project-specific obligations and tasks. The 117 actual contracts were further classified based on project delivery systems (PDSs) and pricing methods, and the impact of PDSs and pricing methods on the functional distribution was examined. The results showed more flexible adaptation and more complex control provisions specified in design-build/engineering, procurement, and construction (DB/EPC) and lump sum contracts. Theoretically, this study provides insights into objective measures in contract research and enriches the body of knowledge on the structure of construction contracts from a multifunctional perspective. Practically, professionals are provided with guidance on managing the complexity of each functional provision.