Automation of BOQ Analysis and Material List Selection Using NLP and Multi-Class Prediction

Tendering and procurement departments play a vital role in securing projects for contractors. However, such departments often deal with repetitive and time-consuming tasks when analyzing owners’ specifications and Bill of quantities (BOQs) data, which diverts their attention from more complex, high-value tasks. In addition, material selection outcomes vary based on engineers’ expertise, leading to a lack of standardization and frequent inaccuracies. This paper aims to provide an automated solution to detect contract items based on BOQs’ descriptions and determine the required materials needed. This is carried out through the utilization of Bidirectional Encoder Representations from Transformers (SBERT) to classify BOQ items based on their descriptions. In addition, the BOQ embeddings acquired from SBERT are then used in a multi-class prediction model that is able to predict the required materials. The presented approach classified BOQ items with an average F1-score of 0.94. Thus, the presented novel approach can decrease the tendering and procurement phases and contribute to developing a competitive and effective construction market.