The construction industry generates 13% of global gross domestic product; however, it accounts for one-third of global greenhouse gas emissions, creating an urgent need to align industry practices with sustainable development goals (SDGs). This research introduces sustainable development goals bidirectional encoder representations from transformers (sdgBERT), a deep-learning model designed for comprehensive SDG mapping of construction discourse through natural language processing techniques. The methodology involved fine-tuning a pretrained bidirectional encoder representations from transformers (BERT) model using 42,065 manually labeled multidomain text extracts to enable systematic SDG classification. The model demonstrates high accuracy with weighted averages of 0.90 for precision, recall, and F1-scores. Validation on 4,912 manually labeled construction domain sentences confirms robust generalization across diverse document types. The model’s explainability is demonstrated through contextual text plots, revealing its capacity for contextually relevant predictions. The model’s practical application is further illustrated through temporal analysis of SDG trends in construction research, establishing a foundation for automated sustainability mapping in the industry. This research advances construction informatics by developing and validating sdgBERT for comprehensive SDG mapping of construction text. The model demonstrates superior accuracy over existing approaches, providing a robust framework that enables systematic sustainability mapping and establishes new directions for automated sustainability discourse analysis in construction management.
Sadick et al. (2026) studied this question.