| description abstract | Public concerns exert far-reaching influence on various phases of megaprojects, requiring the decision makers to achieve dynamic analysis in the aspects of identification, measurement, and management. The study proposes an integrated topic and sentiment modeling approach to analyze the dynamics of public concerns from unstructured project documents. First, the topic-over-time (TOT) model is adopted to identify the public concerns and trace the trend of public popularity on the concerns. Second, the bidirectional encoder representations from transformers (BERT)-based sentiment model is developed to reveal the trend of public sentiment toward each public concern. Finally, a mirror “N” strategic model is proposed considering the trend of public popularity and sentiment, together with the classical public participation strategies: collaboration, consultation, involvement, and information. With the 1,748 official documents from the Hong Kong–Zhuhai–Macao Bridge, the proposed method is validated. As a result, 16 public concerns and their levels of popularity trends are identified in 16 years of project duration by the TOT model. The volatile and mild public sentiment changes are tracked in the timeline by the BERT-based sentiment model. The recommendation of management strategies derived from the mirror “N” strategic model is summarized on public concerns in three project phases: planning, construction, and handover. The dynamic data-driven method bridges the knowledge domains of public participation studies and text-mining technologies for better megaproject management. | |
