Dynamic Stakeholder-Associated Topic Modeling on Public Concerns in Megainfrastructure Projects: Case of Hong Kong–Zhuhai–Macao Bridge

Dynamic Supply Chain Capability Analysis of Hong Kong-Zhuhai-Macao Bridge Construction: A Topic Modeling Approach

“…Ekanayake, Geoffrey Qiping Shen, Mohan M. Kumaraswamy and Jin Xue primary analysis was conducted using the TOT modeling technique. Compared to other topic modeling techniques such as Latent Dirichlet Allocation (LDA), PLSA and probabilistic theory, Topic Over Time (TOT) modeling is considered to be a more effective technique (Xue et al, 2020). This is because TOT captures the low-dimensional data structures, and also, TOT detects the structure's changes over time which is unavailable with the other topic modeling techniques (Wang and McCallum, 2006).…”

“…Then, following the study of Wang and McCallum (2006), each document's timestamp was counted by considering all the word timestamps in a document. The hyperparameters of α and β were kept as α = 50/T and β = 0.1 (Xue et al, 2020) considering the model's simplicity (Wang and McCallum, 2006). Further, TF-IDF filtering was done prior to the modeling to enhance the texts' quality by removing frequent but meaningless texts.…”

“…Finally, the model generated six SCC topics associated with the HZMB project, including 15 feature words with high relevant probabilities in each. Given the feature word list 𝑉 (𝑧) , each topic's coherence [that measures the degree of logical consistency for text], 𝐶(𝑍; 𝑉 (𝑧) ) was calculated and tested using Equation 01, as suggested by Xue et al (2020).…”

“…Further, it was based on static analysis; hence, it lacks the evidence to support the dynamics of SCCs throughout the project phases (Lin et al, 2018). These limitations could be successfully overcome through text mining and case study research (Xue et al, 2020). Topic modeling is a robust text-mining tool that detects the core commonalities among a pool of texts (Wang and McCallum, 2006).…”

“…Topic modeling is a robust text-mining tool that detects the core commonalities among a pool of texts (Wang and McCallum, 2006). Among the topic modeling methods, Topic Over Time (TOT) modeling is considered to be more effective (Xue et al, 2020) as it explores both the content of core concepts and the dynamic concept patterns (Wang and McCallum, 2006).…”