Public sentiment analysis on urban regeneration: A massive data study based on sentiment knowledge enhanced pre-training and latent Dirichlet allocation

Abstract

Background Public satisfaction is the ultimate goal and an important determinant of China’s urban regeneration plan. This study is the first to use massive data to perform sentiment analysis of public comments on China’s urban regeneration. Methods Public comments from social media, online forums, and government affairs platforms are analyzed by a combination of Natural Language Processing, Knowledge Enhanced Pre-Training, Word Cloud, and Latent Dirichlet Allocation. Results (1) Public sentiment tendency toward China’s urban regeneration was generally positive but spatiotemporal divergences were observed; (2) Temporally, public sentiment was most negative in 2020, but most positive in 2021. It has remained consistently negative in 2022, particularly after February 2022; (3) Spatially, at the provincial level, Guangdong posted the most comments and Tibet, Shanghai, Guizhou, Chongqing, and Hong Kong are provinces with highly positive sentiment. At the national level, the east and south coastal, southwestern, and western China regions are more positive, as opposed to the northeast, central, and northwest regions; (4) Topics related to Shenzhen’s renovations, development of China’s urban regeneration and complaints from residents are validly categorized and become the public’s key focus. Accordingly, governments should address spatiotemporal disparities and concerns of local residents for future development of urban regeneration.