Options
From policy to perception: Sentiment analysis and topic modeling on housing issues in Hong Kong
Author(s)
Date Issued
2025
Citation
ICLLS 2025 Seventh International Conference on Linguistics and Language Studies.
Type
Conference Paper
Abstract
This study aims to examine sentiment patterns and thematic evolutioninacorpus of political discourse on housing issues in Hong Kong, utilizing the LargeLanguage Model (LLM)-BERT for analysis. Building on prior research demonstratingthe effectiveness of thematic and sentiment analysis in discourse studies, particularlyfor large-scale datasets (Jacobs & Tschötsche, 2019; Isoaho et al., 2021), this studyanalyses a comprehensive corpus of 40 years of annual policy addresses bytheGovernors (1984 to 1997) and Chief Executives (1997 to 2024) of Hong Kong, consisting of 94,6291 words. Given that “housing” has been a persistent societal challenge in Hong Kong, we anticipate a constant occurrence of this topic inthecorpus, with variations in thematic focus and sentiment reflecting the evolving socio- political context over four decades (1984-2024). The present study uses a mixedresearch approach to address three research questions: 1) What themes arediscussed in the housing discourse in Hong Kong, and how have they evolvedovertime? 2) How did sentiment (positive vs. negative polarity) and subjectivity (opinion- based vs. fact-based focus) shift over time? 3) What are the underlying rhetorical strategies employed when addressing housing issues?
Our preliminary findings from the quantitative topic modelling analysisidentified six themes: 政府公屋计划 (government public housing program), 土地供應 (land supply), 重 建 樓 宇 (building redevelopment), 新界土地發展(NewTerritories Land Development), 啟德發展 (Kai Tai Development), and 大嶼山人工島填海發展 (Lantau Artificial Islands Reclamation). These topics highlight keyareassuch as government housing initiatives, land supply strategies, urban renewal, andthe socio-economic impacts of housing policies. The sentiment analysis revealedanoverall positive trend with a peak around 2007, indicating favorable governmental perceptions of the issue. However, sentiment declined post-2007, with fluctuationsand a sharp drop in 2012, reflecting shifting public attitudes due to changing socio- political contexts. Drawing upon the appraisal system (Martin & White, 2005), further qualitative attitudinal discussion of the thematic and sentiment patterns will be conducted based on the structural, lexicogrammatical and contextual resourcesfrom the corpus.
The present study offers insights into the evolving focus of housing policies andpublic sentiment in Hong Kong. Methodologically, the study advances discourseanalysis by integrating LLMs, enriching the practice of applying State-of-the-art Natural Language Processing techniques such as topic modeling and sentiment analysis in corpus-based discourse studies. Using the LLM-BERT model to explorethematic and sentiment insights in housing discourse could offer potential implications for housing policy development and implementation, possibly helpingpolicymakers address public concerns and refine housing strategies.
Our preliminary findings from the quantitative topic modelling analysisidentified six themes: 政府公屋计划 (government public housing program), 土地供應 (land supply), 重 建 樓 宇 (building redevelopment), 新界土地發展(NewTerritories Land Development), 啟德發展 (Kai Tai Development), and 大嶼山人工島填海發展 (Lantau Artificial Islands Reclamation). These topics highlight keyareassuch as government housing initiatives, land supply strategies, urban renewal, andthe socio-economic impacts of housing policies. The sentiment analysis revealedanoverall positive trend with a peak around 2007, indicating favorable governmental perceptions of the issue. However, sentiment declined post-2007, with fluctuationsand a sharp drop in 2012, reflecting shifting public attitudes due to changing socio- political contexts. Drawing upon the appraisal system (Martin & White, 2005), further qualitative attitudinal discussion of the thematic and sentiment patterns will be conducted based on the structural, lexicogrammatical and contextual resourcesfrom the corpus.
The present study offers insights into the evolving focus of housing policies andpublic sentiment in Hong Kong. Methodologically, the study advances discourseanalysis by integrating LLMs, enriching the practice of applying State-of-the-art Natural Language Processing techniques such as topic modeling and sentiment analysis in corpus-based discourse studies. Using the LLM-BERT model to explorethematic and sentiment insights in housing discourse could offer potential implications for housing policy development and implementation, possibly helpingpolicymakers address public concerns and refine housing strategies.
Availability at HKSYU Library
