Moving Beyond Count-based Approach:
Understanding Hong Kong Nationalism with Topic Network
Justin Chun-ting Ho
University of Edinburgh
Data
Data
- Facebook page of key nationalist organisations in Hong Kong
- 8,783 posts, average of 70 characters per post, mostly Chinese but also English
Why Facebook?
Why Facebook?
- An important platform of political communication (Michael Chan, Hsuan-Ting Chen & Francis L.F. Lee, 2016;
Hsuan-Ting Chen, Michael Chan & Francis L.F. Lee, 2016;
Francis L.F. Lee & Joseph Man Chan, 2015; Gary Tang & Francis L.F. Lee, 2013) - A major channel of mobilization during the Umbrella Movement (Paul S. N. Lee, Clement Y. K. So & Louis Leung, 2015)
- All nationalist groups use Facebook
Methodology
A mixed method approach of Critical Discourse Analysis (CDA) and Computational Text Analysis
Why Mixed Method?
Critical Discourse Analysis
- Close analysis of language
- Revealing the discursive struggle and unveiling the strategies employed by the nationalists to construct the nation.
Challenges of CDA
- Unsystematic manner of applying analytical procedures

(Koller & Mautner, 2004; McEnery & Wilson, 1996) - Unable to deal with large corpus
- Sampling could be problematic
Computational Text Analysis
- Systematic analysis of large-scale text collections (using computer progarmmes)
- Draws on techniques from natural language processing, information retrieval, and machine learning
- Augmenting close reading, not replacing
(Read Grimmer and Stewart (2013) for the pitfalls of Text Analysis)
Topic Modeling
Count-based Approach
Limitation
- Texts serve different purposes
- Count-based approach assumes all texts matter to the discourse equally
- Abundance is not equal to importance
Topic Network Approach
Topic Network Approach
- Concepts within a discourse do not exist in isolation, they are dependent on each other in various ways
- Discourse as network phenomenon
- Centrality to identify core concepts that unite different concepts into a coherent whole
