Options
A corpus-based diachronic analysis of Covid-19 topics, sentiments, metaphors in media discourse: A case study of Hong Kong free pree
Author(s)
Date Issued
2025
Publisher
Hong Kong: Hong Kong Shue Yan University
Description
126 pages
Type
Thesis
Programme
Master of Philosophy of English
Abstract
The COVID-19 pandemic is a global public health crisis that has had profound impacts on all aspects of society. During the COVID-19 pandemic, news media played a crucial role in not only promptly disseminating information on epidemic prevention and control such as wearing masks, maintaining social distancing, and receiving vaccinations but also facilitating effective communication between the government and the public regarding epidemic prevention policies and health guidance. With this significance, considerable scholarly attention has been paid to examining how news media outlets not only framed the pandemic but also modulated their affective valence of its coverage and deployed metaphorical language to interpret the crisis. Nonetheless, little systematic research has been done on these diachronic changes in the context of Hong Kong news media. Furthermore, prior media discourse studies on COVID-19 relied on manual content analysis which is time-consuming and prone to subjectivity. In particular, the use of natural language processing (NLP) technology which has been widely applied in the fields of social media and e-commerce remains unexplored in the field of news media. This study aims to fill these theoretical and methodological voids by investigating the diachronic evolution of topics, sentiments, and metaphor use in Hong Kong news media coverage on COVID-19 using NLP techniques. Specifically, it seeks to find answers to the following questions: (1) How do topics in the Hong Kong news media coverage of COVID-19 evolve over time during the pandemic? (2) How do the emotional tones of the Hong Kong news media coverage of COVID-19 fluctuate over time? (3) How does metaphor use in the Hong Kong news media coverage of COVID-19 change over time?
A total of 2541 reports related to COVID-19 from December 31, 2019 to August 16, 2024. were collected from the Hong Kong Free Press website. BERT was employed to compare the distribution of the topics covered by the Hong Kong news media during the COVID-19 pandemic and their similarities and differences across its different stages, whereas TextBlob was applied to determine and quantify news authors’ attitudes at different stages of the pandemic by classifying their tones as positive, negative, or neutral. The Metaphor Identification Procedure (MIP) and its extended form Metaphor Identification Procedure Vrije Universiteit (MIPVU) were also applied to detect metaphorical expressions.
The research found that eight relatively recurrent news themes were identified, including Country and official, Law and security, Test and health, Quarantine and travel, Hospital and infection, Mask and worker, City and restaurant, and Vaccine and biontech, which are in line with the complexity of Hong Kong's social and political situation. There are strong internal connections among dimensions such as government measures, medical systems, law enforcement and social security. These discussions not only respond to the public health crisis but are also influenced by Hong Kong's unique social and political landscape. The topic modeling framework tracks the structural changes in narrative focus, public concern and media priorities. In terms of emotional diachronic change, the initial stage was mostly characterized by fear and uncertainty, but as vaccines were rolled out and government strategies gradually improved, the tone of the news gradually turned neutral and even cautiously optimistic. However, when epidemic prevention policies sparked controversy or the infection rate rose again, negative emotions also rose significantly. The metaphor analysis revealed that the “war” metaphor (fight, threat, etc.) was prevalent across all stages of the pandemic, whereas “journey” metaphors (e.g., milestone, journey) surged during its middle-to-late stage. Overall, Hong Kong’s COVID‑19 news discourse is highly dynamic and appears to shape the public’s perception of pandemic risk, their judgment of the legitimacy of government public‑health policies, and their willingness to comply with prevention measures.
This research has multiple significances in the field of media discourse. Theoretically, this study demonstrates a clear causal chain: when Hong Kong news media adjusts their framing of the epidemic, the news frames alter public sentiment; these emotional shifts, in turn, pressure policymakers to recalibrate their responses, and the resulting policies ultimately reshape society’s overall understanding of the pandemic. Methodologically, the study weaves together BERT‑based topic modelling, TextBlob lexicon‑driven sentiment scoring, and the multi‑step MIP/MIPVU metaphor‑identification procedure into a unified corpus‑linguistic workflow that delivers high accuracy in tracking thematic shifts, sentiment polarity, and metaphor usage, while offering a systematic, replicable approach adaptable to large‑scale diachronic text analysis.
A total of 2541 reports related to COVID-19 from December 31, 2019 to August 16, 2024. were collected from the Hong Kong Free Press website. BERT was employed to compare the distribution of the topics covered by the Hong Kong news media during the COVID-19 pandemic and their similarities and differences across its different stages, whereas TextBlob was applied to determine and quantify news authors’ attitudes at different stages of the pandemic by classifying their tones as positive, negative, or neutral. The Metaphor Identification Procedure (MIP) and its extended form Metaphor Identification Procedure Vrije Universiteit (MIPVU) were also applied to detect metaphorical expressions.
The research found that eight relatively recurrent news themes were identified, including Country and official, Law and security, Test and health, Quarantine and travel, Hospital and infection, Mask and worker, City and restaurant, and Vaccine and biontech, which are in line with the complexity of Hong Kong's social and political situation. There are strong internal connections among dimensions such as government measures, medical systems, law enforcement and social security. These discussions not only respond to the public health crisis but are also influenced by Hong Kong's unique social and political landscape. The topic modeling framework tracks the structural changes in narrative focus, public concern and media priorities. In terms of emotional diachronic change, the initial stage was mostly characterized by fear and uncertainty, but as vaccines were rolled out and government strategies gradually improved, the tone of the news gradually turned neutral and even cautiously optimistic. However, when epidemic prevention policies sparked controversy or the infection rate rose again, negative emotions also rose significantly. The metaphor analysis revealed that the “war” metaphor (fight, threat, etc.) was prevalent across all stages of the pandemic, whereas “journey” metaphors (e.g., milestone, journey) surged during its middle-to-late stage. Overall, Hong Kong’s COVID‑19 news discourse is highly dynamic and appears to shape the public’s perception of pandemic risk, their judgment of the legitimacy of government public‑health policies, and their willingness to comply with prevention measures.
This research has multiple significances in the field of media discourse. Theoretically, this study demonstrates a clear causal chain: when Hong Kong news media adjusts their framing of the epidemic, the news frames alter public sentiment; these emotional shifts, in turn, pressure policymakers to recalibrate their responses, and the resulting policies ultimately reshape society’s overall understanding of the pandemic. Methodologically, the study weaves together BERT‑based topic modelling, TextBlob lexicon‑driven sentiment scoring, and the multi‑step MIP/MIPVU metaphor‑identification procedure into a unified corpus‑linguistic workflow that delivers high accuracy in tracking thematic shifts, sentiment polarity, and metaphor usage, while offering a systematic, replicable approach adaptable to large‑scale diachronic text analysis.
File(s)
Name
232703M.pdf
Description
Embargo period 2025-2028
Size
1.29 MB
Format
Adobe PDF
Checksum
(not present)
(MD5):c80e0f6d641b1e03b12155a5ff3f4c5b
Availability at HKSYU Library
