Cross-Cultural Differences in Public Discourse on COVID-19 Vaccination in the United States and South Korea: Cross-Sectional Analysis Using Natural Language Processing
JMIR Formative Research · 2026-03-05 · 10 · e84791
DOI: 10.2196/84791 ↗At a glance
Querschnittstudie in JMIR Formative Research, published online on 2026-03-05. Written by Sangpil Youm and 4 co-authors. Institution: Department of Computer and Information Science and Engineering, University of Florida, Gainesville, FL, United States. Major MEDLINE topics: COVID-19 Vaccines, COVID-19, Cross-Cultural Comparison, Natural Language Processing. 88 references. The full text is freely available in PubMed Central.
Abstract (original language)
Background
The COVID-19 vaccine was introduced as a crucial tool to combat the pandemic. However, concerns about its effectiveness, side effects, and misinformation spread remain. Prior research largely relied on survey-based approaches with limited populations. To address these limitations, social media offers a broader, more naturalistic lens into public discourse on COVID-19 vaccination. Accordingly, our study leverages social media data to identify factors shaping vaccine-related information needs, perceptions, and communication dynamics.
Objective
This study investigated public discourse about COVID-19 vaccines on community-driven question-and-answer sites in the United States (Quora; Quora, Inc) and South Korea (Naver Knowledge-iN; Naver Corp) to identify cross-national similarities and differences in vaccine-related information needs, sentiment patterns, and public perceptions over time.
Methods
We analyzed publicly available COVID-19 vaccine–related questions and answers posted between June 27, 2020, and June 27, 2021, on 2 community-driven question-and-answer platforms: Quora (United States) and Naver Knowledge-iN (South Korea). After preprocessing and sample-size matching, the dataset included 3952 question-answer pairs per platform, with one community-selected (most upvoted) answer analyzed per question. Natural language processing (NLP) techniques were applied for topic classification and sentiment analysis. Questions were categorized using a hybrid topic modeling approach combining Latent Dirichlet Allocation (LDA) and Top2Vec, identifying 5 topics on Quora and 7 topics on Naver Knowledge-iN. Answer sentiments were classified using an ensemble of Bidirectional Encoder Representations from Transformers (BERT; Google LLC)– and Efficiently Learning an Encoder that Classifies Token Replacements Accurately (ELECTRA; Google LLC)–based transformer models, and temporal sentiment trends were examined using monthly aggregation.
Results
Five shared information needs emerged, including effects of vaccines, variants, government policy, visiting overseas, and different vaccines, while South Korea uniquely exhibited vaccination appointments (711/3952, 18%) and school and education (513/3592, 13%). Negative sentiment predominated in US (Quora) answers across 4 of 5 topics, whereas positive sentiment exceeded 50% (498/790, 337/474, 367/592, 218/316, 348/553, 562/711, and 364/513) across all 7 topics on Naver Knowledge-iN. Temporally, US sentiment exhibited multiple positive-negative crossovers, whereas Korean sentiment stabilized toward positivity after February 2021, coinciding with the national vaccine rollout. Question-answer sentiment pairs showed contrasting interaction patterns, including negative-negative pairs dominated in the United States (eg, 504/978, 51.5% for different vaccines), while in South Korea, positive-positive and negative-positive pairs accounted for more than 63% (498/790, 337/474, 367/592, 218/316, 348/553, 562/711, and 364/513) of interactions in 7 topics, with positive-positive pairs most prevalent in 6 of 7 topics, except for variants.
Conclusions
Public perceptions of COVID-19 vaccines and related information needs differ between the 2 countries, shaped by cultural context, trust in government, and information-seeking environments. Analysis of social question and answer data from the 2 countries reveals shared information needs but divergent sentiment patterns. These findings highlight the value of social media data for public health research and the need for culturally and platform-specific communication strategies.
Subjects
MeSH terms (MEDLINE)
Author keywords
Substances
Authors and institutions
- Department of Computer and Information Science and Engineering, University of Florida, Gainesville, FL, United States.
- Department of Sociology, Korea University, Seoul, Republic of Korea.
- Luddy School of Informatics, Computing, and Engineering, Indiana University, Bloomington, IN, United States.
- Aml&Regtech Team, Naver Financial, Seongnam, Republic of Korea.
- Department of Applied Artificial Intelligence, College of Computing & Informatics, Sungkyunkwan University, Seoul, Republic of Korea.
ORCID profiles: Sangpil Youm, Sou Hyun Jang, Haewoon Kwak, Jaeyoung Choi, Yong Jeong Yi
Cite
APA
Youm, S., Jang, S. H., Kwak, H., Choi, J., & Yi, Y. J. (2026). Cross-Cultural Differences in Public Discourse on COVID-19 Vaccination in the United States and South Korea: Cross-Sectional Analysis Using Natural Language Processing. JMIR Formative Research, 10, e84791. https://doi.org/10.2196/84791
Vancouver
Youm S, Jang SH, Kwak H, Choi J, Yi YJ. Cross-Cultural Differences in Public Discourse on COVID-19 Vaccination in the United States and South Korea: Cross-Sectional Analysis Using Natural Language Processing. JMIR Formative Research. 2026;10:e84791. doi:10.2196/84791
BibTeX
@article{Youm2026,
author = {Youm, Sangpil and Jang, Sou Hyun and Kwak, Haewoon and Choi, Jaeyoung and Yi, Yong Jeong},
title = {Cross-Cultural Differences in Public Discourse on COVID-19 Vaccination in the United States and South Korea: Cross-Sectional Analysis Using Natural Language Processing},
journal = {JMIR Formative Research},
year = {2026},
volume = {10},
pages = {e84791},
doi = {10.2196/84791}
}
EUVIMED