A Hierarchical Topic Modeling Analysis of Twitter Data on COVID-19 Vaccines

This study employs hierarchical topic modeling (HTM) to analyze the discourse surrounding COVID-19 vaccines in Korea. We collected and preprocessed 123,041 Korean tweets mentioning major vaccine brands using the Twitter API, spanning from February 2021 to January 2023. Three hierarchical topic models were applied and compared on this dataset. To enhance the hierarchical understanding, the dataset was divided into four periods marked by significant vaccination events, and each HTM was applied to these subsets. Our findings show that hLDA consistently outperformed the other models across all periods. The topic hierarchy generated by the best-performing hLDA model reveals that public discourse in Korea primarily focuses on vaccine safety and effectiveness. These insights are expected to inform policy-making efforts aimed at increasing vaccine acceptance and improving public health outcomes.