Agenda-setting effects for covid-19 vaccination: Insights from 10 million textual data from social media and news articles using BERTopic

Abstract

This study investigates the agenda-setting effects of media in the context of the COVID-19 vaccination by leveraging a cutting-edge machine learning framework, BERTopic, to analyze over 10 million textual data points from social media and news articles. The research highlights a significant divergence between public opinion, primarily expressed on Twitter, and the media agenda, challenging traditional agenda-setting theories in public health crises. Specifically, while public discourse centered on vaccination-related concerns and negative sentiments toward vaccination policies, media coverage diversified to include topics such as politics, foreign affairs, and economics. The proposed framework systematically integrates data collection, preprocessing, and advanced topic modeling to enhance interpretability and efficiency. By adopting BERTopic, this study advances beyond traditional Latent Dirichlet Allocation (LDA) models by offering superior clustering and contextual understanding of unstructured text data. The framework demonstrates its utility in identifying actionable insights for public health practitioners, policymakers, and information systems researchers, providing a robust methodology to track and evaluate public sentiment and media narratives during health crises. Ultimately, this study emphasizes the critical need to align media messaging with public concerns to improve vaccination campaigns and public health communication. It contributes to the theoretical understanding of agenda-setting in the digital era while offering practical guidelines for leveraging social big data in multidisciplinary applications.