Presentation description

This study examines how public responses to COVID-19 crisis communication varied based on the message source (government officials vs. politicians) and topic (mask mandates, vaccine efficacy, economic policies). Twitter played a crucial role in shaping public discourse during the pandemic, and this research categorizes 2,000 replies to official COVID-19 tweets into discussion-based (logical and fact-driven) and reaction-based (emotionally charged) responses.

Using machine learning models—including AdaBoost, Neural Networks, Decision Trees, and Gaussian Processes—this study classifies responses, with Gaussian Processes achieving the highest accuracy (74.6%). Additional linguistic features (LIWC cognitive processing and social process scores) and LLaMA NLP techniques were incorporated to refine classification, identifying financial and statistical references in tweets.