What drives popularity and virality of COVID-19 vaccine discourse on Twitter: Insights from text mining and network visualization.

Zhang, Jueman Mandy; Wang, Yi Jasmine; Shi, Molu; Wang, Xiuli

BACKGROUND COVID-19 vaccination is considered as a critical prevention measure to help end the pandemic. Social media such as Twitter has played an important role in public discussion about COVID-19 vaccines. OBJECTIVE This study intended to investigate message-level drivers of the popularity and virality of tweets about COVID-19 vaccines using machine-based text mining techniques. It also examined the topic communities of the most liked and most retweeted tweets using network analysis and visualization. METHODS We collected US-based English-language public tweets about COVID-19 vaccines from January 1, 2020 to April 30, 2021 (n=501,531). Topic modeling and sentiment analysis were used to identify latent topics and valence, which together with auto-extracted information about media presence, linguistic features, and account verification were used in regression models to predict likes and retweets. Among the 2,500 most liked tweets and most retweeted tweets respectively, network analysis and visualization were used to detect topic communities and present the relationship between the topics and the tweets. RESULTS Topic modeling yielded 12 topics. The regression analyses showed that eight topics positively predicted likes and seven topics positively predicted retweets, among which, the topic of vaccine development and people's views and that of vaccine efficacy and rollout had relatively larger effects. Network analysis and visualization revealed that the 2,500 most liked and most retweeted retweets were clustered around the topics of vaccine access, vaccine efficacy and rollout, vaccine development and people's views, and vaccination status. The overall valence of tweets was positive. Positive valence increased likes, but valence did not affect retweets. Media presence and account verification increased likes and retweets. Linguistic features had mixed effects on likes and retweets. CONCLUSIONS The study suggests the public interest in and demand for information about vaccine development and people's views and that about vaccine efficacy and rollout. These topics, along with the use of media and verified accounts, have enhanced the popularity and virality of tweets. These could be addressed in vaccine campaigns to help the diffusion of content on Twitter. CLINICALTRIAL