Mining Public Opinion on COVID-19 Vaccines using Unstructured Social Media Data

The emergence of the novel coronavirus (COVID-19), and the necessary separation of populations led to an unprecedented number of new social media users seeking information related to the pandemic. Nowadays, with an estimated 4.5 billion users worldwide, social media data offer an opportunity for near real-time analysis of large bodies of text related to disease outbreaks and vaccination. This study investigated and compared public discourse related to COVID-19 vaccines expressed on two popular social media platforms, Reddit and Twitter. Approximately 9.5 million Tweets and 70 thousand Reddit comments were analyzed from dates January 1, 2020, to March 1, 2022, and analyzed through topic modeling, sentiment analysis, and semantic network analysis. Sentiment analysis through the fine-tuned DistilRoBERTa model revealed that even though Twitter content was overall more negative than content expressed on Reddit, relatively similar changes in sentiment occurred among users of both online platforms. Reversals in sentiment trends typically occurred within relative proximity to events such as vaccine development news, vaccine release, frequent discussion of side-effects, the discovery of new variants, and pandemic fatigue. Topic modeling and semantic network analysis provided insight into how public discourse related to COVID-19 and vaccinations, misinformation, and vaccine hesitancy evolved over 26 months. Though misinformation and mention of conspiracy theories were detected with the analysis, the occurrence of both was less frequent than expected. This work provides a framework that could be scaled and utilized by public health officials to monitor disease outbreaks in near real-time in large communities as well as smaller local groups. Hopefully, the results from this study will help to guide and facilitate the implementation of targeted digital interventions among vaccine-hesitant populations and provide insights to public health officials to inform decision-making and effective policy development