A Large-Scale Analysis of COVID-19 Twitter Dataset in a New Phase of the Pandemic

Social media have been awash with news and discussions of the COVID-19 pandemic. It is phenomenal to observe that social media has been the focal venue for people to express their reactions, opinions, and interpretations of the pandemic, given the presence of mixed sources of real information and misinformation. Thus, it is essential to conduct professional assessments of the public views and their evolving nature. Our study aims to extract and assess insights into the reflections of sentiments and topics of the public on Twitter and their dynamics along the timeline of the Delta variant. It highlights the extraordinary influence Twitter, or similar major social media, would have on people to comprehend and decide how to cope with the pandemic. We present findings of extracted sentiments and topics from a large-scale dataset of COVID-related tweets collected for the recent phase of the Delta variant of the pandemic (July-September 2021). We utilized a variety of machine learning algorithms for topic modeling and testing the accuracy of sentiment analysis. Our study shows the dramatic dominance of a positive and objective sentiment rather than a negative and subjective sentiment as well as the shift of prevalent topics during the period of study. The findings indicate the importance of conveying real, rational, and accurate information instead of misinformation on social media to foster the public's awareness and preparedness for a major public emergency incident such as the pandemic.