Covid-19 Tweets Sentiment Analysis with Latent Dirichlet Allocation Topic Modeling

Abstract

Analysis of Covid-19  vaccine tweets has been an extensive focus in understanding user trends throughout the pandemic. This project concentrated on the development of a Latent Dirichlet Allocation (LDA)  model along with sentiment analysis to better understand different trends and patterns which have arisen temporally. In addition, the presence of adverse events within the tweet data set was compared with the Vaccine Adverse Event Reporting System (VAERS) COVID-19 World Vaccine Adverse Reactions data to see if there were any distinctions between the reported events. It was discovered that there were distinct peaks in subjectivity and polarity throughout time and a nine-topic LDA model was constructed with the highest coherence score. Topics within the constructed dataset were seen to be diverse varying in subjects such as adverse events. It was also observed that there were notable distinctions in the adverse events reported in the VAERS dataset compared to the tweet data set.