Sentiment analysis on COVID-19 tweets

Abstract

The global proliferation of COVID-19, as well as the growth of platforms as a primary source of information for many individuals, has elicited a wide spectrum of reactions. However, keeping up with this mass scenario is difficult. A large number of people offer their opinions and perspectives on current events on social media, making it difficult for a human to read and comprehend everything. Tweets disseminate a great deal of information. Using public Twitter comments, our study attempted to conduct a sentiment analysis of the entire discourse about COVID-19 in a paper. We improved on earlier methodologies and methods for sentiment analysis. Our primary goal is to examine tweets concerning COVID-19 from the previous year using natural language processing and neural network methods. We used a multiclass dataset and applied it to BOW, TFIDF, and One Hot Encoding. Furthermore, after training four distinct classifiers with these different pre-processing algorithms in each classifier, we performed a competitive analysis and found RoBERTa as the best performing with 90% accuracy.