Covid-19 Vaccine Sentiment Analysis During Second Wave in India by Transfer Learning Using XLNet

Abstract

The Covid-19 pandemic has created a world-wide crisis from the perspectives of health and economy. Vaccination is one of the prime means by which herd immunity could be developed. Social media platforms such as Twitter has played a major role in building public opinion as the vaccination drive got underway in several countries. In this paper, we present a tweet-based sentiment analysis of the two popularly administered vaccines in India Covishield and Covaxin during the second wave of the pandemic in India, from March 2021 to September 2021, which was attributed to the Delta mutant of the coronavirus. We use unlabeled Covid-19 vaccine-related tweets downloaded from a large-scale dataset from March 2021 to September 2021, and employ transfer learning for classifying the unlabeled tweets. The contributions of this paper are: - sentiment analysis of unlabeled vaccine-related tweets by training a transformer model on pre-trained XLNet (transformer) features derived from a labeled non-Covid Twitter dataset, a time-line of public sentiments for the two vaccines administered in India, and word clouds of high-frequency adjective unigrams after sentiment analysis, as evidence.