Sentiment analysis for COVID-19 vaccination news

Abstract

The coronavirus disease, also known as COVID-19, has affected our daily lives hugely. In this technological era, social media platforms play a vital role to spread information regarding this pandemic across the world, as people express their feelings via social networks. With the availability of vaccines, Singapore has rolled out the vaccination program where Singaporeans and long-term residents are hugely encouraged to get vaccinated as part of the preventive measure. People have shared their opinions about vaccinations on social networking sites like Facebook. This project aims to understand public sentiments regarding various vaccines in Singapore and predict message sentiment. The WKW school provided COVID-19 Facebook posts in Singapore was cleaned before sentiment analysis.

Different sentiment analysis models, TextBlob, VADER, Flair and CT-BERT, are used. Several rules are defined through analysis to improve the models' accuracy. The TextBlob model accuracy increased from 44% to 52.67%, the VADER model accuracy increased from 50.67% to 58.33%, the Flair model accuracy increased from 72.97% to 79.05%, and the CT- BERT model accuracy increased from 61.49% to 80.41%. The results show that with rules, the accuracy of the models improves significantly. The improved CT-BERT model outperforms the rest and is used for analysis of various vaccines. A website is developed to display the result. With the development of the website, individuals can easily understand public sentiments about varying vaccines in Singapore, helping to uncover the reasons people refuse to take the vaccine. Users can also input messages on the website and discover its sentiment.