Machine learning and Lexical Semantic-based Sentiment Analysis for Determining the Impacts of the COVID-19 Vaccine

ABSTRACT

In 2020 COVID-19 has taken the world by storm. Scientists from around the world are still working to develop a more effective vaccine for this disease. AstraZeneca, Moderna, Sputnik V and Comirnaty (Pfizer) are just a few of the vaccines that have been developed and are now being used by large populations. Social media is a powerful tool for people to express their opinions on current events, such as COVID-19 and its vaccine. It is highly noticeable that people are becoming increasingly concerned about the availability and effectiveness of these vaccines and other remedies for COVID-19. Healthcare organizations and professionals can acquire useful insights into vaccination safety by evaluating people's sentiments. Furthermore, it can also assist to prevent unnecessary panic and the spread of misinformation among people. In this paper, a comprehensive analysis of people's sentiments regarding the vaccination against COVID-19 is shown. Twitter's data regarding the vaccine for COVID-19 from January to December of 2020 was collected from Kaggle for analysis. Necessary preprocessing techniques have been used to prepare and label the data based on textual sentiment using the lexical semantic methods: TextBlob and VADER. Various machine learning methods like Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), merged model (RNN+CNN) and Logistic Regression have been used to analyze the public sentiments and to visualize their concerns regarding the vaccination against COVID-19 throughout 2020. Then, the results from both TextBlob and VADER were compared in order to obtain the highest possible accuracy and to better understand the reasons for them. © 2021 IEEE.