Deep Learning-Based Sentiment Analysis of COVID-19 Pfizer Vaccine Tweets Using Transformer and Bi-LSTM Architectures

Abstract

Starting in December 2019, the COVID-19 virus has impacted economies globally, infected billions of people worldwide, and created a global health disaster. The discovery of vaccines against SARS-CoV-2, the virus responsible for COVID-19, has proven safe and successful in combating the epidemic. As of July 2021, there were 184 vaccine candidates in preclinical development, 105 in clinical testing, and 18 vaccinations that were authorized for use in emergencies. These efforts represent the hard work of the scientific community in combating the pandemic. Language processing tactics can be used for the guidance of health communication strategies and to reduce misinformation. This investigation focuses on the emotion analysis of Pfizer vaccines using data from the Twitter platform based on different deeplearning methods and transformers. The dataset used in this study includes 11,021 tweets from the Twitter platform, collected from Kaggle, related to the Pfizer and BioNTech vaccines. The survey analyzes recent tweets to find out what people are saying about the Pfizer and BioNTech vaccines. The database was divided based on tweets related to Pfizer vaccines, which were categorized using DL frameworks. The sentiment distribution provides an overview of the opinions on positive, negative, and neutral comments. This can be represented using graphical and chart representations such as word clouds, ROC curves, and precision-recall curves. Deep learning models are employed for sentiment analysis, including Transformers and Bi-LSTM models.Various models, including DistilBERT, Google Electra-base, BiLSTM, and other Transformers, were utilized for this analysis. the results are reported using metrics such as accuracy, F1-score, and other evaluation metrics. The sentiment analysis results from the models show that Model DistilBERT outperformed the others in both accuracy (0.92) and F1-score (0.91), as depicted in the bar chart, where DistilBERT had the highest performance across all models. Such analyses will help healthcare providers, policymakers, and the general public understand the overall sentiment of Pfizer vaccines