ABSTRACT
Nowadays, social media is crucial for informing people about global issues such as the current COVID-19 pandemic. People have expressed different perspectives on the COVID-19 vaccine, and Twitter has proven to be an excellent medium for sharing these. The purpose of this research is to propose a text vectorized neural network (NN) model and compare it with long short-term memory (LSTM), and bidirectional long short-term Memory (BiLSTM) to analyse sentiment on COVID-19 vaccine twitter data. The raw Twitter data is collected and three different raters annotate the data as positive or negative. The label is finalized using the kappa value. Then, Natural Language Processing (NLP) methods are used to process the twitter data. The study concludes that the proposed text vectorized NN model outperforms other models in terms of accuracy, as it achieves 81% in the test dataset, while LSTM and BiLSTM obtain 75% and 74%, respectively. The text vectorized NN model receives a 51% score in Matthews' correlation coefficient (MCC), while LSTM and BiLSTM acquire 37% and 39% scores. Other performance metrics, such as precision, recall, f1-score, and confusion matrix were also used to validate the models more effectively. © 2022 ACM.