TextBlob and BiLSTM for Sentiment analysis toward COVID-19 vaccines

Abstract

Nowadays, social media like Twitter, play a vital role in our life since it is a source of swapping views, thoughts, and feelings towards many issues such as the global pandemic covid-19. Nevertheless, it can a source of diffusion of fake news which can affect negatively the opinions of many people and even change their thoughts behind a lot of sensitive situations such as the COVID-19 vaccines. In this context, it is crucial for public health agencies to understand and identify people's opinions and views toward COVID-19 vaccines. To this end, we propose our model to classify the tweets of people into three classes, negative, neutral, and positive. In fact, we considered a large dataset extracted from Twitter includes 174490 tweets. Tweet analysis was conducted by TextBlob to categorize the sentiment and the Bidirectional LSTM model to classify the sentiments. The proposed model was compared with other studied machine learning classifiers and deep learning algorithms. The aim of this work also is to select the best model between the studied model that is suitable for the sentiment analysis for COVID-19 vaccines. BiLSTM outperformed the other studied models with ahigh accuracy rate of 94.12%.