eXplainable Artificial Intelligence - A Study of Sentiments About Vaccination in Brazil

Sentiment analysis in social networks is a focus in several studies on Machine Learning, this happens because the scope and speed with which opinions and emotions about events, controversial issues and products and services are treated on the Internet make it attractive to analyze this medium to obtain relevant information and of interest. Based on this context, this paper presents a sentiment analysis on social networks, focusing on Twitter, about the COVID-19 vaccination campaign in Brazil, using Machine Learning techniques, more specifically, logistic regression, and subsequently the eXplainable Artificial Intelligence (XAI) with the methods LIME, SHAP and Eli5 to interpret the model output. Although there are several applications in the field of sentiment analysis, this study focuses on using real Twitter data, extracted according to the desired context, for five months, processing, analyzing and preparing them for training, and on the explainability of the results obtained during the analysis. The results obtained show that the sample population was mostly in favor of vaccination for issues such as health and the collective good of the population, while those who were against wondered about compulsion and the power of freedom of choice, and expressed fear of being part of an experiment, given the design time of vaccine development.