eXplainable Artificial Intelligence in sentiment analysis of posts about Covid-19 vaccination on Twitter

Considering the impact of the use of Artificial Intelligence (AI) in the most diverse branches of society and the use of eXplicable Artificial Intelligence (XAI) to improve the interpretability of these intelligent models, this paper aims to analyze some existing XAI methods to verify their effectiveness. To this end, experiments were conducted with LIME, SHAP, and Eli5 solutions in a scenario of sentiment classifications in Twitter posts about the Covid-19 vaccination process in Brazil. Thus, it is observed that the tools provide relevant information about the aspects that interfere in the classification of tweets as favorable or not favorable to vaccination, which allows concluding that the methods bring the necessary transparency to confirm the AI decisions regarding the sentiments related to the vaccination process in Brazil.