Sentiment analysis aims to extract subjective information, such as opinions and feelings, from natural language texts. This paper presents the comparison of different approaches to sentiment analysis in the context of the Portuguese language in order to make these results available in the literature to assist researchers in their future work. As an object of application of this study, tweets related to the volleyball theme were used, and a database was organized that has 2,330 tweets divided into 1,032 negatives and 1,298 positives. Lexical, committee and machine learning approaches were used. The committee approach in the Stacking model formed by the Decision Tree, Naive Bayes and Support Vector Machine algorithms as level-0 and the Logistic Regression algorithm as level-1 obtained the best performance with approximately 82% Accuracy, Precision, Recall and F1 Score in prediction of texts.

--

Sentiment analysis,Analysis of tweets,Natural language processing

• Ciências da Computação e da Informação - Ciências Naturais