FuzzySentClass: Interval-valued fuzzy approach to the Sentiment Analysis Problem via SentiWordNet

Abstract

Sentiment analysis, especially social network analy sis (SNA), is a relevant research area. In recent years, this domain has become an active research question in data mining, natural language processing, and sentiment analysis (opinion mining). It consists of analyzing and extracting emotions, opinions, or attitudes from reviews of products, services, music, and movies, classifying them into positive, neutral, and negative, or even extracting the degree of importance (polarity). In this article, we propose a new approach using the Interval-valued Fuzzy Logic called FuzzySentClass to classify tweets based on lexicon using SentiWordnet. Our approach consists of classifying tweets according to three classes: positive, neutral, and negative, apply ing the FuzzySentClass that considers the steps of Fuzzification, Inference, and Defuzzification of an interval-valued fuzzy system. For the stage of obtaining results, the Juzzy platform was considered. The obtained results are evaluated based on the accuracy of the classifications obtained in the executions varying the type reducer in the FuzzySentClass Defuzzification step. In addition, the interval entropy approach is used to measure the imprecision information of achieved results. Our approach reached an accuracy of 83.22 with the centroid type reducer and 82.63 with the center of sets type reducer. And, resulting on the values of 0.117469 and 0.149853 as the maximum diameter of interval entropy for IvFS related to input and output variables, respectively. Index Terms—Interval-valued Fuzzy Logic, Interval-valued Fuzzy Sets, Sentiment Analysis, Interval Entropy