A Framework for Efficient Multilevel Polarity-Based Sentiment Analysis Using Fuzzy Logic

Abstract

The evaluation of any product or event on social media with the opinion or emotion of peoples is known as sentiment analysis (SA). A great deal of attention has been attracted in recent years, toward both science and industry fields for a variety of uses. Machine learning and the text mining uses this most widely known application area of sentiment analysis. This paper presents a framework for efficient multilevel sentiment analysis using fuzzy logic for the classification of online test reviews polarity as strong positive, positive, negative and strong negative. This proposed model can use the fuzzy logic classifier to enhance the degree of sentiment polarity of reviews. Here, fuzzy logic classifier is used for finding the sentiment classes. This also utilizes the mechanism of imputation of missing sentiment for integrating non-opinionated sentences in generating precise results. Results show that the proposed method has a capability of extracting opinions and classify them in an effective way. The proposed method has a capability to predict the degree of sentiment polarity for the reviews on a social media. The better precision and F1-scores are obtained for an objective/subjective classification and polarity (positive/negative) classification on twitter dataset.