Supervised and Traditional Term Weighting Methods for Sentiment Analysis

Abstract

Sentiment analysis is a text classifying problem and because of its popularity and commercial revenue, it has been widely studied. The most important point in text categorization is how to represent the texts. Instead of traditional methods, supervised term weighting methods which include terms' distribution of classes has been started to be used. In this study, these methods are compared in different dimensions on two datasets which consist Turkish Twitter posts. In conclusion, supervised term weighting methods are found more successful and applicable.