Using stylometric features for sentiment classification

Abstract

This paper is a comparative study about text feature extractionmethods in statistical learning of sentiment classification. Featureextraction is one of the most important steps in classification systems.We use stylometry to compare with TF-IDF and Delta TF-IDF baselinemethods in sentiment classification. Stylometry is a research areaof Linguistics that uses statistical techniques to analyze literary style.In order to assess the viability of the stylometry, we create a corpus ofproduct reviews from the most traditional online service in Portuguese,namely, Buscapé. We gathered 2000 review about Smartphones. We usethree classifiers, Support Vector Machine (SVM), Naive Bayes, and J48to evaluate whether the stylometry has higher accuracy than the TFIDFand Delta TF-IDF methods in sentiment classification. We foundthe better result with the SVM classifier (82,75%) of accuracy with stylometryand (72,62%) with Delta TF-IDF and (56,25%) with TF-IDF.The results show that stylometry is quite feasible method for sentimentclassification, outperforming the accuracy of the baseline methods.We may emphasize that approach used has promising results.