A novel representation for classification of user sentiments

Abstract

In this paper we present a term sequence preserving representation for text documents called as label matrix for classification of opinions. This is an efficient yet effective representation in polarity of opinions with the use of only three parts of speech feature set viz verb, adverb and adjective. To draw out the efficiency of our proposed technique we led experimentation on one publically accessible extremity audit dataset furthermore our own made motion picture survey and item survey datasets. We have explored quantitative comparative analysis between existing classifiers and proposed method.