Twitter Sentiment Polarity Classification using Barrier Features

Abstract

English. A crucial point for the applicability of sentiment analysis over Twitter is represented by the degree of manual intervention necessary to adapt the approach to the considered domain. In this work we propose a new sentiment polarity classifier exploiting barrier features, originally introduced for the classification of textual data. Empirical tests on SemEval2014 competition data sets show that such approach overcomes performance of baseline systems in nearly all cases.