Arabic Sexist Comments Detection in Youtube: A Context-Aware Opinion Analysis Approach

Abstract

In this chapter, we present an approach to automatize the assessment of attitudes toward violence against women and women’s rights, by analyzing Youtube comments, written in Arabic. More specifically, we propose a context-aware approach to opinion analysis in comments by taking into account the polarity of videos to which comments are associated. We build a training set and use it to train a classifier that predicts videos’ polarity. The accuracy and precision of the produced video classifier are 98 % and 94 %, respectively.