Prediction of Aggressive Comments in Social Media: an Exploratory Study

Abstract

This paper presents a set of techniques for predicting aggressive comments in social media. In a time when cyberbullying has, unfortunately, made its entrance into society and Internet, it becomes necessary to find ways for preventing and overcoming this phenomenon. One of these concerns the use of machine learning techniques for automatically detecting cases of cyberbullying; a primary task within this cyberbullying detection consists of aggressive text detection. We concretely explore different computational techniques for carrying out this task, either as a classification or as a regression problem, and our results suggest that a key feature is the identification of profane words.