Optimized Twitter Cyberbullying Detection based on Deep Learning

Cyberbullying is a crime in which a perpetrator targets a person with online harassment and hate. Many cyberbullying detection approaches have been introduced, but they were largely based on textual and user features. Most of the research found in the literature aimed to improve detection by introducing new features. Although, as the number of features increases, the feature extraction and selection phases become harder. In addition, another drawback of such improvements is that some features-for example, user age-can be easily fabricated. In this paper, we propose optimised Twitter cyberbullying detection based on deep learning (OCDD), a novel approach to address the above challenges. Unlike prior work in this field, OCDD does not extract features from tweets and feed them to a classifier; rather, it represents a tweet as a set of word vectors. In this way, the semantics of words is preserved, and the feature extraction and selection phases can be eliminated. As for the classification phase, deep learning will be used, along with a metaheuristic optimisation algorithm for parameter tuning.