A new algorithm that recognizes bullies and online attackers has been developed by a group of researchers from the University of Binghamton. Specifically, the researchers developed an algorithm that, with an accuracy of 90% according to the press release, recognizes the bullies on Twitter.
More and more IT and Artificial Intelligence laboratories and researchers are spending their time trying to develop methods for automatically recognizing bullying and aggression on the internet, in order, quite clearly, to benefit even large companies, they keep the social networks that, at least for the time being, mostly use human moderators.
Jeremy Blackburn, an American university computer scientist, is trying to bridge this gap by analyzing the behavioral patterns of “bullies” on Twitter and comparing them to those of “normal” users. It is precisely for this reason that the researcher, together with his colleagues, has created special crawlers to collect data from Twitter faster and more efficiently.
He then relied on natural language processing algorithms and other tools already available for social network analysis and was able to develop an algorithm that automatically classifies two models of offensive behavior online: cyberbullying and cyber aggression.
The accuracy of the algorithm would be 90%. The algorithm is able to identify tricky behavior, for example users who launch threats make racist comments to other users.