Research pick: Detecting and blocking cyberbullying - "Improved cyberbully detection techniques using multiple correlation coefficient from forum corpus"

Bullying is as old as humanity, but in today’s world of ubiquitous and always-connected devices, there is a whole realm of bullying that can take place out of sight but be just as devastating to its victims – cyberbullying. Detecting and so having the opportunity to prevent cyberbullying in open online forums and social networking sites, for instance, requires technology that can automatically detect trollish and thuggish behaviour. Once detected, the problems that victims face might be addressed but more importantly, the cyberbullies might be shut down or otherwise punished.

Writing in the International Journal of Autonomic Computing, a team from India reveals their algorithm which detects and weighs the words in forums and calculates whether or not particular clusters of words are associated with cyberbullying behaviour.

The team explains the problem and why it matters so much: “Cyberbullying has emerged as a major problem along with the recent development of online communication and social media. Cyberbullying has also been extensively recognised as a serious national health problem, in which victims demonstrate a significantly high risk of suicidal ideation,” they write. They add that “This proposed framework shows better results while the action is to stop the online users becoming the victims of cyberbully.”

Sheeba, J.I., Devaneyan, S.P. and Tata, P. (2018) ‘Improved cyberbully detection techniques using multiple correlation coefficient from forum corpus’, Int. J. Autonomic Computing, Vol. 3, No. 2, pp.152–171.