Academics develop innovative technology to tackle cyberbullies

Academics develop innovative technology to tackle cyberbullies

University of Wolverhampton academics are teaching computers to identify early warning signs of abusive behaviour online in a bid to outsmart cyberbullies.

Innovative ways to use technology to tackle abusers are being devised by the University’s Computational Linguistics Research Group.

Through automatic learning of features, clues, and patterns of cyberbullying related contents using natural language processing and machine learning computers will be able to recognise signs of abuse, which users and moderators can act upon.

Dr Vinita Nahar, Post-Doctoral Research Fellow in Natural Language Processing, is particularly focusing on understanding swearing use in Twitter.

She said it was a challenge to differentiate between expressing opinion and abuse of others but the project sought to examine these issues and establish the sentiment of the tweet.

“We are interested in developing predictive models to identify emerging cyberbullying activities in social networks, and identify the most influential bullies and the most offended victims,” said Dr Nahar.

“When bullying takes place in a virtual space, it is difficult for adults to monitor the actions of young children/students. Having an automatic system in place that can detect early signs of suspicious, abusive and offensive messages, and which can alert users or a moderator of its repetitive occurrence, can enable better prevention of potentially tragic outcomes, including suicide attempts by the cyberbullying victims.

“Cyberbullying detection has a positive impact on people’s lives.”

After the research is complete the model can be applied by teachers, parents, social media platforms and anywhere else where there is a potential risk of abuse. A pilot scheme is being planned with regional secondary schools.

Ends

 For media inquiries please contact the Media Relations Team on 01902 322736 or email emmap@wlv.ac.uk<mailto:emmap@wlv.ac.uk>.