Citation

Salawu S, Lumsden J, He Y. Int. J. Bullying Prev. 2022; 4(1): 66-88.

Copyright

(Copyright © 2022, Holtzbrinck Springer Nature Publishing Group)

DOI

10.1007/s42380-021-00115-5

PMID

unavailable

Abstract

A negative consequence of the proliferation of social media is the increase in online abuse. Bullying, once restricted to the playground, has found a new home on social media. Online social networks on their part have intensified efforts to tackle online abuse, but unfortunately, such is the scale of the problem that many young people are still regularly subjected to a wide range of abuse online. Research in automated detection of online abuse has increased considerably in recent times. However, existing studies on online abuse detection typically focus on developing newer algorithms to improve predictions, and little research is done on developing impactful tools that leverage these algorithms to tackle online abuse. In this paper, we present BullStop, a mobile application that can use different machine learning models to detect cyberbullying. A new cyberbullying dataset containing 62,587 tweets annotated using a taxonomy of different cyberbullying types was created to facilitate the classifier's training. BullStop was developed using a participatory and user-centred design approach involving young people, parents, educators, law enforcement and mental health professionals. Additionally, the application incorporates online training for the ML models using ground truth supplied by the user as additional training data, and in this way, it can create a personalised classifier for each user. Furthermore, on detecting online abuse, the application automatically initiates punitive actions such as deleting offensive messages and blocking cyberbullies on behalf of the user. BullStop is freely available on the Google Play Store and has been downloaded by hundreds of users.

Language: en

Keywords

Cyberbullying; Machine learning; Mobile application; Offensive language detection; Social media