Automatic cyberbullying detection: A systematic review

Automatic cyberbullying detection is a task of growing interest, particularly in the Natural Language Processing and Machine Learning communities. Not only is it challenging, but it is also a relevant need given how social networks have become a vital part of individuals' lives and how dire the consequences of cyberbullying can be, especially among adolescents. In this work, we conduct an in-depth analysis of 22 studies on automatic cyberbullying detection, complemented by an experiment to validate current practices through the analysis of two datasets. Results indicated that cyberbullying is often misrepresented in the literature, leading to inaccurate systems that would have little real-world application. Criteria concerning cyberbullying definitions and other methodological concerns seem to be often dismissed. Additionally, there is no uniformity regarding the methodology to evaluate said systems and the natural imbalance of datasets remains an issue. This paper aims to direct future research on the subject towards a viewpoint that is more coherent with the definition and representation of the phenomenon, so that future systems can have a practical and impactful application. Recommendations on future works are also made.