Classical Machine Learning vs Deep Learning for Detecting Cyber-Violence in Social Media

Cyber-violence is a largely addressed problem in e-health researches, its focus is the detection of harmful behavior from the online user-generated text in order to prevent and protect victims. In this work, we tackle the problem of Social Media (SM) text analysis to detect the harmful content that is the common characteristic of cyber-violence acts. For that, we use classical Machine Learning (ML) based on user psychological features that we compare with Deep Learning (DL) techniques in a small dataset setting. The results were in favor of classical ML. The findings highlight that psychological characteristics extracted from user-generated text are strong predictors of his harmful behavior. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.