AI to prevent cyber-violence: harmful behaviour detection in social media

摘要

Social media has allowed people to communicate freely. This total freedom has led to the emergence of cyber-violence with a growing number of victims. Many researches in psychology and e-health have been conducted to detect the act of cyber-violence. In computational field, most of works have focused on multiple aspects of cyber-violence, but none of them, to our knowledge, have studied the perpetrator's harmful behaviour from an emotional dimension. Our goal in this work is to discover the relationship between the emotional state of social media users and their harmful behaviour while engaged in the act of cyber-violence. Our approach is based on Ensemble Machine Learning and engineered features related to Plutchik wheel of basic emotions extracted with semantic similarity and word embedding. The results show a significant association between the individual's emotional state and the harmful intent, which may be a good indicator for cyber-violence detection.