A Study of Contact Network Generation for Cyber-bullying Detection

Cyber-bullying is widely recognized as a serious social problem, especially for adolescents and it is also becoming a threat to the viability of online social networks. Several researchers have proposed digital-communication based frameworks whose goal is to help detect bullying in social networks. In this paper, our study focuses on detecting relation-based cyber-bullying, which is an indirect attack on a human, e.g., isolating a victim by ignoring the victim's messages. Recently, relationship-based cyber-bullying has received attention as a new type of cyber-bullying, and detecting it is still a novel problem. As it attacks a human relationship, the detection should monitor the change of the human relationship. In this paper, for the first step of relation-based cyber-bullying detection, we propose a framework to generate a contact network. The framework consists of two phases for a reduction of false negative, i.e. students are friends in the school but detected as non-friend in the Social Networking Service (SNS), which is a serious problem for the cyber-bullying detection. Finally, this paper analyzes the collected SNS data with the actual human relationships, and evaluates the proposed framework.