A framework for early detection of antisocial behavior on Twitter using natural language processing

Online antisocial behavior is a social problem and a public health threat. A manifestation of such behavior may be fun for a perpetrator, however, can drive a victim into depression, self-confinement, low self-esteem, anxiety, anger, and suicidal ideation. Online platforms such as Twitter and Facebook can sometimes become breeding grounds for such behavior. These platforms may have measures in place to deter online antisocial behavior, however, such behavior still prevails. Most of the measures rely on users reporting to platforms for intervention. In this paper, we advocate a more proactive approach based on natural language processing and machine learning that can enable online platforms to actively look for signs of antisocial behavior and intervene before it gets out of control. By actively searching for such behavior, social media sites can possibly prevent dire situations that can lead to someone committing suicide.