D-BullyRumbler: a safety rumble strip to resolve online denigration bullying using a hybrid filter-wrapper approach

Abstract

Denigration is a specialized form of cyberbullying which describes a recurrent, sustained and intentional attempt to damage the victim’s reputation or ruin the friendships that he or she has by spreading unfounded gossip or rumors online. It is the most common bullying tactic involving character assassination of public figures like celebrities and politicians. As a comprehensive approach to match to the scale of social media this research put forwards a D-BullyRumbler model for automatic detection and resolution of denigration cyberbullying in online textual content using a hybrid of lexicon-based and machine learning-based techniques. The model processes textual, content-based and user-based features to uncover denigration from two perspectives. Firstly, a direct explicit content analysis is done to look for denigration markers as features for model training and testing. Concurrently, potentially harmful messages, rumors, are identified as candidates and examined for target profile type to reveal the case of denigration. An additional OR operation is done to maintain the holistic framework. Another novelty of the work includes the use of hybrid filter-wrapper method, Chi-square filter and cuckoo search wrapper algorithm to improve the performance of reputation rumor classification module. Experimental results on social media datasets show the superior classification performance. The results validate the effectiveness of the proposed model which facilitates timely intervention by buzzing an alarm to the moderators and further forming a rumble safety strip to inhibit the production and dissemination of inappropriate content to protect the victims.