DENOISING ENCODER WITH SEMANTICS AND EXCLUSION FOR SPOTTING CYBERBULLY

The rapid growth of social networking is supplementing the progression ofcyberbullying activities. Most of the individuals involved in these activities belong to theyounger generations, especially teenagers, who are at more risk of suicidal attempts.Cyberbullying is the process of using the Internet, cell phones, or other devices to send orpost text or images intended to hurt or embarrass another person. Through machine learningtechniques, we can detect language patterns used by bullies and their victims, and developrules to automatically detect cyberbullying content. Here, we introduce a new machinelearning method to deal with this problem. Our method named Semantic-EnhancedMarginalized Stacked Denoising Auto-Encoder (smSDA) is developed via a semanticextension of the popular deep learning model. The smSDA method detects the hiddenattributes of the bullying information. Our approach experiments on two public cyberbullyingcorpora i.e. twitter and MySpace. The outcome of our proposed method is better than theother text representation learning methods.