Identification of Online Abuse and It’s Inhibition

Abstract

Online abuse is an act of attacking an individual repeatedly with an intent to harm. This has a very disturbing effect on many individual irrespective of the age group. In this paper, we propose a new representation learning method to solve this problem. Our method named Naïve Bayes classifier is a simple probabilistic classifier based on Bayes’ theorem with naïve independence assumptions between the features. Naive Bayes is an highly scalable with different number of parameters linear in number of predictors in a learning problem. The proposed method is able to exploit the hidden feature structure of abusive information and learn a robust and discriminative representation of text. We have implemented our algorithm using five lakhs of tweets and around one thousands of users.