Vulgar and spam comment identification using Gluon Natural Language Processing and Convolution Neural Networks

Abstract

With the advancement of technology, the virtual platform and social media have become an important part of people’s daily life. The social media allows users to communicate, to express their feelings and to discuss on various topics. But cyber bullying and vulgar or toxic comments become an alarming problem in social media. In this paper, a system has been proposed to detect and classify vulgar comments using convolutional neural networking with gluon natural language processing. The system can classify toxic, hate speech, insult, obscene, threat and normal text by training from a huge dataset obtained from Kaggle, extracting comments from YouTube, facebook through extension of browser. The system shows 95.4% accuracy to detect and classify the vulgar comments.