Detection of Cyber-Aggressive Comments on Social Media Networks

Abstract

The spread of aggressive tweets, status and comments on social network are increasing gradually. People are using social media networks as a virtual platform to troll, objurgate, blaspheme and revile one another. These activities are spreading animosity in race-to-race, religion to religion etc. So, these comments should be identified and blocked on social networks. This work focuses on extracting comments from social networks and analyzes those comments whether they convey any blaspheme or revile in meaning. Comments are classified into three distinct classes; offensive, hate speech and neither. Document similarity analyses are done to identify the correlations among the documents. A well defined text pre-processing analysis is done to create an optimized word vector to train the classification model. Finally, the proposed model categorizes the comments into their respective classes with more than 93% accuracy.