Learning to Classify Hate and Extremism Promoting Tweets

Research shows that Twitter is being misused as a platform for online radicalization and contains several hate and extremism promoting users and tweets violating the community guidelines of the website. Manual identification of such tweets is practically impossible due to millions of tweets posted every day and hence solutions to automate the task of tweet classification is required for Twitter moderators or an intelligence and security analyst. We formulate the problem of hate and extremism promoting tweet identification as a one-class classification problem and propose several linguistic features. Experimental results on large and real-world dataset demonstrate that the proposed approach is effective.