Variants of Naïve Bayes Algorithm for Hate Speech Detection in Text Documents
Vijay Vijay, Pushpneel Verma
Abstract
Social media provides an inexpensive way to communicate with millions of users. These websites provide a platform to propagate your ideas to millions of people. The power of social media is often misused. Freedom of expressing opinions and beliefs on social media websites has been resulted in spread of hate speech. It has become a major challenge to check the dissemination of hate speech on websites of social media. We used Naive Bayes algorithm to classify text tweets into three classes i.e., tweets containing hate speech, tweets containing offensive language and tweets containing neither hate speech nor offensive language. It is a supervised machine learning based algorithm. We performed experiments with three variations of Naive Bayes i.e., Bernoulli's, multinomial and Gaussian Naive Bayes classifier. Bernoulli’s and Multinomial Naïve Bayes delivered the best accuracy values.