NLP Based Hate Speech Detection And Moderation

As Social media penetration increases in day-to-day life so does the growth of hate speech. After 2016 due to affordability of internet many user on boarded the internet which not only increased social media interactions but also growth of hate speech. be it religion phobia, homophobia, gender, toxicity etc. To control hate speech, NLP based machine learning model has been proposed. The proposed model uses TFIDF feature generation method to which binarized naive bayes is applied to calculate maximum likelihood feature for each labels. Model has been trained with logistic regression on the maximum likelihood feature. which helps to classify hate speech and moderate accordingly. The proposed model produces an overall accuracy of 83 percent and able to achieve 5 percent improvement compared to multinomial naive bayes' production in identifying hate speech and classifying it under various labels.