Hate speech detection using DNN (deep neural network)

Abstract

Hate Speech is defined as offensive or threatening speech that expresses prejudice against a particular group of people on the basis of race, stereotype, epithet, threat, gender, sexual orientation, religion, organization, country. Through the availability of the Internet and Social media, anonymity has made hate speech hard to detect. For the detection of hate speech, the DNN (Deep Neural Network) model can be very effective. Also, there is no such research available that gives the best result when it comes to detecting hate speech from audio speech. In this paper, feature extraction of audio has been done with the help of many audio feature extraction methods- MFCCs, ZCR etc. The DNN (Deep Neural Network) deep learning method is used as it gives better accuracy in sequential data like audio. Moreover, for classification purposes, two different modern classifiers are used to classify the dataset that has been made. SVM(Support Vector Machine) and XGboost classification models are used in our dataset to compare results. From these three models, DNN(Deep Neural Network) performs the best applying the dataset. Lastly, after applying these three different kinds of approaches, the research has been completed by doing analysis and by predicting whether it is hate speech or not.