Using Machine Learning to Predict the Targets of Hate Speech on Social Media

Abstract

Social media facilitates people having a diverse set of personalities to communicate with each other. People are free to communicate with others without any limitations. Occa sionally such a communication results in to use of hate speech against others. Hate speech is the use of violent, aggressive, and offensive language. Though social media websites do not allow the use of hate speech, but the size of these platforms makes it nearly impossible to manage all their content. Consequently, several studies have been conducted for automatically detecting hate speech on social media. Focus of these stud ies is to detect the hateful content. Majority of these studies ignore predicting the tar get of the hate speech on social media. Focus of this study is to predict targets of hate speech. In this regard, firstly, a new balanced Hate Speech Targets Dataset (HSTD) is developed. HSTD contains tweets labeled for targets and non-targets of hate speech. Secondly, a novel framework Hate–speech Targets Prediction FrameworK (HTPK) is pro posed to predict the targets of hate speech on social media.For this purpose, we have used machine learning algorithms. There are many algorithms used for prediction in machine learning but we have applied the algorithms used in binary class prediction to HTPK and chose the algorithms that performed best. Comparison with state-of-the-art methods shows that HTPK performs better than these methods.