A Comparative Study of Automatic Hate Speech Detection Using Machine Learning

There is no denying that social media’s ubiquitous use and the knowledge it seamlessly disseminates have improved humanity. But despite its many benefits, this growth has also given rise to urgent worries, including the spread of hate speech. Modern research have embraced a variety of feature engineering techniques and machine learning algorithms as a remedy to this increasing difficulty inside the world of social media platforms. These initiatives aim to automatically identify hate speech across several datasets, offering a promising way to lessen this pervasive problem. As far as we are aware that no research has directly compared different feature engineering methods with different machine learning algorithms to determine which method produces the best results on a representative public dataset. This article’s goal is to evaluate how well three different feature engineering approaches work with eight different machine learning algorithms using an open source free publicly available datasets that include 03 different classes. The results of the experiment revealed that the combination of bigram features and the (SVM) support vector machine algorithm performed the best overall, with an amazing accuracy rate of “79%”. Our research consequences touch on actual situations, making it a landmark study that potentially laid the foundation for future efforts aimed at automatically identifying hate speech. Further, the results of these comparisons will serve as state-of-the-art approaches against which further studies of automated text classification can be measured.