Detecting Hate Speech On Social Media

The Information exchange and the increasing use of social media have brought great benefits to humanity. However, this also causes some issues, such as the dissemination and exchange of hate speech messages. Therefore, to solve the issue, subsequent studies have employed a variety of feature engineering techniques and machine learning algorithms to automatically identify hate speech posts across a variety of datasets that is growing on social media sites. To our knowledge, no research has been done to contrast various feature engineering methodologies and machine learning algorithms in order to ascertain which feature engineering strategy and algorithm performs best on a widespread dataset that is publicly accessible. On a publicly available dataset with three different classes, this study examines the performance of three feature engineering strategies and eight machine learning algorithms. The experimental results show that the support vector machine approach, with an overall accuracy rate of 79%, works best when paired with bigram features. Our research has real- world applications and can serve as a benchmark in the field of automatically identifying hate speech texts. Additionally, the results of various comparisons will be used to compare upcoming studies for current automated text classification approaches using state-of-the-art methods.