Appliance of Deep Learning on hate speech detection: a systematic review

Abstract

Hate speech is defined as offensive viewpoints toward individuals or groups divided by ethnicity, religion, or gender. Hate speech is not restricted by law in many countries, and may be seen as the protection of free speech since whether a statement is hate speech is highly related to the individual. Hate speech is widely recognized and has been researched extensively in recent years due to its nebulous definition and quick spread. With deep learning's ongoing growth, some researchers have started to apply deep neural networks to hate speech detection (HSD). Although a lot of progress has been made in research-based work around the task, a comprehensive overview of the progress of the task is lacking. Therefore, an overview of the progress that using deep neural networks to address hate speech problems is necessary. The overview presents what methods have been employed recently to enhance the task's performance, and also analyzes the potential problems of the task.