Study on BERT Model for Hate Speech Detection

The last decade has witnessed a massive rise in the problem of hate speech due to revolutionary growth in online content. Urgent countermeasures are being taken by government bodies, social websites, multinational companies, research communities to solve the problem of hate speech detection, which is a text classification problem, by the development of automated methods of hate speech detection. A lot of manual effort and time is dedicated to find the solution for the problem of detecting hate speech from online content, but the authors of the paper feel that there is a lot of scope to improve the accuracy of the existing automated methods of hate speech detection. The authors observed that recently, transfer learning models have exhibited good results in the area of text classification for question-answering, summarization, and nextword prediction, but these learning models have not been used for the problem of hate speech detection to date. The authors of the article anticipate that these networks may give better results in another task of text classification, i.e., hate speech detection. Therefore, this article proposes a novel method of hate speech detection based on the concept of attention networks using the BERT attention model.