Systematic Literature Review Of Hate Speech Detection With Text Mining

Abstract

Along with the increasing activity on social media, hate speech is getting out of control. Hate speech detection can be done by utilizing text mining technology. There have been many hate speech detection studies conducted. To identify and analyze research trends, data sources, methods and features used in hate speech detection, this systematic literature review was created. Until early 2020, the topics of hate speech were found, including hate speech against minorities, religion, women, the general election agenda, and politics. Sources of data that are widely used to be used as datasets come from twitter. Hate speech is not only classified into HS (hate speech) and Non-HS (non-hate speech) but can be further classified into racism, sexism, offensive, abusive, threats of violence and others. Of the 38 studies that meet inclusion and exclusion, there are 26 algorithms and 28 features that have been used to detect hate speech. However, these methods and features do not necessarily guarantee a good hate detection performance. Hate speech classification performance is also influenced by the dataset, the features chosen, the number of classes and mutually exclusive classes.