Arabic Hate Speech Detection Using Different Machine Learning Approach

Hate speech is defined as an expression that targets an individual or community on the aspects like religion, sexual orientation, race, political opinion, and origin. Recently, hate speech on social media especially in the Arabic language has been exponentially increased and led to severe causes. Various studies had been conducted on social media platforms adopted by people to broadcast their opinions. This work aims to develop a model that is able to handle detection and classification of Arabic hate speech and offensive language. The experiments are carried out in using various machine learning (ML) and deep learning (DL) models. In this work, Arabic Hate Speech Detection (AHSD) model is proposed which composed of pre-processing, feature extraction, detection, and classification to identify hate speech on the Arabic benchmark dataset. The proposed model shows improved results. The transfer learning approach model exhibits superior performance compared to all other ML models in terms of accuracy, precision, recall, and F1 scores, achieving improvements of 84%, 79%, 80%, and 79%, respectively.