A Review of the Recent Research in Arabic Offensive Language Detection Using Machine Learning

Abstract

This paper aims to review the status of research on the detection of offensive content in the
Arabic language in social media. A search and selection process of four pages of Google
Scholar yielded 32 papers for this review. From the variety of the models, changes, and
additions used in the models, it is difficult to conclude that a particular model will be the
most suitable for detecting offensive Arabic language used in social media. Each researcher
or group addressed a specific aspect of detection. Future research could focus on using hybrid
models, combining models, transfer learning, and ensembled models for detecting offensive
content and the context of Arabic dialects needs to be researched more deeply.