Transformer-Based Models for Arabic Text Sentiment Analysis: A Systematic Literature Review
Aissam Izimi, Amal Battou · 2024
This Systematic Literature Review (SLR) provides an overview of the use and effectiveness of transformer-based models in Arabic Sentiment Analysis (SA). The study involves an extensive search of six scientific databases, extracting detailed summaries of experiments conducted in the reviewed articles. The review aims to answer three research questions: 1) The most used transformer-based models in Arabic text, 2) Their effectiveness compared to traditional machine-learning models, and 3) The challenges faced by these models for Arabic SA. The most used transformer-based model was Arabert. But the choice of a model can be influenced by various factors such as task specificity, language and dialect, computational resources, analysis level, and the need for emotion detection. The review concludes with perspectives and future research directions in this field.