Sentiment Analysis is a research field that studies human opinion, sentiment, evaluation, and emotions towards entities such as products, services, organizations, events, topics, and their attributes. It is also a task of natural language processing. However, sentiment analysis research has mainly been carried out for the English language. Although the Arabic language is one of the most used languages on the Internet, only a few studies have focused on Arabic language sentiment analysis.
In this paper, a review of the most important research works in the field of Arabic text sentiment analysis using deep learning algorithms is presented. This review illustrates the main steps used in these studies, which include pre-processing, feature extraction and classification, as well as the datasets used. In the end, all the research works are compared in terms of their methodology and results. The findings demonstrated that the majority of deep learning models, including CNN and LSTM, outperformed many of the machine learning models analyzed, and that the size of the training datasets had a direct correlation with the model's performance. Where larger datasets resulted in more successful model training.
