Analyzing Machine Learning Algorithms for Sentiments in Arabic Text

The studies addressing the application of machine and deep learning models to analyze the sentiments of Arabic online reviews related to the real-estate and automobile fields are not mature. To fill this gap, this research has focused on classifying three types of sentiments in Arabic real-estate and automobile online reviews, which are negative, positive, and mixed sentiments. The research focused on analyzing the reviews written in both Gulf Cooperation Council (GCC) dialects and modern standard Arabic (MSA). The research also explained the natural language processing strategies that were adopted to prepare the text for classification. The research discussed the details of collecting and annotating the data, preprocessing procedures, and feature selection methods. Following this, the research highlighted the adopted strategies for balancing and splitting the datasets, and it showed the analysis of the classification results for both machine and deep learning models. Finally, the suggestions for future work were provided in this research.