Analyzing Sentiment in Arabic Tweets: A Study Using Machine Learning and Deep Learning Techniques

Natural language processing is a field of study that encompasses various tasks, including text generation, language modeling, and classification. The extraction of valuable information from text documents allows for better decision-making in political decisions and analyzing public opinion; those techniques are also deployed in support systems. In the proposed study, we aim to analyze sentiment in Arabic Twitter comments written in Moroccan dialectal Arabic and modern standard Arabic. The process starts with data collection using the Twitter API, followed by data pre-processing. The polarity determination is based on emoji annotations, which is a major key in the model training. Word sequences are used to create word vectors, which are then processed through dense layers to produce customized word embeddings. Sentiment analysis is performed using both deep learning (DL) and machine learning (ML) models to extract opinions from Arabic text. The Arabic-collected dataset has been presented and evaluated using several indicators in order to compare the analysis performance between both machine learning and deep learning algorithms. The results of this study show promising results and satisfactory outcomes, which is very encouraging for additional studies in this field.