Vectorization and Sentiment Analysis of Arabizi Text

Abstract In recent years, a new form of Arabic has emerged to facilitate communication between younger generations, particularly with the advent of social media platforms. Globalization was one of the primary factors that increased the importance of the English language, particularly with the widespread adoption of technology and the dominance of various technological devices and platforms \cite{HAL}. This new form of Arabic is 'Arabizi, a portmanteau of Araby-Englizi, meaning Arabic-English, is a digital trend in texting Non-Standard Arabic using Latin script \cite{taha01}. The intensified use of Arabizi has given rise to a plethora of new research concerns about how to interpret this type of language using various machine learning approaches. Consequently, Natural Language Processing (NLP) might aid in deriving substantial insights, allowing sentiment analysis. This study will review the various approaches presented in the literature to address this topic. Then we tested a set of machine learning models, deep learning models, and tested out ensembles made with both. The results were that a fine-tuned Support Vector Machine (SVC) produced the best results with an accuracy of 0.63 and F1 score of 0.59.