Sentiment Analysis: Implementing ML techniques for the Arabic language to improve the accuracy of current techniques

Mouneera AlQahtani
Bachelor’s Degree Student, Information Technology, Imam Muhammad bin Saud Islamic University, Riyadh, Saudi Arabia
Riyad Almakki
Assistant Professor – PhD in Information Systems, Imam Muhammad bin Saud Islamic University, Riyadh, Saudi Arabia
Modhi Al Shaikh
ssistant Professor – PhD in Information Systems, Imam Muhammad bin Saud Islamic University, Riyadh, Saudi Arabia
Email : alqahtanimouneera@gmail.com

Abstract

Sentiment analysis in the Arabic language faces unique challenges due to its complex morphology, diverse dialects, and limited linguistic resources. While substantial research has been conducted in the field, achieving high accuracy in sentiment classification remains a pressing issue. In this paper, we systematically review 31 studies on sentiment analysis for Arabic product reviews published between 2018 and 2024. We focus on recent advancements in machine learning (ML) and deep learning (DL) techniques, examining methodologies, datasets, and the performance results achieved. Our review categorizes the sentiment analysis techniques into lexicon-based, machine-learning-based, and hybrid approaches, with a particular emphasis on the prevalent use of ML models in Arabic sentiment analysis. The studies reviewed employ a variety of algorithms including Naïve Bayes, Decision Trees, SVM, CNN, and AraBERT models, among others. Furthermore, our analysis highlights the common preprocessing and feature extraction techniques utilized, as well as evaluation metrics employed to determine the efficacy of these models. Despite notable advancements, our findings indicate that many existing approaches fall short of delivering optimal results. We argue that future research should consider implementing alternative machine learning models and leveraging comprehensive datasets to enhance the accuracy of current techniques in Arabic sentiment analysis.

Keywords: Arabic sentiment analysis, machine learning, deep learning, lexicon-based techniques, sentiment classification, natural language processing