Abstract—Formal Arabic is less commonly used in everyday communication, especially on social media, where colloquial Arabic dominates. However, most Arabic Natural Language Processing (NLP) tools, including sentiment analysis systems, are primarily developed for Modern Standard Arabic (MSA), making it difficult to analyze dialectal content. This study addresses this gap by focusing on the Palestinian dialect, which is widely used across the Levant region. Two main approaches were explored: machine learning and lexicon-based methods. A manually labeled dataset of Facebook telecom service reviews was collected, cleaned, and split into training and testing sets. Five machine learning algorithms and a neural network model (Multilayer Perceptron) were evaluated, while the lexicon-based approach used a predefined sentiment lexicon for colloquial Arabic. The machine learning models outperformed the lexicon-based method, achieving accuracy scores between 92.2% and 94.0%, with Naïve Bayes yielding the highest F1-Score (95.0%). The neural network model achieved 95.0% in both accuracy and F1-Score, while the lexicon-based method scored 81.0% accuracy and a 67.0% F1-Score. Although no hybrid model was implemented, the findings suggest future research could explore combining lexical and machine learning approaches.
 
Keywords—language, social media, colloquial Arabic, machine learning, lexicon-based approach, sentiment analysis, multilayer perceptron classifier


Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).