Sentiment Analysis On Arabic Companies Reviews

• Published in: 2024 6th International Conference on Computing and Informatics (ICCI) pp. 418 - 428
• Main Authors: Fouda, Aya E., Ahmed, Karim Salah, Ashraf Mohamed, Karim, Noshy, Mayer Mamdouh, ElKattan, Youssef, Mhran, Amany Ahmed, Abdelbaky, Ibrahim, Fouad, Khaled M.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 06.03.2024
• Subjects: Adaptation models; Analytical models; Arabic companies reviews; Arabic diacritics; Arabic language; Companies; Logistic regression; Machine learning models; Predictive models; PySpark; Reviews; Sentiment analysis
• DOI: 10.1109/ICCI61671.2024.10485056

This study introduces an innovative approach to sentiment analysis, specifically tailored for the Arabic language, a domain that poses unique challenges due to its complex morphology and diverse dialects. Utilizing a substantial dataset of over 108,000 reviews related to Arabic companies, our primary objective was to develop a robust and reliable sentiment scoring system that caters to the intricacies of the Arabic language, aimed at assisting businesses in understanding customer sentiments more effectively.Our methodology encompassed an extensive preprocessing phase, crucial for preparing the dataset for accurate analysis. This phase included converting emojis and emoticons into textual descriptions a contemporary and expressive facet of digital communication, removing Arabic diacritics to standardize text input, and normalizing the text for consistency. The preprocessing was carefully designed to address the specific challenges posed by the Arabic script and dialectal variations.For the core sentiment analysis, we employed two robust machine learning models: Logistic Regression and Naive Bayes. Each model was meticulously chosen and adapted for its proven effectiveness in text classification, especially in handling the nuances of sentiment analysis within the Arabic text. Our approach not only leverages the strengths of these models but also adapts them to accommodate the linguistic characteristics of Arabic. The results of our study provide significant contributions to the field of sentiment analysis in Arabic. By addressing the unique challenges of the language and adapting conventional machine learning techniques accordingly, our research offers valuable insights and tools for businesses and researchers focused on the Arab market. These insights are instrumental for companies seeking to understand and respond to customer sentiments in a linguistically diverse and culturally rich region.