An analysis of customer perception using lexicon-based sentiment analysis of Arabic Texts framework

Area of Science:

• Natural Language Processing (NLP)
• Computational Linguistics
• Social Media Analytics

Background:

• Growing demand for sentiment analysis in Arabic due to global social media reach.
• Existing research inadequately addresses Arabic customer perceptions of products like coffee.
• Need for robust Arabic text analysis tools to understand consumer feedback.

Purpose of the Study:

• To propose and validate a comprehensive Lexicon-based Sentiment Analysis on Arabic Texts (LSAnArTe) framework.
• To analyze customer perceptions of coffee and coffee products using Arabic social media data.
• To evaluate the performance of the LSAnArTe framework against existing tools.

Main Methods:

• Developed the LSAnArTe framework utilizing the AraSenTi dictionary and Qalasadi platform for lemmatization.
• Classified word and sentence sentiment based on lexicon scores.
• Validated the framework using a manually annotated dataset of 10,769 tweets from X (formerly Twitter).

Main Results:

• The LSAnArTe framework achieved a high accuracy of 93.79% in sentiment analysis.
• The proposed framework significantly outperformed Amazon Comprehend, which achieved 51.90% accuracy.
• Manual annotation ensured high accuracy and reliability for framework validation.

Conclusions:

• The LSAnArTe framework provides a highly accurate solution for Arabic sentiment analysis on social media data.
• This research addresses a critical gap in analyzing Arabic customer feedback for products like coffee.
• The developed framework offers a valuable tool for businesses seeking to understand Arabic-speaking consumer markets.