Emotion detection and sentiment dynamics in customer-generated reviews: A corpus-based study of Saudi Arabian food delivery applications

Abstract

The rapid growth of online food delivery platforms has resulted in large volumes of user-generated feedback, creating a critical need for automated systems that can extract meaningful insights for service improvement. Traditional sentiment analysis methods mostly rely on polarity classification, which frequently falls short of capturing the complex emotional elements that influence customer satisfaction and discontent. The emotion-aware feedback analysis system presented in this study is intended to convert customer reviews into practical suggestions for Saudi Arabian restaurant owners and food delivery services. The suggested system classifies customer reviews submitted in structured Excel format at the sentence level using a transformer-based deep learning model. Batch-based inference and lightweight text preprocessing are used to guarantee computational efficiency and practical usability. In order to produce understandable recommendations that focus on important operational areas like food quality, delivery performance, hygiene, pricing, and customer support, detected emotional patterns are combined with domain-specific service rules.the system incorporates geographic visualization to contextualize major food delivery platforms, including Marsool, Talabat, and HungerStation, within Saudi Arabia. The system successfully detects dominant emotional trends and generates useful insights that go beyond conventional sentiment polarity, according to experimental evaluation using actual customer review data. The findings show that emotion-aware analysis offers a deeper and more insightful comprehension of customer feedback, facilitating data-driven decision-making for service enhancement. The suggested method emphasizes the usefulness of emotion-aware natural language processing and computational linguistics in practical service-oriented applications.