Applying sentiment analysis to evaluate consumer responses to Vietnamese brands

This study evaluates the application of natural language processing to analyzing consumer responses from online feedback data. It compares traditional machine-learning models with pretrained language models and combines sentiment classification, aspect-based analysis, and a Sentiment Index. The findings indicate that pretrained language models, particularly models optimized for Vietnamese such as PhoBERT, can improve sentiment classification performance by better capturing linguistic context. Combining the Sentiment Index with aspect-based analysis enables identification of both overall sentiment trends and specific product or brand attributes influencing consumer responses. The study demonstrates how large volumes of unstructured feedback can be transformed into structured, actionable information, providing a data-driven basis for brand management, customer-experience monitoring, and managerial decision-making.