<![CDATA[Fast-food restaurants in Indonesia continue to grow, including Restaurant X, which competes to attract and retain customers. Reviews on digital platforms such as Google Maps have become a crucial provider of information for understanding customer satisfaction, complaints, and service aspects that require enhancement. The present study seeks to examine customer sentiment and identify key topics in customer reviews using a combination of Bidirectional Long Short-Term Memory (BiLSTM) and BERTopic. The originality of this research is reflected in the integratio of supervised and unsupervised learning approaches to provide a more holistic understanding of customer perceptions in the Indonesian fast-food restaurant industry. A total of 10,451 reviews from nine outlets with having the largest number of customer reviews during the 2016–2025 period were gathered and examined. The findings suggest that most reviews exhibit positive sentiment, with 7,762 reviews classified as positive. The BiLSTM model attained an accuracy rate of 78.6%, a precision value of 81.6%, a recall value of 83.5%, and an F1-score value of 82.5%, demonstrating strong classification performance. Meanwhile, BERTopic successfully identified dominant topics related to service quality, service speed, food spiciness, and menu variety. The research contribution lies in the development of a unified analytical framework capable of simultaneously identifying customer sentiment and uncovering the key issues influencing customer experiences. These findings can function as a data-based decision support system for enhancing service quality and customer satisfaction.]]>
