Background: Conventional SERVPERF-based hotel service quality assessments often rely on questionnaires, which may lead to midpoint bias, limited engagement, and subjective interpretation. Therefore, online customer reviews offer an alternative data source for capturing authentic service experiences. Objective: This study aims to evaluate service quality in five-star hotels in Surabaya by integrating SERVPERF, Analytical Hierarchy Process (AHP), text mining, sentiment analysis, and the 2-tuple linguistic model. Methods: A total of 1,200 TripAdvisor reviews from five five-star hotels in Surabaya were collected and analyzed. Expert judgments were obtained through AHP to determine the weight of each SERVPERF dimension. Review texts were segmented, classified into SERVPERF dimensions using a Large Language Model, and analyzed using TextBlob sentiment polarity. The results were then integrated into the 2-tuple linguistic model. Results: Tangibles was the most frequently discussed and most positively perceived dimension in customer reviews. However, Reliability received the highest expert-based AHP weight and became the most influential dimension after weighted SERVPERF scoring, followed by Responsiveness, Assurance, Empathy, and Tangibles. Conclusion: The proposed hybrid model provides a more comprehensive and data-driven approach to hotel service quality assessment by combining authentic customer reviews with expert-based weighting. This method helps hotel managers identify priority areas for service improvement more objectively.