Semester of Graduation
Summer 2026
Degree Type
Thesis
Degree Name
Master of Science in Information Technology
Department
Information Technology
Committee Chair/First Advisor
Shirley Tian
Second Advisor
Zhigang Li
Third Advisor
Xu Tao
Abstract
This study develops an AI-driven framework to assess hotel service quality from online reviews. By using a large corpus of TripAdvisor hotel reviews, we apply BERTopic to uncover latent service topics at the sentence level and a transformer-based sentiment model to classify sentiment polarity (positive, neutral, and negative) and generate continuous sentiment scores. Topics are then mapped to the service quality model SERVQUAL dimensions (Tangibles, Reliability, Responsiveness, Assurance, Empathy). For each review, we compute dimension salience and performance and derive weighted indicators to capture both what guests discuss and how they feel. This study uses OLS regression model to examine the associations between the weighted SERVQUAL indicators and overall hotel ratings, with review length included as a control variable. The results show that Tangibles is the most salient dimension and has the strongest standardized association with overall rating, while Empathy shows the highest sentiment performance and the second strongest standardized association. Reliability and Responsiveness show relatively lower salience and weaker sentiment performance but remain significantly associated with overall rating. The controlled model explains 48.6% of the variance in overall hotel ratings. These findings show that AI-driven text analytics can transform unstructured hotel reviews into interpretable, theory-aligned service quality indicators and can support scalable service quality monitoring for hotel managers.