Analyzing tourist satisfaction in remote island homestays: a sentiment-based study of Nam Du island tourism, Vietnam

Purpose This study examines tourist satisfaction with homestay accommodations on Nam Du Island, Vietnam, using user-generated online reviews. It aims to identify overall sentiment, explore key service attributes associated with satisfaction and dissatisfaction, and examine differences between reviewer groups based on platform engagement. Design/methodology/approach A mixed-methods approach integrating natural language processing and thematic analysis was employed. A total of 2,056 Google Maps reviews from 41 homestays were analyzed using the VADER sentiment model. Keyword extraction (term frequency–inverse document frequency, TF-IDF), topic modeling (latent Dirichlet allocation, LDA) and manual thematic interpretation were applied to identify dominant patterns. Independent-samples t-tests with effect sizes and robustness checks were conducted to compare sentiment scores and ratings between Google Local Guides and non-Local Guide reviewers. Additionally, Spearman's rank correlation was used to assess the relationship between automated sentiment scores and explicit star ratings. Findings Results indicate that 75.5% of reviews are positive, highlighting scenic views, cleanliness, reasonable pricing and hospitality as commonly associated with favorable evaluations. Negative reviews (17.4%) focus on service attitude, infrastructure constraints and inconsistencies between expectations and actual service. While parametric t-tests identified statistically significant differences between reviewer groups based on platform engagement, these differences are not robust to nonparametric checks and represent trivial effect sizes. Consequently, platform-based variations in perception are statistically marginal, and the comparative analysis is ultimately inconclusive. Practical implications The findings suggest that improving service consistency, communication transparency and basic infrastructure reliability can enhance guest satisfaction. Implications are differentiated into operator-level and broader actions. Originality/value This study provides unique contextual and theoretical insights into service quality in remote island homestays. Rather than claiming methodological novelty, its originality lies in demonstrating how environmental and infrastructural constraints (such as utility instability and host-coordinated isolated logistics) adapt traditional performance-based service quality frameworks. Furthermore, it establishes a critical theoretical bridge by empirically demonstrating that automated NLP sentiment analysis effectively indexes perceived service quality in emerging, informal destinations lacking formal institutional reputation systems.