AI-based sentiment analysis of tourist reviews for improving municipal tourism services: case study of Isparta Municipality

Urban municipalities occupy a structurally pivotal yet analytically underserved position within smart city research: they are simultaneously infrastructure providers, public space curators, and de facto destination managers, yet the literature has rarely demonstrated how AI-driven governance tools can be embedded in their routine workflows rather than deployed as one-off research interventions. This study addresses that gap by presenting a fully operational, replicable closed-loop framework in which zero-shot BERT-based sentiment analysis and keyword-guided topic classification are integrated into the complaint-handling and service-quality processes of Isparta Municipality in Turkey. Drawing on dual-channel visitor feedback from Google Maps and a municipal WhatsApp hotline at the Kirazli Creek tourism facility, the framework demonstrates how AI-derived sentiment signals can be automatically routed to the relevant municipal sub-unit, eliminating manual triage and enabling precision resource allocation without specialist data science infrastructure. The procedural complexity of the system, encompassing multilingual preprocessing, nested cross-validation benchmarking across seven classifiers and two lexicon-based tools, zero-shot BERT classification via cosine-similarity matching, and a structured pre-post outcome assessment, constitutes the primary methodological contribution, as this layered architecture is precisely what makes the framework transferable to other municipalities and Destination Management Organisations. Following municipality-enacted service improvements, visitor sentiment shifted significantly and footfall increased substantially, providing empirical evidence that the smart city-smart tourism symbiosis theorised extensively in the literature is achievable at the medium-city level. Designed to scale, the framework opens substantive new perspectives for cities seeking to transition from reactive service management to genuinely adaptive, data-driven urban tourism stewardship.