Mapping themes of heritage tourism with BERTopic—evidence from Nánxún Ancient Town

World Cultural Heritage sites are not only historical legacies but also spaces where visitors’ experiences and emotions are articulated. As the first Jiangnan ancient town inscribed on the UNESCO World Heritage List, Nánxún Ancient Town represents a unique case for examining how the public perceives and emotionally engages with heritage. This study integrates BERTopic for unsupervised topic modeling with a fine-tuned RoBERTa model for sentiment analysis to investigate large-scale social media data related to Nánxún. The analysis identifies several core themes of public concern, such as Historical Culture and Traditional Landscapes, Commercial Experience and Local Characteristics, Modern Entertainment and Cultural Activities, Transportation and Local Development, Daily Life and Community Interaction, and Ancient Bridges and Natural Landscapes. Sentiment classification further reveals the emotional distribution associated with each theme, highlighting both positive resonance and critical perceptions. Temporal and seasonal analyses demonstrate how public sentiments and topics evolve over time, reflecting dynamic patterns of visitor engagement. The findings offer data-driven insights for heritage managers, supporting evidence-based strategies in heritage conservation, tourism planning, and cultural communication.