Built Heritage Perception and Spatial Patterns of Tourist Affect in Dali Bai Autonomous Prefecture, China: A Social Media–Based Assessment

Against the backdrop of the concurrent processes of tourism development, spatial commercialisation, and architectural heritage conservation, identifying tourists’ cognitive images of destinations and differences in textual valence provides an important basis for evaluating the quality of tourism built environments and implementing fine-grained spatial governance. Taking Dali Bai Autonomous Prefecture in Yunnan Province as the study area, this study analyses 32,992 Ctrip attraction reviews using high-frequency term identification, two-level theoretical coding, a semantic co-occurrence network, and affective analysis to construct an attraction-level image-affect analytical framework. The results show that, under a theory-guided codebook, review content can be organised into four dimensions: natural landscapes, built heritage, ethnic culture, and services and accessibility. Natural mountain–lake scenery has the broadest review coverage, while strong semantic associations among architectural character, Bai vernacular dwellings, and Bai culture indicate that built heritage is an important spatial representation through which tourists recognise local culture. Textual valence is predominantly positive: positive, neutral, and negative affective terms account for 71.87%, 22.65%, and 5.48%, respectively. Positively coded propositions mainly concern natural and ecological landscapes, accounting for 44.12%, whereas 71.92% of negatively coded propositions concern services and accessibility, particularly prices and commercialisation, crowding and order, transport routes, and staff service. The complete corpus covers 228 attractions, five of which lack classifiable valence text. Using 43.49% for high-attention text and 88.44% for high-valence text as internal benchmarks, the remaining 223 attractions are classified as high-attention-high-valence, high-attention-low-valence, low-attention-high-valence, and low-attention-low-valence, accounting for 37.67%, 26.01%, 21.52%, and 14.80%, respectively. These categories reflect relative textual focus and textual valence rather than absolute levels of tourist satisfaction. The study identifies an asymmetric distribution between positive resource-related expressions and negative issues associated with the visiting process, providing a basis for architectural heritage conservation, tourism-service improvement, commercial-development regulation, and subsequent stratified on-site verification at representative places.