How do cultural ecosystem services affect visitor emotions? Evidence from social media data in Shanghai urban parks

Traditional Cultural Ecosystem Services (CES) assessments rely largely on questionnaires or interviews, which are limited in capturing large-scale user perceptions and often lack classification frameworks suitable for Chinese linguistic and cultural contexts. With increasing demand for high-quality urban park experiences, understanding how CES shape visitors' emotional responses has become essential for park planning and management.

Using social media review data from five representative urban parks in Shanghai, this study developed an integrated analytical framework that combines LDA topic modeling, a CES classification system adapted to Chinese contexts, sentiment analysis (SnowNLP and BERT), and linear mixed-effects modeling (LMM). A total of 22,564 valid user reviews were processed through text cleaning, segmentation, topic extraction, CES semantic classification, and sentiment quantification to evaluate how different CES types influence emotional outcomes.

Five CES categories were identified: Recreation and Family Engagement, Symbolic and Inspirational Landscapes, Physical and Mental Well-being, Aesthetic and Emotional Experience, and Education and Cognitive Engagement. Recreation and Family Engagement and Symbolic and Inspirational Landscapes were the most frequently perceived CES across parks, while Education and Cognitive Engagement appeared least often. Sentiment analysis showed overwhelmingly positive emotional expression, with "Very Positive" exceeding 83% of all reviews and displaying clear seasonal patterns-higher positivity in spring and autumn, and lower scores in summer and winter. LMM results indicated that Recreation and Family Engagement and Symbolic and Inspirational Landscapes significantly enhanced positive emotions, whereas Education and Cognitive Engagement showed a weak negative association in certain contexts.

The findings demonstrate the effectiveness of social media text analysis for large-scale CES quantification and highlight the differentiated emotional impacts of various CES types. This integrated LDA-CES-sentiment-LMM framework provides methodological innovation for data-driven CES assessment and offers practical insights for emotion-sensitive urban park planning, visitor experience enhancement, and the design of culturally responsive public spaces.