Emotional landscape analysis of cultural ecosystem services in heritage parks: a deep learning approach using social media data

Heritage parks provide diverse Cultural Ecosystem Services (CES), yet the emotional connections between landscape elements and CES categories, particularly in culturally significant sites, remain understudied. This research pioneers the integration of deep learning and large-scale social media text analysis to explore these connections. Analyzing 86,647 comments from five social media platforms (Dianping, Meituan, Ctrip, Tongcheng, and Weibo), eight CES categories were identified using a Text-CNN model, revealing key correlations and clusters. Sentiment and social network analyses further highlighted emotional impacts and semantic associations. Recreation (43.77%), aesthetics (15.77%), and social relationships (13.55%) were dominant in CES discussions. Spirituality was negatively correlated with most other CES, while recreation was positively correlated with all except spirituality, with the strongest link between social relationships and recreation (p = 0.55). Culture, biological environments, and sentiment significantly influenced CES categories. Cultural and biological factors were primary predictors of sentiment, which in turn shaped CES perceptions. Despite challenges such as data heterogeneity and cultural differences in sentiment interpretation, this study demonstrates the value of combining deep learning and social media analysis for CES-focused park management.