Public emotions and visual perception of historic district : An AI approach using social media data

Abstract

Historic districts possess unique visual landscape features that can influence public emotional experiences. Advances in artificial intelligence, particularly multimodal large models, enable integrated analysis of image and text data, offering new possibilities for understanding the complex relationship between visual environments and emotions. This study focuses on Nanluoguxiang, a historic district in Beijing, using 2,851 text reviews and 13,553 associated images collected from the social media platform Dianping between January 2024 and January 2025. An AI-based emotion analysis framework was developed, and XGBoost was applied to examine the associations between seven types of visual scenes and emotional responses. Results indicate that joy dominates public sentiment; commercial scenes are highly perceived; and architectural and streetscape elements are linked to positive emotions, while storefronts correlate with negative emotions.