Exploring the Emotional Spatial Patterns and Functional Perceptions of Internet-Famous Streets Based on Large Language Models: A Case Study of Nanjing Old City

Abstract

This study analyzes over one million geo-tagged Sina Weibo posts (2019–2024) to evaluate emotional patterns and influencing factors on internet-famous streets in Nanjing Old City. First, high-traffic streets and their functional zones are identified using ArcGIS-based spatial analysis. Second, advanced large language models (LLMs) are used to detect residents’ and tourists’ emotions through a four-dimensional structure: primary and secondary emotions, intensity, and polarity. These emotional labels are mapped to street-level sentiment profiles. Spatial clustering and semantic analysis reveal distinct emotional dynamics driven by spzeatial semantics and functional typologies.