Application of Large Language Model Text Analysis in the Study of Urban Image

To improve the accuracy of emotion discrimination in massive text analysis and analyze urban international image in detail, so as to provide a reference for urban construction and capital image building. In this paper, more than 80,000 social media comments were collected with the key words “Beijing” and “Peking”. Large language models (LLMs) was used to mine multi-level information of social media comments related to the evaluation of the city of Beijing, introduce in detail the relevant technical methods for large-scale text analysis using the LLMs, and comprehensively evaluate its performance. Our comparison results with the manual review results showed that the LLMs could accurately identify themes and discriminate emotions, which is expected to significantly improve the efficiency of text analysis and help researchers quickly grasp the themes and intentions of massive text information. Large Language Models (LLMs) significantly enhance text analysis efficiency. When combined with modular Python programming and iterative testing, they optimize performance and streamline debugging. Compared to traditional software, LLMs excel in interpreting complex semantics like irony and puns, and they adeptly handle multilingual social media comments. Their profound understanding of internet language and popular expressions further underscores their superiority in modern text analysis. Our text analysis on Beijing's image as an international city showed that Beijing's overall urban image on social platforms was positive and favorable, with higher attention paid to its economic, scientific, political and cultural dimensions, and higher subjective ratings in traditional culture, urban landscape, economic and technological fields, municipal facilities and public services. The international city image of Beijing in social media is dynamic, with multi-level and multi-theme relative independence and evolutionary stability.