Quantitative Research on Regional Large Language Models Policy Texts Based on Bertopic

Policy plays a significant role as a special factor in the current development of China's AI industry. This paper selects 27 representative policy documents on AI large models from various provinces and cities across China for study. Through the use of Bertopic and LDA models, analyses are conducted based on word frequency, policy semantics, theme characteristics, and policy similarity. The research finds that provincial policies primarily focus on promoting AI technological innovation, application scenarios, and industrial development. There is a high degree of similarity among policy documents across different provinces, but the similarity between those from central-western regions and eastern coastal areas is lower. In contrast, there is higher similarity within the central region itself and among the eastern coastal provinces and cities.

I. Introduction

In recent years, China's artificial intelligence industry has garnered substantial attention from governments at all levels and received key support from national industrial policies. Following the issuance of the “New Generation Artificial Intelligence Development Plan” by the State Council in 2017, over ten targeted policies and directives pertaining to AI advancement have emerged from multiple governmental departments. In its 2024 edition, the “Government Work Report” advocated for an extensive implementation of the “Artificial Intelligence Plus” strategy. The rapid advancement of generative artificial intelligence and large language model technologies has prompted local governments to implement supportive policies and action plans aimed at sustaining the growth of the large-language-model industry. Presently, all regions have developed specific implementation plans and deployment strategies for Large Language Models (LLMs), generative AI, and general artificial intelligence. However, research on China's AI policies, particularly those concerning LLMs, remains relatively sparse and lacks comprehensive analysis of pertinent content. Regional disparities in development characteristics necessitate a comparative analysis of artificial intelligence policies across different regions. This research is crucial as it not only identifies common priorities but also elucidates the unique regional development traits by examining policy variances.