Analyzing the Evolution of Jiangsu’s Science and Technology Policies Based on Topic Modeling and Fine-Tuned RoBERTa

This paper examines the evolution of science and technology policies in Jiangsu Province from 2019 to 2023, utilizing advanced analytical techniques to reveal significant thematic changes and stakeholder dynamics. The study employs Latent Dirichlet Allocation (LDA) for topic modeling and fine-tunes the Robustly optimized BERT approach (RoBERTa) with transfer learning for entity extraction. The results depict intricate relationships between policy themes over different years, marking a significant shift from an initial focus on business incubation and foreign talent to a strategic emphasis on artificial intelligence (AI). Moreover, the research characterizes the complex co-occurrence network relationships between various stakeholders and themes, uncovering valuable network insights. The government and enterprises are central to Jiangsu’s innovation ecosystem, with the government providing essential policy guidance and enterprises driving technological progress. Social organizations and citizens, although less central, still make meaningful contributions. These findings underscore the necessity of strengthening stakeholder collaboration. Overall, this study not only offers a novel computational approach and perspective for understanding science and technology policy but also contributes to effective policymaking. To address current limitations, future research should incorporate more variables and explore regional differences to gain more comprehensive understanding of policy dynamics.