本文以1949~2025年上海科技政策为研究对象，基于GraphRAG构建政策知识图谱，运用 PageRank算法识别7个历史阶段的关键政策节点。研究表明：上海科技政策呈现治理主体由少到多、部门间协同由简到繁、政策对象范畴由小到大的演变规律。本文将上海科技政策实证研究的时间起点首次延展至1949年，并为科技政策研究引入大语言模型驱动的结构化关系抽取与全局中心性测度方法。

Using GraphRAG and PageRank, this study builds a knowledge graph of Shanghai’s S&T policies (1949~2025) and identifies key policy nodes across seven historical stages.Findings reveal three evolutionary patterns: actors expanded from few to many, coordination evolved from simple to complex, and policy targets broadened from narrow to wide. The study pushes the empirical timeline back to 1949 and introduces LLM-driven relation extraction with global centrality measurement to S&T policy research.