大小模型协同驱动的标准文献语义匹配方法与应用研究

摘要：

【目的】文本语义匹配是推动标准数智化深入发展的关键技术，本文提出面向标准数智化转型的标准文献语义匹配框架以支撑标准数智化建设。【方法】结合新兴技术与标准数智化需求，开展了大小模型协同驱动下的标准文献语义匹配方法与应用研究。【结果】构建了融合大模型与轻量级小模型的标准文献语义匹配框架，并以ICS与国民经济行业分类的跨体系映射为例验证了该范式的可行性，形成了ICS与国民经济行业分类在中类层级上的映射数据集。【结论】为涉及标准文献语义匹配的标准数智化应用场景建设提供技术参考，助推标准数智化的深入发展。

Abstract:

[Objective] Text semantic matching is a key technology driving the in-depth development of standard digitalization and intelligentization. This paper proposes a standard literature semantic matching framework for the transformation towards standard digitalization and intelligentization. [Methods] Combining emerging technologies with the needs of standard digitalization and intelligentization, research on standard literature semantic matching methods and applications driven by the collaboration of large and small models has been conducted. [Results] A standard literature semantic matching framework integrating large models and lightweight small models has been developed, and the feasibility of this paradigm has been verified through an example of cross-system mapping between the ICS and industrial classification for national economic activities, resulting in a mapping dataset of the ICS and industrial classification for national economic activities at the medium category level. [Conclusion] This provides technical reference for the construction of standard digitalization and intelligentization application scenarios involving standard literature semantic matching, promoting the in-depth development of standard digitalization and intelligentization.

引用格式：方思怡. 大小模型协同驱动的标准文献语义匹配方法与应用研究[J]. 标准化学报，2026(11): 17-23.

基金项目：本文受上海市质量和标准化研究院院立项目“生成式人工智能视角下的标准专题库建设及应用研究”（项目编号：YRY202610）资助。

作者简介：方思怡，硕士，工程师、知识产权师，研究方向为标准数智化、标准知识服务。