[目的/意义]尽管传统元数据方法主要聚焦资源的组织、分类与复用，但在人工智能技术广泛融合的背景下，图书馆元数据体系正面临从静态描述向动态语义组织的深层转型。以智能元数据为核心，旨在界定其能力边界与技术内涵，探索面向生成式人工智能环境的图书馆元数据重构路径与支撑框架。[方法/过程]在梳理大语言模型在资源生成、语义组织、内容适配、计算友好性与服务集成等方面实践基础的前提下，提出智能元数据的五维能力模型，并构建以“数据层—语义层—应用层”为基础的三层技术架构。进一步选取图书《美育：从培养孩子画画开始》作为案例，开展章节级语义建模实验，检验智能元数据在语义抽取、结构化生成与生成式任务支撑中的可行性。[结果/结论]智能元数据能够在较少人工干预的条件下，实现章节级语义要素的自动抽取与结构化表达，在编目优化、语义检索与知识服务等图书馆关键业务中具有一定的应用潜力。同时，由于缺乏统一标准体系与系统化评价框架，亟须通过跨学科协作强化语义基础设施的规范建设，以更好支撑图书馆面向生成式人工智能时代的智能化演进。
[Purpose/Significance] Although traditional metadata approaches have mainly focused on the organization, classification, and reuse of resources, the library metadata system is now undergoing a profound transformation from static description to dynamic semantic organization under the pervasive integration of artificial intelligence technologies. Focusing on smart metadata as the core concept, this paper aims to define its capability boundaries and technical connotations, and to explore reconstruction paths and supporting frameworks for library metadata in a generative AI environment. [Method/Process] On the basis of reviewing practical applications of large language models in resource generation, semantic organization, content adaptation, computational friendliness, and service integration, the paper proposes a five-dimensional capability model of smart metadata and constructs a three-layer technical architecture comprising a data layer, semantic layer, and application layer. Furthermore, the book Aesthetic Education: Starting from Teaching Children to Paint is selected as a case, and a chapter-level semantic modeling experiment is conducted to examine the feasibility of smart metadata in semantic extraction, structured generation, and support for generative tasks. [Result/Conclusion] The findings indicate that smart metadata can, with minimal human intervention, achieve automatic extraction and structured representation of chapter-level semantic elements, showing clear application potential in key library functions such as cataloging optimization, semantic retrieval, and knowledge services. At the same time, due to the lack of a unified standards system and systematic evaluation framework, there is an urgent need to strengthen the normative construction of semantic infrastructure through interdisciplinary collaboration, so as to better support the intelligent evolution of libraries in the era of generative artificial intelligence.
[1] GREENBERG J. Big metadata, smart metadata, and metadata capital: toward greater synergy between data science and metadata[J]. Journal of data and information science, 2017, 2(3): 19-36.
[2] NISO. Understanding metadata[EB/OL]. [2025-12-05]. https://groups.niso.org/higherlogic/ws/public/download/17446/Understanding%20Metadata.pdf.
[3] ULRICH H, KOCK-SCHOPPENHAUER A K, DEPPENWIESE N, et al. Understanding the nature of metadata: systematic review[J]. Journal of medical internet research, 2022, 24(1): e25440.
[4] LEIPZIG J, NÜST D, HOYT C T, et al. The role of metadata in reproducible computational research[J]. Patterns, 2021, 2(9): 1-28.
[5] GREENBERG J, WU M, LIU W, et al. Metadata as data intelligence[J]. Data intelligence, 2023, 5(1): 1-5.
[6] OYIGHAN D, UKUBEYINJE E S, DAVID-WEST B T, et al. The role of AI in transforming metadata management: insights on challenges, opportunities, and emerging trends[J]. Asian journal of information science and technology, 2024, 14(2): 20-26.
[7] 吴跃. AACR2与RDA的联系及在图书著录部分的区别[J]. 大学图书馆学报, 2010, 28(4): 77-83. (WU Y. From AACR2 to RDA— The difference between AACR2 and RDA in monograph description[J]. Journal of academic libraries, 2010, 28(4): 77-83.)
[8] 胡小菁. BIBFRAME核心类演变分析[J]. 中国图书馆学报, 2016, 42(3): 20-26. (HU X J. Evolution of BIBFRAME core classes[J]. Journal of library science in China, 2016, 42(3): 20-26.)
[9] YANG W, FU R, AMIN M B, et al. Impact and influence of modern AI in metadata management[J]. arXiv preprint arXiv:2501.16605, 2025.
[10] ZAVALIN V, ZAVALINA O L. Are we there yet? Evaluation of AI-generated metadata for online information resources[J]. Information research an international electronic journal, 2025, 30(iConf): 732-740.
[11] 郭利敏, 付雅明, 刘悦如. 检索增强思维在图书馆知识挖掘服务中的应用研究[J]. 图书情报工作, 2025, 69(21): 108-119. (GUO L M, FU Y M, LIU Y R. Research on the application of retrieval augmented thoughts in library knowledge mining services[J]. Library and information service, 2025, 69(21): 108-119.)
[12] 刘炜, 刘倩倩, 付雅明, 等. 人工智能时代的元数据方法论[J]. 图书馆理论与实践, 2023(4): 16-29. (LIU W, LIU Q Q, FU Y M, et al. Metadata methodology in artificial intelligence era[J]. Library theory and practice, 2023(4): 16-29.)
[13] SINGH M, KUMAR A, DONAPARTHI S, et al. Leveraging retrieval augmented generative LLMs for automated metadata description generation to enhance data catalogs[J]. arXiv preprint arXiv:2503.09003, 2025.
[14] XU Y, LIU Z, CHEN J, et al. Multi-modal knowledge graph generation with semantics-enriched prompts[J]. arXiv preprint arXiv:2504.13631, 2025.
[15] LEE J, WANG Y, LI J, et al. Multimodal reasoning with multimodal knowledge graph[J]. arXiv preprint arXiv:2406.02030, 2024.
[16] PENG J, HU X, HUANG W, et al. What is a multi-modal knowledge graph: a survey[J]. Big data research, 2023, 32: 100380.
[17] LU Y, ZHAO W, SUN N, et al. Enhancing multimodal knowledge graph representation learning through triple contrastive learning[C]//Proceedings of the thirty-third international joint conference on artificial intelligence. Jeju: International joint conferences on artificial intelligence, 2024: 5963-5971.
[18] LI C, DENG Y, HU H, et al. Incorporating external knowledge and goal guidance for LLM-based conversational recommender systems[J]. arXiv preprint arXiv:2405.01868, 2024.
[19] 罗鹏程, 王继民, 聂磊. 基于生成式大语言模型的文献资源自动分类研究[J]. 情报理论与实践, 2024, 47(12): 174-182. (LUO P C, WANG J M, NIE L. Research on automatic classification of literature resources based on generative large language model[J]. Information studies: theory & application, 2024, 47(12): 174-182.)
[20] 方卿, 陈瑞. ChatGPT下科技期刊语义出版的变革[J]. 图书情报工作, 2024, 68(18): 4-12. (FANG Q, CHEN R. The changes in semantic publishing of scientific journals under ChatGPT[J]. Library and information service, 2024, 68(18): 4-12.)
