Libraly Journal

Libraly Journal ›› 2026, Vol. 45 ›› Issue (11): 87-97.

Previous Articles     Next Articles

Metadata at the Core: Bridging Human Knowledge and AI Innovation—A Review of the 2025 Dublin Core Metadata Initiative Conference (DCMI2025)

Qiu Kaiyu   

  • Online:2026-11-15 Published:2026-11-09
  • About author:Qiu Kaiyu

Abstract: The 2025 annual conference of the Dublin Core Metadata Initiative (DCMI) was successfully held at the University of Barcelona, Spain, under the theme “Metadata at the Core: Bridging Human Knowledge and AI Innovation”. The conference focused on the evolving role and future development of metadata in the AI era. Based on the conference output, the paper first addressed the fundamental question of “whether metadata is still needed in the AI era”, and argued that metadata had evolved from a traditional descriptive tool to an infrastructure provider in the AI era, serving as the core bridge between traditional knowledge systems and cutting-edge AI innovation. It articulated metadata's threefold roles as quality training fuel, an anchor of truth, and a knowledge association engine, at the construction, knowledge generation, and application stages of AI systems. Building on this perspective, the paper established a three-dimensional analytical framework of “production-application-governance” to systematically elaborate on the multifaceted roles of metadata within the AI knowledge ecosystem. At the production level, it analyzed the pathways for reshaping metadata as AI infrastructure from three dimensions: collaborative organizational models, a standardized interoperability technical framework, and curation tools. At the practical application level, it focused on cutting-edge practical methods for leveraging the mutual empowerment of AI and metadata to drive innovation in knowledge services. At the governance level, it delved into the challenges of risk governance in building a trustworthy AI ecosystem and proposed response pathways such as data sovereignty protection, source traceability, and bias mitigation. These findings provided a significant theoretical foundation and practical reference for constructing an AI knowledge service system that balances efficiency with trustworthiness.

Key words: Metadata,&emsp, DCMI,&emsp, Artificial&emsp, intelligence,&emsp, large&emsp, language&emsp, models,&emsp, Knowledge&emsp, graph