Download PDFOpen PDF in browserPractical Prompt Engineering for AI-Supported Metadata Processing in CRIS11 pages•Published: October 19, 2026AbstractThe growing digitalization of research management has transformed the availability, quality, and interoperability of research information into a central strategic factor within contemporary science systems. Despite increasing demands, metadata workflows in many organizations remain highly manual and heterogeneous, relying heavily on individual interpretation and localized convention. Research in information science has long identified metadata quality as a persistent bottleneck in digital knowledge infrastructures. This challenge is exacerbated in the CRIS context, where metadata no longer serve merely descriptive purposes but underpin strategic decision-making, reporting to funders, international benchmarking, and institutional visibility.Recent advances in large language models offer a new approach: the systematic use of prompt engineering to generate high-quality metadata through text-based instruction alone. Prompt engineering emerges as a scalable, transparent, and institutionally practical method for overcoming long-standing challenges in metadata workflows, improving the governance, interoperability, and strategic value of research information across the scholarly ecosystem. Building on this conceptual foundation, the paper focuses on concrete application scenarios and hands-on examples demonstrating how artificial intelligence and prompt engineering can be used to support the processing of research project metadata in CRIS environments. Keyphrases: artificial intelligence (ai), automation, classification, human in the loop, large language models (llms), metadata, metadata quality, prompt engineering In: Pablo de Castro and Joachim Schöpfel (editors). Proceedings of 17th International Conference on Current Research Information Systems, vol 5, pages 107-117.
|
