Enriching Digital Heritage with LLMs and LOD

Cultural heritage institutions such as libraries, museums, and archives hold vast collections that document human history. However, much of this heritage remains difficult to discover or connect digitally because the metadata describing these objects is often incomplete, inconsistent, or unstructured. The goal of this project was to explore how Large Language Models (LLMs) and Linked Open Data (LOD) can be combined to improve cultural heritage metadata in a reliable, transparent, and reusable way. The project brought together experts from cultural heritage, digital humanities, semantic web technologies, and artificial intelligence in an intensive international workshop.

Rather than treating LLMs as a replacement for existing infrastructure, the project identified where AI adds value and where traditional methods work better. The main result is the Cultural Heritage AI Cookbook, a freely available, practical guide that shows how to recognise named entities (such as people and places), link them to authoritative identifiers, and describe their relationships to heritage objects using a hybrid of LLMs and LOD.

The impact of the project lies in changing how researchers and practitioners think about using AI in heritage contexts. Instead of “black box” automation, the project demonstrated step-by-step workflows that are inspectable, testable, and adaptable. This lowers the barrier for heritage institutions to experiment responsibly with AI while maintaining professional and ethical standards.

The project met its objectives and refined them along the way. Initially focused on experimentation, it evolved toward producing reusable outputs that non-experts can apply to their own collections. The primary audience for these results includes cultural heritage professionals, digital humanities researchers and research software engineers.

Next steps include wider dissemination, community feedback, conference publications, and using the Cookbook as a foundation for future international funding proposals. The call-to-action is simple: heritage practitioners are invited to explore, reuse, and adapt the Cookbook workflows and contribute their own experiences back to the community via GitHub and related networks.