A perfectly cataloged museum record can still be practically undiscoverable. Not becaus...

A perfectly cataloged museum record can still be practically undiscoverable. Not because the metadata is wrong. Because the person searching does not know the vocabulary the cataloger used.

That gap is structural. No amount of metadata improvement closes it. Neither does fuzzy search.

I have been building LinkedCulture to address this, and this week I published the first in a series of white papers documenting the architecture and findings.

The paper is titled: Discovery architecture for cultural heritage: layered retrieval, institutional authority, and the limits of keyword search.

The argument is not that AI should replace how museums and libraries surface their collections. The architectural principle the paper proposes is more precise: AI belongs at the representation layer, not the interpretation layer. The system helps users find records. It does not generate, summarize, or interpret them. The authoritative cataloged record is the only thing the end user ever sees.

LinkedCulture is the working demonstration. Built across eight institutions and more than two hundred thousand records using fully open source infrastructure, it documents three observations from operation: that layered retrieval surfaces records keyword search alone misses, that neither retrieval mode dominates unconditionally, and that a shared representational space appears to mediate vocabulary inconsistency across institutional boundaries in ways single-institution search cannot replicate.

The live system and project overview are at https://linkedculture.org/. Try a conceptual search and see what surfaces across eight institutions at once.

The papers are published on Zenodo with a permanent DOI. Paper 1: https://linkedculture.org/research/discovery-architecture-for-cultural-heritage Paper 2: https://linkedculture.org/research/embedding-cultural-heritage-metadata

This is the first paper in a planned series. Paper two will address embedding methodology and pipeline architecture. Paper three will address the emergent semantic cluster structure observed across the combined index.

"Institutions interested in the project or the research questions it raises are welcome to reach out.