About LinkedCulture

LinkedCulture

Search cultural collections by meaning, not just matching words.

Explore open-access museum records through semantic search across institutions, themes, materials, symbols, and visual ideas.

Search

Try a concept, visual description, material, cultural theme, or interpretive phrase.

Complete verification to search LinkedCulture.

    Research context

    LinkedCulture tests whether semantic search can improve discovery across cultural heritage metadata when users search by concept, mood, symbol, material, or meaning instead of exact catalog terms.

    The prototype currently searches a unified index of 217,643 open-access records across six museum collections using Ollama embeddings and Qdrant vector search.

    Counts updated 2026-05-09.

    217,643 unified records
    6 active collections
    Ollama embeddings
    Qdrant vector search
    Read methodology

    Traditional keyword search works well when the query matches the record. It struggles when the search is conceptual, ambiguous, visual, or metaphorical. LinkedCulture evaluates whether vector embeddings can surface useful relationships across records even when the words do not match directly.

    Object metadata from open-access museum collections is transformed into high-dimensional vectors using nomic-embed-text:latest served via Ollama. Those vectors are indexed in Qdrant and returned through source-specific and cross-collection search modes.

    Active collections

    Art Institute of Chicago

    58,443 records

    Rijksmuseum

    47,156 records

    Cleveland Museum of Art

    41,279 records

    Getty Museum

    59,979 records

    Harvard Art Museums

    6,509 records

    Met

    4,277 records

    Under review: Museum of New Zealand Te Papa Tongarewa — coming soon

    Open-source process

    MetadataOllama EmbeddingsQdrant SearchRanked Results