The first LinkedCulture paper made the argument. The second paper shows how the system...

The first LinkedCulture paper made the argument. The second paper shows how the system actually works.

I've published Paper 2 in the LinkedCulture Research Paper Series on Zenodo: Embedding Cultural Heritage Metadata: Pipeline Design, Multilingual Retrieval, and Hybrid Search in LinkedCulture.

Read Paper 2: https://linkedculture.org/research/embedding-cultural-heritage-metadata

This paper covers the implementation behind LinkedCulture: how structured cultural heritage metadata gets translated into embeddable prose, how records from eight institutions are normalized into a shared schema, how OpenSearch and Qdrant support keyword and semantic retrieval, and how Reciprocal Rank Fusion combines both ranking signals.

It also documents one of the more important lessons from the prototype. Cross-institutional search becomes harder once collections cross language boundaries.

An early version of the index used an English-centered embedding setup. When Dutch Rijksmuseum records entered the system, English-language conceptual queries began returning confident but wrong results. The system did not fail visibly. It returned results. They looked plausible. They were just semantically wrong.

That failure pushed the prototype toward multilingual embedding.

This is not a benchmark study. It is a technical note on pipeline design, implementation choices, observed failure modes, and the tradeoffs involved in building hybrid retrieval across heterogeneous museum metadata.

The research behind this series is documented in several papers: Paper 1: Discovery architecture for cultural heritage: layered retrieval, institutional authority, and the limits of keyword search https://linkedculture.org/research/discovery-architecture-for-cultural-heritage

Paper 2: Discovery architecture for cultural heritage: layered retrieval, institutional authority, and the limits of keyword search https://linkedculture.org/research/embedding-cultural-heritage-metadata

Paper 3: Semantic Clusters and Language Partitioning in a Multi-Institutional Cultural Heritage Index https://linkedculture.org/research/semantic-clusters-language-partitioning

Learn more about LinkedCulture and the work behind it: https://linkedculture.org/

The YouTube channel discussing the project and presents related collections videos from LinkedCulture: https://lnkd.in/eTEKcArq

Qdrant / Ollama / OpenSearch Project "#MuseumTech #CulturalHeritage #SemanticSearch #VectorSearch #DigitalHumanities #LinkedCulture #OpenSearch #Qdrant #Ollama #Metadata