ArtiFact: A Large-Scale Multi-Modal Cultural Heritage Dataset
ArtiFact is an extensive multi-modal cultural heritage dataset comprising 651,045 museum records. This dataset integrates data from the Metropolitan Museum of Art in New York City, the Art Institute of Chicago, and the Rijksmuseum in Amsterdam, featuring tables, text, and images. To enhance cross-modal error detection, a taxonomy of seven error categories has been applied to 130,209 records. Despite advancements, challenges remain in addressing material anachronisms and semantic queries related to cultural proximity and ambiguous object types. The dataset is detailed in the arXiv paper 2606.09648v2.
Key facts
- ArtiFact is a multi-modal cultural heritage dataset.
- It contains 651,045 museum records.
- Data sourced from the Metropolitan Museum of Art, the Art Institute of Chicago, and the Rijksmuseum.
- The dataset combines tables, text, and images.
- A taxonomy of seven error categories was injected into 130,209 records for cross-modal error detection.
- Detecting material anachronisms and temporal shifts remains an open challenge.
- Current systems struggle with semantic queries involving cultural proximity and ambiguous object types.
- The dataset is described in arXiv paper 2606.09648v2.
Entities
Institutions
- Metropolitan Museum of Art
- Art Institute of Chicago
- Rijksmuseum
Locations
- New York City
- Chicago
- Amsterdam