Real-World AAT Materials Dataset: 189,523 Cultural Heritage Object Descriptions

189,523 real-world examples pair material descriptions from cultural heritage collections with their corresponding Art & Architecture Thesaurus (AAT) classifications. The dataset is designed for training models to extract material information from cultural heritage object descriptions. It was created by 'small-models-for-glam' and last updated on September 19, -2025.

Use Cases

• Train text classification models to map free-text material descriptions to standardized AAT terms.
• Fine-tune NLP models for information extraction from cultural heritage collection records.
• Develop tools for automating metadata enrichment in museum and archival databases based on material descriptions.
• Benchmark model performance on real-world, domain-specific text classification tasks.

Strengths

• Contains 189,523 real-world examples, providing substantial volume for model training.
• Focuses on a specific, practical task: mapping cultural heritage descriptions to a standardized vocabulary (AAT).
• Data is sourced from real cultural heritage collections, likely reflecting authentic use cases.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but other scale details like file formats and size are unknown.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface, author: small-models-for-glam

Collection Method

Likely compiled from cultural heritage collection records.

Time Range

null

Freshness

Last updated 2025-09-19 16:18:56; freshness should be verified.

Geography

null