Glossary term
Glossary term
Governance and Compliance
A structured summary of a dataset's source, composition, collection method, labeling, intended uses, limitations, privacy considerations, and known biases. Dataset cards help reviewers understand whether data is appropriate for the intended population, decision context, geography, time period, and risk profile.
Gebru et al. (2018) introduced Datasheets for Datasets, which became the foundation for modern dataset cards.
Hugging Face Dataset Cards are widely used for open data including LAION, RedPajama, and The Pile, providing structured documentation.
The IBM AI FactSheets initiative extended dataset card concepts into enterprise documentation for regulated industries.