Dataset card
By Gareth HoyleLast updated How we research
A dataset card is a short document published with a dataset that describes its contents, origin, intended uses, limitations and license.
A dataset card is a standard document published with a dataset that describes what it contains, how it was collected, what it should and should not be used for, its known biases and its license. Hugging Face uses dataset cards as the README file for each dataset on its Hub.
Where did dataset cards come from?
Dataset cards grew out of Datasheets for Datasets, a 2018 paper by Timnit Gebru and colleagues. It proposed that every dataset ship with a datasheet answering questions on motivation, composition, collection, preprocessing, uses, distribution and maintenance.
How is a dataset card different from a data manifest?
A dataset card differs from a data manifest in purpose. The card is a narrative summary for anyone considering the data, while the manifest is a precise inventory used to price and check one delivery. Private licensing deals usually need both.