Create a dataset card

Each dataset should be accompanied with a Dataset card to promote responsible usage, and alert the user to any potential biases within the dataset. This idea is inspired by the Model Cards proposed by Mitchell, 2018. Dataset cards help users understand the contents of the dataset, context for how the dataset should be used, how it was created, and considerations for using the dataset. This guide shows you how to create your own Dataset card.

  1. Create a new Dataset card by opening the online card creator, or manually copying the template from here.

  2. Next, you need to generate structured tags. The tags help users discover your dataset on the Hub. Create the tags with the online Datasets Tagging app.

  3. Select the appropriate tags for your dataset from the dropdown menus, and save the file once you are done.

  4. Expand the Show YAML output aggregating the tags section on the right, copy the YAML tags, and paste it under the matching section on the online form. Paste the tags into your README.md file if you manually created your Dataset card.

  5. Expand the Show Markdown Data Fields section, paste it into the Data Fields section under Data Structure on the online form (or your local README.md). Modify the descriptions as needed, and briefly describe each of the fields.

  6. Fill out the Dataset card to the best of your ability. Refer to the Dataset Card Creation Guide for more detailed information about each section of the card. For fields you are unable to complete, you can write [More Information Needed].

  7. Once you are done filling out the card with the online form, click the Export button to download the Dataset card. Place it in the same folder as your dataset.

Feel free to take a look at these examples of good Dataset cards for inspiration: