xsum

ds = tfds.load('huggingface:xsum')

Extreme Summarization (XSum) Dataset.

There are three features:
  - document: Input news article.
  - summary: One sentence summary of the article.
  - id: BBC ID of the article.

{
    "document": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "summary": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}