פיאף

ds = tfds.load('huggingface:piaf/plain_text')

Piaf is a reading comprehension dataset. This version, published in February 2020, contains 3835 questions on French Wikipedia.

{
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "title": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "context": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "question": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "answers": {
        "feature": {
            "text": {
                "dtype": "string",
                "id": null,
                "_type": "Value"
            },
            "answer_start": {
                "dtype": "int32",
                "id": null,
                "_type": "Value"
            }
        },
        "length": -1,
        "id": null,
        "_type": "Sequence"
    }
}