Attend the Women in ML Symposium on December 7 Register now

sberquad

References:

sberquad

Use the following command to load this dataset in TFDS:

ds = tfds.load('huggingface:sberquad/sberquad')
  • Description:
Sber Question Answering Dataset (SberQuAD) is a reading comprehension dataset, consisting of questions posed by crowdworkers on a set of Wikipedia articles, where the answer to every question is a segment of text, or span, from the corresponding reading passage, or the question might be unanswerable. Russian original analogue presented in Sberbank Data Science Journey 2017.
  • License: No known license
  • Version: 1.0.0
  • Splits:
Split Examples
'test' 23936
'train' 45328
'validation' 5036
  • Features:
{
    "id": {
        "dtype": "int32",
        "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"
    }
}