References:
Use the following command to load this dataset in TFDS:
ds = tfds.load('huggingface:gigaword')
- Description:
Headline-generation on a corpus of article pairs from Gigaword consisting of
around 4 million articles. Use the 'org_data' provided by
https://github.com/microsoft/unilm/ which is identical to
https://github.com/harvardnlp/sent-summary but with better format.
There are two features:
- document: article.
- summary: headline.
- License: No known license
- Version: 1.2.0
- Splits:
Split | Examples |
---|---|
'test' |
1951 |
'train' |
3803957 |
'validation' |
189651 |
- Features:
{
"document": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"summary": {
"dtype": "string",
"id": null,
"_type": "Value"
}
}