参照:
次のコマンドを使用して、このデータセットをTFDSにロードします。
ds = tfds.load('huggingface:newsph_nli')
- 説明:
First benchmark dataset for sentence entailment in the low-resource Filipino language. Constructed through exploting the structure of news articles. Contains 600,000 premise-hypothesis pairs, in 70-15-15 split for training, validation, and testing.
- ライセンス:フィリピン語-テキスト-ベンチマークはGNU General PublicLicensev3.0の下でライセンスされています
- バージョン:1.0.0
- 分割:
スプリット | 例 |
---|---|
'test' | 9000 |
'train' | 420000 |
'validation' | 90000 |
- 特徴:
{
"premise": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"hypothesis": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"label": {
"num_classes": 2,
"names": [
"0",
"1"
],
"names_file": null,
"id": null,
"_type": "ClassLabel"
}
}