References:
Use the following command to load this dataset in TFDS:
ds = tfds.load('huggingface:quora')
- Description:
The Quora dataset is composed of question pairs, and the task is to determine if the questions are paraphrases of each other (have the same meaning).
- License: No known license
- Version: 0.0.0
- Splits:
Split | Examples |
---|---|
'train' |
404290 |
- Features:
{
"questions": {
"feature": {
"id": {
"dtype": "int32",
"id": null,
"_type": "Value"
},
"text": {
"dtype": "string",
"id": null,
"_type": "Value"
}
},
"length": -1,
"id": null,
"_type": "Sequence"
},
"is_duplicate": {
"dtype": "bool",
"id": null,
"_type": "Value"
}
}