paws_wiki

• Description:

Existing paraphrase identification datasets lack sentence pairs that have high lexical overlap without being paraphrases. Models trained on such data fail to distinguish pairs like flights from New York to Florida and flights from Florida to New York. This dataset contains 108,463 human-labeled and 656k noisily labeled pairs that feature the importance of modeling structure, context, and word order information for the problem of paraphrase identification.

For further details, see the accompanying paper: PAWS: Paraphrase Adversaries from Word Scrambling at https://arxiv.org/abs/1904.01130

This corpus contains pairs generated from Wikipedia pages, containing pairs that are generated from both word swapping and back translation methods. All pairs have human judgements on both paraphrasing and fluency and they are split into Train/Dev/Test sections.

All files are in the tsv format with four columns:

1. id: A unique id for each pair.
2. sentence1: The first sentence.
3. sentence2: The second sentence.
4. (noisy_)label: (Noisy) label for each pair.

Each label has two possible values: 0 indicates the pair has different meaning, while 1 indicates the pair is a paraphrase.

• Additional Documentation: Explore on Papers With Code

• Homepage: https://github.com/google-research-datasets/paws

• Source code: tfds.datasets.paws_wiki.Builder

• Versions:

  • 1.0.0: Initial version.
  • 1.1.0 (default): Adds configs to different subset and support raw text.

• Download size: 57.47 MiB

• Feature structure:

FeaturesDict({
    'label': ClassLabel(shape=(), dtype=int64, num_classes=2),
    'sentence1': Text(shape=(), dtype=string),
    'sentence2': Text(shape=(), dtype=string),
})

• Feature documentation:

Feature	Class	Shape	Dtype	Description
	FeaturesDict
label	ClassLabel		int64
sentence1	Text		string
sentence2	Text		string

• Supervised keys (See as_supervised doc): None

• Figure (tfds.show_examples): Not supported.

• Citation:

@InProceedings{paws2019naacl,
  title = { {PAWS: Paraphrase Adversaries from Word Scrambling} },
  author = {Zhang, Yuan and Baldridge, Jason and He, Luheng},
  booktitle = {Proc. of NAACL},
  year = {2019}
}

paws_wiki/labeled_final_tokenized (default config)

• Config description: Subset: labeled_final tokenized: True

• Dataset size: 17.96 MiB

• Auto-cached (documentation): Yes

• Splits:

Split	Examples
'test'	8,000
'train'	49,401
'validation'	8,000

• Examples (tfds.as_dataframe):

paws_wiki/labeled_final_raw

• Config description: Subset: labeled_final tokenized: False

• Dataset size: 17.57 MiB

• Auto-cached (documentation): Yes

• Splits:

Split	Examples
'test'	8,000
'train'	49,401
'validation'	8,000

• Examples (tfds.as_dataframe):

paws_wiki/labeled_swap_tokenized

• Config description: Subset: labeled_swap tokenized: True

• Dataset size: 8.79 MiB

• Auto-cached (documentation): Yes

• Splits:

Split	Examples
'train'	30,397

• Examples (tfds.as_dataframe):

paws_wiki/labeled_swap_raw

• Config description: Subset: labeled_swap tokenized: False

• Dataset size: 8.60 MiB

• Auto-cached (documentation): Yes

• Splits:

Split	Examples
'train'	30,397

• Examples (tfds.as_dataframe):

paws_wiki/unlabeled_final_tokenized

• Config description: Subset: unlabeled_final tokenized: True

• Dataset size: 177.89 MiB

• Auto-cached (documentation): Yes (validation), Only when shuffle_files=False (train)

• Splits:

Split	Examples
'train'	645,652
'validation'	10,000

• Examples (tfds.as_dataframe):