damt.py

# coding=utf-8
"""Multi-domain German-English parallel dataset for Domain Adapted Machine Translation."""

import datasets

_CITATION = """\
@inproceedings{koehn-knowles-2017-six,
    title = "Six Challenges for Neural Machine Translation",
    author = "Koehn, Philipp  and
      Knowles, Rebecca",
    booktitle = "Proceedings of the First Workshop on Neural Machine Translation",
    month = aug,
    year = "2017",
    address = "Vancouver",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/W17-3204",
    doi = "10.18653/v1/W17-3204",
    pages = "28--39",
}
@inproceedings{aharoni2020unsupervised,
  title={Unsupervised domain clusters in pretrained language models},
  author={Aharoni, Roee and Goldberg, Yoav},
  booktitle = "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
  year={2020},
  url={https://arxiv.org/abs/2004.02105},
  publisher = "Association for Computational Linguistics"
}
"""

_URL = "https://drive.google.com/file/d/1yvB-pvlojtT2UpOX1JvwtD6rw9joQ49A/view"

_HOMEPAGE = "https://github.com/roeeaharoni/unsupervised-domain-clusters"

_DOMAIN = ["it", "koran", "law", "medical", "subtitles"]


class DAMTConfig(datasets.BuilderConfig):
    """BuilderConfig for DAMT Dataset"""

    def __init__(self, domain=None, **kwargs):
        """
        Args:
            domain: domain name.
            **kwargs: keyword arguments forwarded to super.
        """
        super(DAMTConfig, self).__init__(
            name=domain,
            description="multi-domain German-English parallel dataset for Domain Adapted Machine Translation.",
            version=datasets.Version("1.0.0", ""),
            **kwargs,
        )

        # Validate domain name.
        assert domain in _DOMAIN

        self.domain = domain


class DAMT(datasets.GeneratorBasedBuilder):
    """Multi-domain German-English parallel dataset for Domain Adapted Machine Translation."""

    BUILDER_CONFIGS = [DAMTConfig(domain=d) for d in _DOMAIN]

    def _info(self):
        return datasets.DatasetInfo(
            # This is the description that will appear on the datasets page.
            description="multi-domain German-English parallel dataset for Domain Adapted Machine Translation",
            # datasets.features.FeatureConnectors
            features=datasets.Features(
                {"translation": datasets.features.Translation(languages=("en", "de"))}
            ),
            # If there's a common (input, target) tuple from the features,
            # specify them here. They'll be used if as_supervised=True in
            # builder.as_dataset.
            supervised_keys=None,
            # Homepage of the dataset for documentation
            homepage=_HOMEPAGE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""

        domain = self.config.domain

        def _get_drive_url(url):
            return f"https://drive.google.com/uc?id={url.split('/')[5]}"

        dl_dir = dl_manager.download_and_extract(_get_drive_url(_URL))
        files = {
            "train": {
                "en_file": f"{dl_dir}/{domain}/train.en",
                "de_file": f"{dl_dir}/{domain}/train.de",
            },
            "validation": {
                "en_file": f"{dl_dir}/{domain}/dev.en",
                "de_file": f"{dl_dir}/{domain}/dev.de",
            },
            "test": {
                "en_file": f"{dl_dir}/{domain}/test.en",
                "de_file": f"{dl_dir}/{domain}/test.de",
            },
        }

        return [
            datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs=files["train"]),
            datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs=files["validation"]),
            datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs=files["test"]),
        ]

    def _generate_examples(self, en_file, de_file):
        """Yields examples."""

        id_ = 0
        with open(en_file, "r", encoding="utf-8") as en_f:
            with open(de_file, "r", encoding="utf-8") as de_f:
                for en, de in zip(en_f, de_f):
                    yield id_, {"translation": {"en": en.strip(), "de": de.strip()}}
                    id_ += 1