WMT-16-PubMed.py

# coding=utf-8
# Source: https://github.com/huggingface/datasets/blob/master/templates/new_dataset_script.py

"""WMT'16 Biomedical Translation Task - PubMed parallel datasets"""

import gzip
import datasets
import pandas as pd

logger = datasets.logging.get_logger(__name__)

_CITATION = """
@inproceedings{bojar-etal-2016-findings,
    title = Findings of the 2016 Conference on Machine Translation,
    author = {
      Bojar, Ondrej  and
      Chatterjee, Rajen  and
      Federmann, Christian  and
      Graham, Yvette  and
      Haddow, Barry  and
      Huck, Matthias  and
      Jimeno Yepes, Antonio  and
      Koehn, Philipp  and
      Logacheva, Varvara  and
      Monz, Christof  and
      Negri, Matteo  and
      Neveol, Aurelie  and
      Neves, Mariana  and
      Popel, Martin  and
      Post, Matt  and
      Rubino, Raphael  and
      Scarton, Carolina  and
      Specia, Lucia  and
      Turchi, Marco  and
      Verspoor, Karin  and
      Zampieri, Marcos
    },
    booktitle = Proceedings of the First Conference on Machine Translation: Volume 2, Shared Task Papers,
    month = aug,
    year = 2016,
    address = Berlin, Germany,
    publisher = Association for Computational Linguistics,
    url = https://aclanthology.org/W16-2301,
    doi = 10.18653/v1/W16-2301,
    pages = 131--198,
}
"""

_LANGUAGE_PAIRS = ['en-pt', 'en-es', 'en-fr']
_LANGUAGE_PAIRS_TUPLES = [('en','pt'), ('en','es'), ('en','fr')]

_LICENSE = """
This work is licensed under a <a rel="license" href="https://creativecommons.org/licenses/by/4.0/">Attribution 4.0 International (CC BY 4.0) License</a>.
"""

_DESCRIPTION = """
WMT'16 Biomedical Translation Task - PubMed parallel datasets
http://www.statmt.org/wmt16/biomedical-translation-task.html
"""

_URL = "https://huggingface.co/datasets/qanastek/WMT-16-PubMed/resolve/main/WMT16.csv.gz"


class WMT_16_CONFIG(datasets.BuilderConfig):
    def __init__(self, *args, lang1=None, lang2=None, **kwargs):
        super().__init__(
            *args,
            name=f"{lang1}-{lang2}",
            **kwargs,
        )
        self.name = f"{lang1}-{lang2}"
        self.lang1 = lang1
        self.lang2 = lang2

class WMT_16_PubMed(datasets.GeneratorBasedBuilder):
    """WMT-16-PubMed dataset."""

    DEFAULT_CONFIG_NAME = "en-fr"

    
    BUILDER_CONFIGS = [
        WMT_16_CONFIG(
            lang1=lang1,
            lang2=lang2,
            description=f"Translating {lang1} to {lang2} or vice versa",
            version=datasets.Version("16.0.0"),
        )
        for lang1, lang2 in _LANGUAGE_PAIRS_TUPLES
    ]
    BUILDER_CONFIG_CLASS = WMT_16_CONFIG
    
    # BUILDER_CONFIGS = [
    #     datasets.BuilderConfig(name=name, version=datasets.Version("16.0.0"), description=_DESCRIPTION) for name in _LANGUAGE_PAIRS
    # ]

    def _info(self):
        src, target = self.config.name.split("-")
        pair = (src, target)
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {"translation": datasets.features.Translation(languages=pair)}
            ),
            supervised_keys=(src, target),
            homepage="https://www.statmt.org/wmt16/biomedical-translation-task.html",
            citation=_CITATION,
            # license=_LICENSE,
        )

    def _split_generators(self, dl_manager):

        # Download the CSV
        data_dir = dl_manager.download(_URL)

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "filepath": data_dir,
                    "split": "train",
                }
            ),
        ]

    def _generate_examples(self, filepath, split):

        logger.info("⏳ Generating examples from = %s", filepath)

        key_ = 0
        
        with open(filepath, 'rb') as fd:
            gzip_fd = gzip.GzipFile(fileobj=fd)
            df = pd.read_csv(gzip_fd)

        # df = pd.read_csv(filepath, compression='gzip', header=0, sep=',')

        for index, row in df.loc[df['lang'] == self.config.name].iterrows():

            # Get langue pair
            src, target = str(row['lang']).split("-")

            yield key_, {
                "translation": {
                    src: str(row['source_text']).strip(),
                    target: str(row['target_text']).strip(),
                },
            }

            key_ += 1