wikiculture.py

import datasets

class WikiCultureConfig(datasets.BuilderConfig):
    def __init__(self, data_files, **kwargs):
        super().__init__(**kwargs)
        self.data_files = data_files


class WikiCulture(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.0.0")

    BUILDER_CONFIGS = [
        WikiCultureConfig(
            name="all",
            description="All cultures combined",
            data_files=[
                "AF.parquet", "AS.parquet", "AU.parquet", "CH.parquet",
                "EU.parquet", "LA.parquet", "ME.parquet", "NA.parquet",
            ],
        ),
        WikiCultureConfig("AF", "Africa", ["AF.parquet"]),
        WikiCultureConfig("AS", "Asia", ["AS.parquet"]),
        WikiCultureConfig("AU", "Australia & Oceania", ["AU.parquet"]),
        WikiCultureConfig("CH", "China", ["CH.parquet"]),
        WikiCultureConfig("EU", "Europe", ["EU.parquet"]),
        WikiCultureConfig("LA", "Latin America", ["LA.parquet"]),
        WikiCultureConfig("ME", "Middle East", ["ME.parquet"]),
        WikiCultureConfig("NA", "North America", ["NA.parquet"]),
    ]

    def _info(self):
        return datasets.DatasetInfo(
            features=datasets.Features({
                "page_title": datasets.Value("string"),
                "qitem": datasets.Value("string"),
                "bucket_geo": datasets.Value("string"),
                "wiki_id": datasets.Value("int64"),
                "text": datasets.Value("string"),
            })
        )

    def _split_generators(self, dl_manager):
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"files": self.config.data_files},
            )
        ]

    def _generate_examples(self, files):
        import pyarrow.parquet as pq
        idx = 0
        for file in files:
            table = pq.read_table(file)
            for row in table.to_pylist():
                yield idx, row
                idx += 1