modern_haiku.py

import datasets as ds

import json

_SUBSET_NAMES = ["all", "spring", "summer", "autumn", "winter", "none", "kigo"]

_COMMON_FEATURES = {
    "id": ds.Value("int64"),
    "haiku": ds.Value("string"),
    "author": ds.Value("string"),
    "foreword": ds.Value("string"),
    "source": ds.Value("string"),
    "comment": ds.Value("string"),
    "reviewer": ds.Value("string"),
    "note": ds.Value("string"),
}
_KIGO_FEATURES = {
    "id": ds.Value("int64"),
    "word": ds.Value("string"),
    "kana": ds.Value("string"),
    "old_kana": ds.Value("string"),
    "season": ds.Value("string"),
    "subtitle": ds.Sequence(ds.Value("string")),
}

# TODO: Add description of the dataset here
# You can copy an official description
_DESCRIPTION = """\
This new dataset is designed to solve this great NLP task and is crafted with a lot of care.
"""

# TODO: Add a link to an official homepage for the dataset here
_HOMEPAGE = ""

# TODO: Add the licence for the dataset here if you can find it
_LICENSE = ""

_FEATURES = {
    "all": ds.Features(
        {**_COMMON_FEATURES, "season": ds.Value("string"), "kigo": _KIGO_FEATURES}
    ),
    "spring": ds.Features({**_COMMON_FEATURES, "kigo": _KIGO_FEATURES}),
    "summer": ds.Features({**_COMMON_FEATURES, "kigo": _KIGO_FEATURES}),
    "autumn": ds.Features({**_COMMON_FEATURES, "kigo": _KIGO_FEATURES}),
    "winter": ds.Features({**_COMMON_FEATURES, "kigo": _KIGO_FEATURES}),
    "none": ds.Features(_COMMON_FEATURES),
    "kigo": ds.Features(_KIGO_FEATURES),
}

_DATA_URL = "https://pub-6dee886ee0a5425c8fb25fe18f3acc73.r2.dev/public/datasets/modern_haiku/data.json"


class ModernHaikuDataset(ds.GeneratorBasedBuilder):
    VERSION = ds.Version("0.0.1")

    BUILDER_CONFIGS = [ds.BuilderConfig(name=subset) for subset in _SUBSET_NAMES]
    DEFAULT_CONFIG_NAME = "all"

    def _info(self):
        return ds.DatasetInfo(
            description=_DESCRIPTION,
            features=_FEATURES.get(self.config.name),
            homepage=_HOMEPAGE,
            license=_LICENSE,
        )

    def _split_generators(self, dl_manager):
        data_dir = dl_manager.download(_DATA_URL)
        return [
            ds.SplitGenerator(
                name=ds.Split.TRAIN,
                gen_kwargs={
                    "filepath": data_dir,
                },
            ),
        ]

    def _generate_examples(self, filepath):
        with open(filepath, "r", encoding="utf-8") as f:
            data: list[dict] = json.load(f)

        key = 0

        if self.config.name == "all":
            for row in data:
                yield key, row
                key += 1

        elif self.config.name == "none":
            data = [row for row in data if row["season"] == self.config.name]
            for row in data:
                row.pop("season")
                row.pop("kigo")

                yield key, row
                key += 1

        elif self.config.name != "kigo":
            data = [row for row in data if row["season"] == self.config.name]
            for row in data:
                row.pop("season")

                yield key, row
                key += 1

        else:
            all_kigo = {}
            for row in data:
                if row["kigo"] is None:
                    continue
                kigo = row["kigo"]
                id = kigo["id"]
                if all_kigo.get(id) is None:
                    all_kigo[id] = kigo
                else:
                    continue

            for id, kigo in all_kigo.items():
                yield key, kigo
                key += 1