ru_anglicism.py

import json
from typing import List

import datasets


_DESCRIPTION = "Russian dataset for detection and substraction of anglicisms."
_URLS = {
    "train": "data/train.jsonl",
    "test": "data/test.jsonl"
}
_LICENSE = "apache-2.0"


class RuAnglDataset(datasets.GeneratorBasedBuilder):
    """Ru Anglicism Dataset"""

    VERSION = datasets.Version("0.2.0")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(name="default", version=VERSION, description=""),
    ]

    DEFAULT_CONFIG_NAME = "default"

    def _info(self):
        features = datasets.Features(
            {
                "word": datasets.Value("string"),
                "form": datasets.Value("string"),
                "sentence": datasets.Value("string"),
                "paraphrase": datasets.Value("string")
            }
        )
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=("sentence", "paraphrase"),
        )

    def _split_generators(self, dl_manager):
        downloaded_files = dl_manager.download_and_extract(_URLS)
        return [
            datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": downloaded_files["train"]}),
            datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": downloaded_files["test"]})
        ]

    def _generate_examples(self, filepath):
        with open(filepath, encoding="utf-8") as f:
            for id_, row in enumerate(f):
                data = json.loads(row)
                yield id_, data