political.py

import os
from pathlib import Path
import datasets
from typing import List

STYLE, CLASSIFIER = "style", "classifier"
_CITATION = """\
@inproceedings{style_transfer_acl18,
    title={Style Transfer Through Back-Translation},
    author={Prabhumoye, Shrimai and Tsvetkov, Yulia and Salakhutdinov, Ruslan and Black, Alan W},
    year={2018},
    booktitle={Proc. ACL}
    }
"""

_DESCRIPTION = """\
Political slant transfer dataset. Contains two classes of political tweets between Democratic and Republican Politicans. This dataset can be used for classification tasks.
"""
_HOMEPAGE = "https://github.com/shrimai/Style-Transfer-Through-Back-Translation"
_LICENSE = ""  # could not find.


class PoliticalDataset(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.0.0")
    DEFAULT_CONFIG_NAME = STYLE

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name=STYLE,
            version=VERSION,
            description="Political Tweets Dataset, used for Style Transfer tasks.",
        ),
        datasets.BuilderConfig(
            name=CLASSIFIER,
            version=VERSION,
            description="Political Tweets Dataset, Used for classification tasks.",
        ),
    ]

    def _info(self):
        features = datasets.Features(
            {
                "text": datasets.Value("string"),
                "label": datasets.Value("string"),
            }
        )
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=("text", "label"),
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager) -> List[datasets.SplitGenerator]:
        data_dir = "political_data"

        splits: List[datasets.SplitGenerator] = []
        if self.config.name == STYLE:
            splits.append(
                datasets.SplitGenerator(
                    name=datasets.Split.TRAIN,
                    gen_kwargs={
                        "filepaths": [
                            os.path.join(data_dir, "republican_only.train.en"),
                            os.path.join(data_dir, "democratic_only.train.en"),
                        ],
                        "split": "train",
                    },
                )
            )
        else:
            splits.append(
                datasets.SplitGenerator(
                    name=datasets.Split.TRAIN,
                    gen_kwargs={
                        "filepaths": [os.path.join(data_dir, "classtrain.txt")],
                        "split": "train",
                    },
                )
            )

        splits += [
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "filepaths": [
                        os.path.join(data_dir, "republican_only.dev.en"),
                        os.path.join(data_dir, "democratic_only.dev.en"),
                    ],
                    "split": "dev",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "filepaths": [
                        os.path.join(data_dir, "republican_only.test.en"),
                        os.path.join(data_dir, "democratic_only.test.en"),
                    ],
                    "split": "test",
                },
            ),
        ]
        return splits

    def _generate_examples(self, filepaths: List[str], split: str):
        for filepath in filepaths:
            filename = Path(filepath).name
            label = filename.split(".")[0].split("_")[0]
            with open(filepath, encoding="utf-8") as f:
                for key, row in enumerate(f):
                    text = row.strip()
                    if split != "test":
                        # label only exists in train/eval files.
                        text = text.split()
                        label, text = text[0], text[1:]
                        text = " ".join(text)

                    yield (
                        key,
                        {"text": text, "label": label},
                    )


if __name__ == "__main__":
    from tqdm import tqdm

    dataset = PoliticalDataset(config_name="classifier")
    dataset = dataset.as_streaming_dataset()
    print(dataset)
    for row in tqdm(dataset["train"]):
        row["text"] = "hello"