cscomm.py

import json

import datasets

from pathlib import Path

import pandas as pd

BASE_DATA_PATH = Path("./data")


class CSCOMMConfig(datasets.BuilderConfig):
    """BuilderConfig for CSCOMM."""

    def __init__(self, key, pretraining=False, data_path="./data", **kwargs):
        """BuilderConfig for CSCOMM.
        Args:
          key: `string`
          **kwargs: keyword arguments forwarded to super.
        """
        # Version history:
        # 0.0.1: Initial version.
        super(CSCOMMConfig, self).__init__(version=datasets.Version("0.0.1"), **kwargs)
        self.key = key
        self.pretraining = pretraining


class CSCOMM(datasets.GeneratorBasedBuilder):
    BUILDER_CONFIGS = [
        CSCOMMConfig(
            name="AP",
            key="ap"
        ),
        CSCOMMConfig(
            name="AP+P",
            key="ap_p"
        ),
        CSCOMMConfig(
            name="AP+J",
            key="ap_j"
        ),
        CSCOMMConfig(
            name="AP+PJ",
            key="ap_pj"
        ),
        CSCOMMConfig(
            name="BA",
            key="ba"
        ),
        CSCOMMConfig(
            name="BA+P",
            key="ba_p"
        ),
        CSCOMMConfig(
            name="BA+J",
            key="ba_j"
        ),
        CSCOMMConfig(
            name="BA+PJ",
            key="ba_pj"
        ),
        CSCOMMConfig(
            name="pretrain-unlabeled",
            key="pt_un",
            pretraining=True
        ),
        CSCOMMConfig(
            name="pretrain-labeled",
            key="pt_la",
            pretraining=True
        ),
        CSCOMMConfig(
            name="pretrain-both",
            key="pt_unla",
            pretraining=True
        ),
    ]

    def _info(self):
        features = {
            "round_id": datasets.Value("string"),
            "source": datasets.Value("string")
        }
        if not self.config.pretraining:
            features["commentary"] = datasets.Value("string")

        return datasets.DatasetInfo(
            features=datasets.Features(features),
        )

    def _split_generators(self, dl_manager):
        dl_dir = dl_manager.download_and_extract({
            "train": f"./data/{self.config.key}/train.csv",
            "valid": f"./data/{self.config.key}/valid.csv",
            "test": f"./data/{self.config.key}/test.csv"
        })

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "data_file": dl_dir["train"],
                    "split": datasets.Split.TRAIN,
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "data_file": dl_dir["valid"],
                    "split": datasets.Split.VALIDATION,
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "data_file": dl_dir["test"],
                    "split": datasets.Split.TEST,
                },
            ),
        ]

    def _generate_examples(self, data_file, split):
        df = pd.read_csv(data_file)
        for i, row in enumerate(df.itertuples()):
            example = {"round_id": row.round_id, "source": row.source}
            if not self.config.pretraining:
                example["commentary"] = row.commentary

            yield i, example