ctkfacts_nli.py

import os
import pathlib
from typing import overload
import datasets
import json

from datasets.info import DatasetInfo

_VERSION = "0.0.6"

_URL= "data/"

_URLS = {
    "train": _URL + "train.jsonl",
    "validation": _URL + "validation.jsonl",
    "test": _URL + "test.jsonl"
}

_DESCRIPTION = """\
CtkFactsNLI is a NLI version of the Czech CTKFacts dataset
"""

_CITATION = """\
@article{DBLP:journals/corr/abs-2201-11115,
  author    = {Jan Drchal and
               Herbert Ullrich and
               Martin R{\'{y}}par and
               Hana Vincourov{\'{a}} and
               V{\'{a}}clav Moravec},
  title     = {CsFEVER and CTKFacts: Czech Datasets for Fact Verification},
  journal   = {CoRR},
  volume    = {abs/2201.11115},
  year      = {2022},
  url       = {https://arxiv.org/abs/2201.11115},
  eprinttype = {arXiv},
  eprint    = {2201.11115},
  timestamp = {Tue, 01 Feb 2022 14:59:01 +0100},
  biburl    = {https://dblp.org/rec/journals/corr/abs-2201-11115.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}
"""

datasets.utils.version.Version
class CtkfactsNli(datasets.GeneratorBasedBuilder):
    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "id": datasets.Value("int32"),
                    "label": datasets.ClassLabel(names=["REFUTES", "NOT ENOUGH INFO", "SUPPORTS"]),
                    # datasets.features.Sequence({"text": datasets.Value("string"),"answer_start": datasets.Value("int32"),})
                    "evidence": datasets.Value("string"),
                    "claim": datasets.Value("string"),
                }
            ),
            # No default supervised_keys (as we have to pass both question
            # and context as input).
            supervised_keys=None,
            version=_VERSION,
            homepage="https://fcheck.fel.cvut.cz/dataset/",
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager: datasets.DownloadManager):
        downloaded_files = dl_manager.download_and_extract(_URLS)

        return [
            datasets.SplitGenerator(datasets.Split.TRAIN, {
                "filepath": downloaded_files["train"]
            }),
            datasets.SplitGenerator(datasets.Split.VALIDATION, {
                "filepath": downloaded_files["validation"]
            }),
            datasets.SplitGenerator(datasets.Split.TEST, {
                "filepath": downloaded_files["test"]
            }),
        ]

    def _generate_examples(self, filepath):
        """This function returns the examples in the raw (text) form."""
        key = 0
        with open(filepath, encoding="utf-8") as f:
            for line in f:
                datapoint = json.loads(line)
                yield key, {
                    "id": datapoint["id"],
                    "evidence": " ".join(datapoint["evidence"]),
                    "claim": datapoint["claim"],
                    "label": datapoint["label"]
                }
                key += 1