mintakaqa.py

import json
from typing import List

import datasets

LANGUAGES = ["ar", "de", "es", "fr", "hi", "it", "ja", "pt"]
DATA_PATH = "mintaka_test.json"


class MintakaQAConfig(datasets.BuilderConfig):
    def __init__(self, language, **kwargs):
        super().__init__(**kwargs)
        self.language = language


class MintakaQA(datasets.GeneratorBasedBuilder):
    BUILDER_CONFIG_CLASS = MintakaQAConfig

    BUILDER_CONFIGS = [
        MintakaQAConfig(name=language, language=language) for language in LANGUAGES
    ]

    def _info(self):
        return datasets.DatasetInfo(
            description="Mintaka: A Complex, Natural, and Multilingual Dataset for End-to-End Question Answering.",
            features=datasets.Features(
                {
                    "id": datasets.Value("string"),
                    "question": datasets.Value("string"),
                    "answer": datasets.Value("string"),
                }
            ),
            homepage="https://github.com/amazon-science/mintaka",
            citation="https://aclanthology.org/2022.coling-1.138",
        )

    def _split_generators(
        self, dl_manager: datasets.DownloadManager
    ) -> List[datasets.SplitGenerator]:
        downloaded_file = dl_manager.download_and_extract(DATA_PATH)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={"filepath": downloaded_file}
            ),
        ]

    def _generate_examples(self, filepath):
        with open(filepath) as f:
            data = json.load(f)
            id_ = 0
            for example in data:
                if example["answer"]["answerType"] == "entity":
                    question = example["translations"][self.config.language]
                    answer = (
                        example["answer"]["answer"][0]["label"][self.config.language]
                        if example["answer"]["answer"]
                        else example["answer"]["mention"]
                    )
                    if not answer or not question:
                        continue

                    yield id_, {"id": id_, "question": question, "answer": answer}

                    id_ += 1