squad_tr.py

# coding=utf-8
# Copyright 2020 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# This file is based off of the dataset loader script for the original SQuAD2.0
# dataset.
#
#     https://huggingface.co/datasets/squad_v2

# Lint as: python3
"""SQuAD-TR Dataset"""


import json

import datasets
from datasets.tasks import QuestionAnsweringExtractive


logger = datasets.logging.get_logger(__name__)

_HOMEPAGE = "https://github.com/boun-tabi/squad-tr"

_CITATION = """\
@article{
  budur2023squadtr,
  title={Building Efficient and Effective OpenQA Systems for Low-Resource Languages},
  author={todo},
  journal={todo},
  year={2023}
}
"""

_DESCRIPTION = """\
SQuAD-TR is a machine translated version of the original SQuAD2.0 dataset into
Turkish.
"""

_VERSION = "1.0.0"

_DATA_URL = _HOMEPAGE + "/raw/beta/data"
_DATA_URLS = {
    "default": {
        "train": f"{_DATA_URL}/squad-tr-train-v{_VERSION}.json.gz",
        "dev": f"{_DATA_URL}/squad-tr-dev-v{_VERSION}.json.gz",
    },
    "excluded": {
        "train": f"{_DATA_URL}/squad-tr-train-v{_VERSION}-excluded.json.gz",
        "dev": f"{_DATA_URL}/squad-tr-dev-v{_VERSION}-excluded.json.gz",
    }
}


class SquadTRConfig(datasets.BuilderConfig):
    """BuilderConfig for SQuAD-TR."""

    def __init__(self, **kwargs):
        """BuilderConfig for SQuAD-TR.
        Args:
          **kwargs: keyword arguments forwarded to super.
        """
        super(SquadTRConfig, self).__init__(**kwargs)


class SquadTR(datasets.GeneratorBasedBuilder):
    """SQuAD-TR: Machine translated version of the original SQuAD2.0 dataset into Turkish."""

    VERSION = datasets.Version(_VERSION)

    BUILDER_CONFIGS = [
        SquadTRConfig(
            name="default",
            version=datasets.Version(_VERSION),
            description="SQuAD-TR default version.",
        ),
        SquadTRConfig(
            name="excluded",
            version=datasets.Version(_VERSION),
            description="SQuAD-TR excluded version.",
        ),
        SquadTRConfig(
            name="openqa",
            version=datasets.Version(_VERSION),
            description="SQuAD-TR OpenQA version.",
        ),
    ]

    DEFAULT_CONFIG_NAME = "default"

    def _info(self):

        # We change the contents of the "answers" field based on the
        # configuration selected. Specifically, we are excluding the
        # "answer_start" field for the "excluded" and "openqa" configurations.
        if self.config.name in ["excluded", "openqa"]:
            answers_feature = datasets.features.Sequence({
                "text": datasets.Value("string"),
            })
        else:
            answers_feature = datasets.features.Sequence({
                "text": datasets.Value("string"),
                "answer_start": datasets.Value("int32"),
            })
  
        # Constructing our dataset features.
        features = datasets.Features({
            "id": datasets.Value("string"),
            "title": datasets.Value("string"),
            "context": datasets.Value("string"),
            "question": datasets.Value("string"),
            "answers": answers_feature
        })

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=None,
            homepage=_HOMEPAGE,
            citation=_CITATION,
            task_templates=[
                QuestionAnsweringExtractive(question_column="question", context_column="context", answers_column="answers")
            ],
        )

    def _split_generators(self, dl_manager):

        # If the configuration selected is "default" or "excluded", we directly
        # load the files from the URLs in _DATA_URLS. For the "openqa"
        # configuration, we combine the datapints from the two different files
        # used in the "default" and "excluded" configurations.
        if self.config.name == "openqa":
            default_files = dl_manager.download_and_extract(_DATA_URLS["default"])
            excluded_files = dl_manager.download_and_extract(_DATA_URLS["excluded"])
            train_file_paths = [default_files["train"], excluded_files["train"]]
            dev_file_paths = [default_files["dev"], excluded_files["dev"]]

            return [
                datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath_list": train_file_paths}),
                datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath_list": dev_file_paths}),
            ]
        else: 
            config_urls = _DATA_URLS[self.config.name]
            downloaded_files = dl_manager.download_and_extract(config_urls)

            return [
                datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": downloaded_files["train"]}),
                datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": downloaded_files["dev"]}),
            ]


    def _generate_examples(self, filepath=None, filepath_list=None):
        """This function returns the examples in the raw (text) form."""
        assert filepath or filepath_list
        if filepath:
            filepath_list = [filepath]

        # Combining the generators for the different filepaths
        generators = [self._generate_examples_from_filepath(f) for f in filepath_list]
        for generator in generators:
            for element in generator:
                yield element

    def _generate_examples_from_filepath(self, filepath):
        logger.info("generating examples from = %s", filepath)
        key = 0
        with open(filepath, encoding="utf-8") as f:
            squad = json.load(f)
            for article in squad["data"]:
                title = article.get("title", "")
                for paragraph in article["paragraphs"]:
                    context = paragraph["context"]  # Do not strip leading blank spaces GH-2585
                    for qa in paragraph["qas"]:
                        # Constructing our answers dictonary. Note that the
                        # answers_dictionary won't include the answer_start
                        # field in the "excluded" and "openqa" modes.
                        answers_dictionary = {
                            "text": [answer["text"] for answer in qa["answers"]],
                        }
                        if self.config.name not in ["excluded", "openqa"]:
                            answers_dictionary["answer_start"] = [answer["answer_start"] for answer in qa["answers"]]

                        # Constructing our datapoint
                        datapoint = {
                            "title": title,
                            "context": context,
                            "question": qa["question"],
                            "id": qa["id"],
                            "answers": answers_dictionary,
                        }

                        # Features currently used are "context", "question", and "answers".
                        # Others are extracted here for the ease of future expansions.
                        yield key, datapoint
                        key += 1