GlotSparse.py

# coding=utf-8
# Copyright 2023 The GlotSprase Authors.
# Lint as: python3
"""
                                           GlotSprase
"""

""" This dataset loading script is built based on Hugging Face tutorial and OSCAR-2301's and CulturaX dataset script. """

import os
import collections

import pandas as pd

import datasets

logger = datasets.logging.get_logger(__name__)

_DESCRIPTION = """\
    GlotSprase \
"""

_URL = "https://huggingface.co/datasets/kargaranamir/GlotSparse"

_LICENSE = """
We do not own any of the text from which these data has been extracted.
We license the actual packaging, the metadata and the annotations of these data under the CC BY 4.0.

If you are a website/dataset owner and do not want your data to be included in this corpra, please send us an email at [email protected] .
"""

_CITATION = r"""\
@misc{GlotSparse,
  author = {Kargaran, Amir Hossein},
  title = {GlotSparse Corpus},
  year = {2023},
  publisher = {Github},
  journal = {Github Repository},
  howpublished = {{\\url{https://github.com/kargaranamir/GlotSparse}}},
}
"""

_BASE_DATA_PAT_FORMAT_STR = "{language}/{language}.csv"
_BASE_CHECKSUM_FILE_NAME = "checksum.sha256"


def _languages():
    """Create the sorted dictionary of language codes, and language names.
    Returns:
      The sorted dictionary as an instance of `collections.OrderedDict`.
    """
    langs = {
        "Balochi_Arab": "bal_Arab",
        "Twi_Latn": "twi_Latn",
        "Fanti_Latn": "fat_Latn",
        "South-Azerbaijani_Arab": "azb_Arab",
        "Southern-Kurdish_Arab": "sdh_Arab",
        "Gurani-Arab": "hac_Arab",
        "Southern-Uzbek": "uzs_Arab",
        "Kirmanjki-Latn": "kiu-Latn",
        "Southern-Uzbek_Arab": "uzs_Arab",
        "Gilaki": "glk_Arab",
    }

    langs = {v: k for k, v in langs.items()}
    return collections.OrderedDict(sorted(langs.items()))


class GlotConfig(datasets.BuilderConfig):
    """GlotSprase corpus."""

    def __init__(self, language: str, **kwargs):
        """BuilderConfig for GlotSprase.
        Args:
            language (str): It has to contain 3-letter coded strings following the writing script with an underline in between. For example: "glk_Arab", "fat_Latn".
            **kwargs: Keyword arguments forwarded to super.
        """
        # Validate the language.
        if language not in _languages():
            raise ValueError("Invalid language: %s " % language)

        name = f"{language}"
        description = (
            f"Original {_languages()[language]} GlotSprase dataset from 2023"
        )
        super(GlotConfig, self).__init__(
            name=name, description=description, **kwargs
        )

        # Additional attributes
        self.language = language
        self.base_data_path = _BASE_DATA_PAT_FORMAT_STR.format(language=language)


class Glot(datasets.ArrowBasedBuilder):
    """GlotSprase"""

    BUILDER_CONFIGS = [
        GlotConfig(  # pylint: disable=g-complex-comprehension
            language=language,
            version=datasets.Version("1.0.0"),
        )
        for language in _languages()
    ]
    BUILDER_CONFIG_CLASS = GlotConfig

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "Source": datasets.Value("string"),
                    "Content": datasets.Value("string"),
                    "Length": datasets.Value("int64"),
                    "Script": datasets.Value("string"),
                    "ISO639-3": datasets.Value("string"),
                    "Language": datasets.Value("string"),
                }
            ),
            supervised_keys=None,
            homepage=_URL,
            citation=_CITATION,
            license=_LICENSE,
        )

    def _split_generators(self, dl_manager):

        data_urls = [self.config.base_data_path]
        doc_files = dl_manager.download(
            [url for url in data_urls if url.endswith(".csv")]
        )
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN, gen_kwargs={"doc_files": doc_files}
            ),
        ]


    def _generate_tables(self, doc_files):
        """This function returns the data by iterating on all the files."""
        for doc_i, doc_path in enumerate(doc_files):
            df = pd.read_csv(doc_path)

            for index, row in df.iterrows():
                yield f"{doc_i}_{index}", {
                            "ISO639-3": row["ISO639-3"],
                            "Language": row["Language"],
                            "Content": row["Content"],
                            "Script": row["Script"],
                            "Length": row["Length"],
                            "Source": row["Source"],
                            }