czech_parliament_plenary_hearings.py

import os
import tarfile

import soundfile as sf

import datasets

_LICENSE = "https://creativecommons.org/licenses/by/4.0/"
_HOMEPAGE = "https://lindat.mff.cuni.cz/repository/xmlui/handle/11234/1-3126"
_DATASET_URL = "https://lindat.mff.cuni.cz/repository/xmlui/bitstream/handle/11234/1-3126/snemovna.tar.xz"

_DESCRIPTION = "Large corpus of Czech parliament plenary sessions, originaly released 2019-11-29 by Kratochvíl Jonáš, Polák Peter and Bojar Ondřej\
            The dataset consists of 444 hours of transcribed speech audio snippets 1 to 40 seconds long.\
            Original dataset transcriptions were converted to true case from uppercase using spacy library."

_CITATION = """\
 @misc{11234/1-3126,
 title = {Large Corpus of Czech Parliament Plenary Hearings},
 author = {Kratochv{\'{\i}}l, Jon{\'a}{\v s} and Pol{\'a}k, Peter and Bojar, Ond{\v r}ej},
 url = {http://hdl.handle.net/11234/1-3126},
 note = {{LINDAT}/{CLARIAH}-{CZ} digital library at the Institute of Formal and Applied Linguistics ({{\'U}FAL}), Faculty of Mathematics and Physics, Charles University},
 copyright = {Creative Commons - Attribution 4.0 International ({CC} {BY} 4.0)},
 year = {2019} } """


class MyAudioDataset(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.0.0")

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "id": datasets.Value("string"),
                    "audio": datasets.features.Audio(),
                    "transcription": datasets.Value("string"),
                    "audio_sampling_rate": datasets.Value("int32"),
                }
            ),
            supervised_keys=None,
            homepage=_HOMEPAGE,
            citation=_CITATION,
            license=_LICENSE
        )

    def _split_generators(self, dl_manager):
        archive_path = dl_manager.download_and_extract(_DATASET_URL)

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"directory": os.path.join(
                    archive_path, "ASR_DATA", "train")},
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={"directory": os.path.join(
                    archive_path, "ASR_DATA", "dev")},
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={"directory": os.path.join(
                    archive_path, "ASR_DATA", "test")},
            ),
        ]

    def _generate_examples(self, directory):
        for root, _, files in os.walk(directory):
            for filename in files:
                if filename.endswith(".wav"):
                    audio_path = os.path.join(root, filename)
                    transcription_path = os.path.join(root, filename + ".trn")

                    # Load audio and transcription
                    audio, sampling_rate = sf.read(audio_path)
                    with open(transcription_path, "r", encoding="utf-8") as f:
                        transcription = f.read().strip()

                    yield f"{audio_path}-{transcription}", {
                        "audio": audio,
                        "transcription": transcription,
                        "audio_sampling_rate": sampling_rate,
                    }