multispeaker-tts-sinhala.py

import gzip
import json
import re
import os
import datasets

logger = datasets.logging.get_logger(__name__)
_DESCRIPTION = """\\nThis data set contains multi-speaker high quality transcribed audio data for Sinhala. The data set consists of wave files, and a TSV file. 
The file si_lk.lines.txt contains a FileID, which in tern contains the UserID and the Transcription of audio in the file.
The data set has been manually quality checked, but there might still be errors.

Part of this dataset was collected by Google in Sri Lanka and the rest was contributed by Path to Nirvana organization.
"""
_CITATION = """
@inproceedings{Sodimana2018,
  author={Keshan Sodimana and Pasindu {De Silva} and Supheakmungkol Sarin and Oddur Kjartansson and Martin Jansche and Knot Pipatsrisawat and Linne Ha},
  title={{A Step-by-Step Process for Building TTS Voices Using Open Source Data and Frameworks for Bangla, Javanese, Khmer, Nepali, Sinhala, and Sundanese}},
  year=2018,
  booktitle={Proc. The 6th Intl. Workshop on Spoken Language Technologies for Under-Resourced Languages},
  pages={66--70},
  doi={10.21437/SLTU.2018-14},
  url={http://dx.doi.org/10.21437/SLTU.2018-14}
}
"""
_URL = "https://www.openslr.org/30/"
_DATA_URL = "https://huggingface.co/datasets/keshan/wit-dataset/resolve/09d263477614f9fa8c8af0d8ad78f6d4e410a43c/data.tar.gz"
_DATA_FILE_URL = "https://huggingface.co/datasets/keshan/wit-dataset/resolve/09d263477614f9fa8c8af0d8ad78f6d4e410a43c/file_index.tsv"
_LICENSE = "https://www.openslr.org/resources/30/LICENSE.txt"
_LANGUAGES = [
    "si",
]


class SiTTSConfig(datasets.BuilderConfig):
    """BuilderConfig for SiTTS."""

    def __init__(self, *args, **kwargs):
        """BuilderConfig for SiTTS.
        Args:
            languages (:obj:`List[str]`): list of languages to load
            **kwargs: keyword arguments forwarded to super.
        """
        super().__init__(
            *args, **kwargs,
        )


class SiTTS(datasets.GeneratorBasedBuilder):
    """SiTTS, a  manually quality checked, Sinhala multi-speaker TTS corpora."""

    BUILDER_CONFIGS = [SiTTSConfig(languages=[lang]) for lang in _LANGUAGES]
    BUILDER_CONFIG_CLASS = SiTTSConfig

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "id": datasets.Value("string"),
                    "sentence": datasets.Value("string"),
                    "file_path": datasets.Value("string"),
                }
            ),
            supervised_keys=None,
            homepage=_URL,
            citation=_CITATION,
            license=_LICENSE,
        )

    def _split_generators(self, dl_manager):
        abs_path_to_clips = dl_manager.download_and_extract(_DATA_URL)
        abs_path_to_data = dl_manager.download(_DATA_FILE_URL)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "filepath": abs_path_to_data,
                    "path_to_clips": abs_path_to_clips,
                },
            ),
        ]

    def _generate_examples(self, filepath, path_to_clips):
        data_fields = list(self._info().features.keys())
        path_idx = data_fields.index("file_path")

        with open(filepath, encoding="utf-8") as f:
            lines = f.readlines()

            for id_, line in enumerate(lines):
                id_value, sentence = line.strip().split("\t")
                # sentence = re.findall(r'"(.*?)"', line)[0].strip()
                # id_value = re.findall(r"(sin_[^\s]+)", line)[0]
                file_path = "{0}.wav".format(id_value)
                field_values = [id_value, sentence, file_path]

                # set absolute path for wav audio file
                field_values[path_idx] = os.path.join(
                    path_to_clips, field_values[path_idx]
                )

                # if data is incomplete, fill with empty values
                if len(field_values) < len(data_fields):
                    field_values += (len(data_fields) - len(field_values)) * ["''"]

                yield id_, {key: value for key, value in zip(data_fields, field_values)}