startimes_luganda_recording.py

import os
import csv
import datasets
from .release_stats import STATS

_DATASET_DICT = {
    "Aldrine": ["20240730", "20240801", "20240805", "20240812"],
    "Nabbuto": ["20240730", "20240801", "20240806", "20240812", "20240819", "20240826"],
    "Nakawunde": ["20240819", "20240826"],
    "Namala": ["20240730", "20240807", "20240812", "20240819", "20240826"],
    "William": ["20240730"],
    "Solomon": ["20240805", "20240812", "20240819", "20240826"],
}

_HOMEPAGE = "https://huggingface.co/Loyage/startimes_luganda_recording"

_DATA_URL = "https://huggingface.co/datasets/Loyage/startimes_luganda_recording/resolve/main/data"


class StartimesLugandaRecordingConfig(datasets.BuilderConfig):
    """BuilderConfig for StartimesLugandaRecording."""

    def __init__(self, name, version, **kwargs):
        self.language = "Luganda"

        description = (
            f"Luganda Speaking Voice Datasets recorded by Startimes. "
            f"All rights reserved by Startimes. "
        )

        super(StartimesLugandaRecordingConfig, self).__init__(
            name=name,
            version=datasets.Version(version),
            description=description,
            **kwargs,
        )


class StartimesLugandaRecording(datasets.GeneratorBasedBuilder):
    BUILDER_CONFIGS = [
        StartimesLugandaRecordingConfig(
            name="Luganda",
            version="1.0.0",
        )
    ]

    def _info(self):
        return datasets.DatasetInfo(
            description="Luganda Speaking Voice Datasets recorded by Startimes. All rights reserved by Startimes.",
            features=datasets.Features(
                {
                    "path": datasets.Value("string"),
                    "sentence": datasets.Value("string"),
                    "speaker": datasets.Value("string"),
                }
            ),
        )

    def _split_generators(self, dl_manager):
        dl_manager.download_config.ignore_url_params = True
        archive_paths = {}
        metadata_path = dl_manager.download_and_extract(f"{_DATA_URL}/metadata.tsv")

        local_extracted_archive_paths_list = []
        for speaker, dates in _DATASET_DICT.items():
            audio_urls = {}
            audio_urls["train"] = []
            for date in dates:
                audio_urls["train"].append(f"{_DATA_URL}/{date}-{speaker}.tar")
            archive_paths = dl_manager.download(audio_urls)
            local_extracted_archive_paths = (
                dl_manager.extract(archive_paths) if not dl_manager.is_streaming else {}
            )
            local_extracted_archive_paths_list.append(local_extracted_archive_paths)

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "local_extracted_archive_paths": local_extracted_archive_paths[0].get(
                        "train"
                    ),
                    "archives": [
                        dl_manager.iter_archive(path)
                        for path in archive_paths.get("train")
                    ],
                    "meta_path": metadata_path,
                },
            )
        ]

    def _generate_examples(self, local_extracted_archive_paths, archives, meta_path):
        meta_data = {}
        with open(meta_path, "r", encoding="utf-8") as f:
            reader = csv.DictReader(f, delimiter="\t")
            for row in reader:
                meta_data[row["path"]] = row

        for i, archive in enumerate(archives):
            for path, file_obj in archive:
                if path in meta_data:
                    result = dict(meta_data[path])
                    path = (
                        os.path.join(local_extracted_archive_paths, path)
                        if local_extracted_archive_paths
                        else path
                    )
                    result["audio"] = {"path": path, "bytes": file_obj.read()}
                    result["path"] = path
                    yield path, result