chall.py

import json
import os
import datasets
import soundfile as sf


_DESCRIPTION = "tbd"
_CITATION = "tbd"

_META_FILE = "chall_data.jsonl"


logger = datasets.logging.get_logger(__name__)


class ChallConfig(datasets.BuilderConfig):

    split_into_utterances: bool = False

    def __init__(self, split_into_utterances: bool, **kwargs):
        super(ChallConfig, self).__init__(version=datasets.Version("1.0.0"), **kwargs)
        self.split_into_utterances = split_into_utterances


class Chall(datasets.GeneratorBasedBuilder):

    VERSION = datasets.Version("1.0.0")

    DEFAULT_CONFIG_NAME = "chall_data"

    BUILDER_CONFIGS = [
        ChallConfig(
            name="chall_data",
            split_into_utterances=False
        ),
        ChallConfig(
            name="asr",
            split_into_utterances=True
        )
    ]

    max_chunk_length: int = int

    def __init__(self, *args, max_chunk_length=12, **kwargs):
        super().__init__(*args, **kwargs)
        self.max_chunk_length = max_chunk_length  # max chunk length in seconds

    @property
    def manual_download_instructions(self):
        return (
            "To use the chall dataset you have to download it manually. "
            "TBD Download Instructions. "  # todo
            "Please extract all files in one folder and load the dataset with: "
            "`datasets.load_dataset('chall', data_dir='path/to/folder/folder_name')`"
        )

    def _info(self):

        if self.config.split_into_utterances:
            features = datasets.Features({
                "audio_id": datasets.Value("string"),  # todo maybe shorten to id
                "intervention": datasets.Value("int32"),
                "school_grade": datasets.Value("string"),
                "area_of_school_code": datasets.Value("int32"),
                "background_noise": datasets.Value("bool"),
                "speaker": datasets.Value("string"),
                "words": datasets.features.Sequence(
                    {
                        "start": datasets.Value("float"),
                        "end": datasets.Value("float"),
                        "duration": datasets.Value("float"),
                        "text": datasets.Value("string"),
                    }
                ),
                "audio": datasets.Audio(sampling_rate=16_000)
            })
        else:
            features = datasets.Features({
                "audio_id": datasets.Value("string"),  # todo maybe shorten to id
                "intervention": datasets.Value("int32"),
                "school_grade": datasets.Value("string"),
                "area_of_school_code": datasets.Value("int32"),
                "participants": datasets.features.Sequence(
                    {
                        "pseudonym": datasets.Value("string"),
                        "gender": datasets.Value("string"),
                        "year_of_birth": datasets.Value("int32"),
                        "school_grade": datasets.Value("int32"),
                        "languages": datasets.Value("string"),
                        "estimated_l2_proficiency": datasets.Value("string")
                    }, length=-1
                ),
                "background_noise": datasets.Value("bool"),
                "speakers": datasets.features.Sequence(
                    {
                        "spkid": datasets.Value("string"),
                        "name": datasets.Value("string")
                    }
                ),
                "segments": datasets.features.Sequence(
                    {
                        "speaker": datasets.Value("string"),
                        "words": datasets.features.Sequence(
                            {
                                "start": datasets.Value("float"),
                                "end": datasets.Value("float"),
                                "duration": datasets.Value("float"),
                                "text": datasets.Value("string"),
                            }
                        ),
                    }
                ),
                "audio": datasets.Audio(sampling_rate=16_000)
            })

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            # todo No default supervised_keys (as we have to pass both question and context as input).
            supervised_keys=None,
            homepage="",
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):

        print("_split_generators")

        # todo define splits?

        data_dir = os.path.abspath(os.path.expanduser(dl_manager.manual_dir))

        print(data_dir)

        # todo read ids for splits as we do not separate them by folder

        if not os.path.exists(data_dir):
            raise FileNotFoundError(
                f"{data_dir} does not exist. Make sure you insert a manual dir via `datasets.load_dataset('chall', data_dir=...)` "
                f"that includes files unzipped from the chall zip. Manual download instructions: {self.manual_download_instructions}"
            )
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"filepath": os.path.join(data_dir, "data"), "metafile": os.path.join(data_dir, _META_FILE)},
            ),
            # datasets.SplitGenerator(
            #     name=datasets.Split.TEST,
            #     gen_kwargs={"filepath": os.path.join(data_dir, "data"), "metafile": os.path.join(data_dir, _META_FILE)},
            # ),
            # datasets.SplitGenerator(
            #     name=datasets.Split.VALIDATION,
            #     gen_kwargs={"filepath": os.path.join(data_dir, "data"), "metafile": os.path.join(data_dir, _META_FILE)},
            # ),
        ]

    def _generate_examples(self, filepath, metafile):

        logger.info("generating examples from = %s", filepath)  # todo define logger?
        print("_generate_examples")

        with open(metafile, 'r') as file:
            for line in file:
                data = json.loads(line)

                # load json
                transcript_file = os.path.join(filepath, data["transcript_file"])
                with open(transcript_file, 'r') as transcript:
                    transcript = json.load(transcript)

                audio_id = data['audio_id']
                audio_file_path = os.path.join(filepath, data["audio_file"])

                if self.config.name == "asr":

                    for segment_i, segment in enumerate(transcript["segments"]):

                        id_ = f"{audio_id}_{str(segment_i).rjust(3, '0')}"

                        data["audio_id"] = id_
                        data["speaker_id"] = segment["speaker"]
                        data["words"] = segment["words"]

                        track = sf.SoundFile(audio_file_path)

                        can_seek = track.seekable()
                        if not can_seek:
                            raise ValueError("Not compatible with seeking")

                        sr = track.samplerate
                        start_time = segment["words"][0]["start"]
                        end_time = segment["words"][-1]["end"]

                        start_frame = int(sr * start_time)
                        frames_to_read = int(sr * (end_time - start_time))

                        # Seek to the start frame
                        track.seek(start_frame)

                        # Read the desired frames
                        audio = track.read(frames_to_read)

                        data["audio"] = {"path": audio_file_path, "array": audio, "sampling_rate": sr}

                        yield id_, data
                else:

                    id_ = data["audio_id"]
                    data["speakers"] = transcript["speakers"]
                    data["segments"] = transcript["segments"]

                    audio, samplerate = sf.read(audio_file_path)
                    data["audio"] = {"path": audio_file_path, "array": audio, "sampling_rate": samplerate}

                    yield id_, data