Audio Dataset

speechline.utils.dataset

format_audio_dataset(df, sampling_rate=16000)

Formats Pandas DataFrame as a datasets Dataset. Converts audio path column to audio arrays and resamples accordingly.

Parameters:

Name	Type	Description	Default
df	pd.DataFrame	Pandas DataFrame to convert to Dataset.	required

Returns:

Type	Description
Dataset	datasets' Dataset object usable for batch inference.

Source code in speechline/utils/dataset.py

def format_audio_dataset(df: pd.DataFrame, sampling_rate: int = 16000) -> Dataset:
    """
    Formats Pandas `DataFrame` as a datasets `Dataset`.
    Converts `audio` path column to audio arrays and resamples accordingly.

    Args:
        df (pd.DataFrame):
            Pandas DataFrame to convert to `Dataset`.

    Returns:
        Dataset:
            `datasets`' `Dataset` object usable for batch inference.
    """
    dataset = Dataset.from_pandas(df)
    dataset.save_to_disk(str(config.HF_DATASETS_CACHE))
    saved_dataset = load_from_disk(str(config.HF_DATASETS_CACHE))
    saved_dataset = saved_dataset.cast_column("audio", Audio(sampling_rate=sampling_rate))
    return saved_dataset

prepare_dataframe(path_to_files, audio_extension='wav')

Prepares audio and ground truth files as Pandas DataFrame. Assumes files are of the following structure:

path_to_files
├── langX
│   ├── a.{audio_extension}
│   ├── a.txt
│   └── b.{audio_extension}
│   └── b.txt
└── langY
    └── c.{audio_extension}
    └── c.txt

Parameters:

Name	Type	Description	Default
path_to_files	str	Path to files.	required
audio_extension	str	Audio extension of files to include. Defaults to "wav".	'wav'

Exceptions:

Type	Description
ValueError	No audio files found.

Returns:

Type	Description
pd.DataFrame	DataFrame consisting of: audio (audio path) id language language_code ground_truth

Source code in speechline/utils/dataset.py

def prepare_dataframe(path_to_files: str, audio_extension: str = "wav") -> pd.DataFrame:
    """
    Prepares audio and ground truth files as Pandas `DataFrame`.
    Assumes files are of the following structure:

    ```
    path_to_files
    ├── langX
    │   ├── a.{audio_extension}
    │   ├── a.txt
    │   └── b.{audio_extension}
    │   └── b.txt
    └── langY
        └── c.{audio_extension}
        └── c.txt
    ```

    Args:
        path_to_files (str):
            Path to files.
        audio_extension (str, optional):
            Audio extension of files to include. Defaults to "wav".

    Raises:
        ValueError: No audio files found.

    Returns:
        pd.DataFrame:
            DataFrame consisting of:

        - `audio` (audio path)
        - `id`
        - `language`
        - `language_code`
        - `ground_truth`
    """
    audios = sorted(glob(f"{path_to_files}/*/*.{audio_extension}"))
    audios = [a for a in audios if Path(a).stat().st_size > 0]
    if len(audios) == 0:
        raise ValueError("No audio files found!")

    df = pd.DataFrame({"audio": audios})
    # ID is filename stem (before extension)
    df["id"] = df["audio"].apply(lambda f: Path(f).stem)
    # language code is immediate parent directory
    df["language_code"] = df["audio"].apply(lambda f: Path(f).parent.name)
    df["language"] = df["language_code"].apply(lambda f: f.split("-")[0])
    # ground truth is same filename, except with .txt extension
    df["ground_truth"] = df["audio"].apply(lambda p: Path(p).with_suffix(".txt"))
    df["ground_truth"] = df["ground_truth"].apply(lambda p: open(p).read() if p.exists() else "")
    return df

preprocess_audio_transcript(text)

Preprocesses audio transcript. - Removes punctuation. - Converts to lowercase. - Removes special tags (e.g. GigaSpeech).

Source code in speechline/utils/dataset.py

def preprocess_audio_transcript(text: str) -> str:
    """
    Preprocesses audio transcript.
    - Removes punctuation.
    - Converts to lowercase.
    - Removes special tags (e.g. GigaSpeech).
    """
    tags = [
        "<COMMA>",
        "<PERIOD>",
        "<QUESTIONMARK>",
        "<EXCLAMATIONPOINT>",
        "<SIL>",
        "<MUSIC>",
        "<NOISE>",
        "<OTHER>",
    ]
    chars_to_remove_regex = "[\,\?\.\!\-\;\:\"\“\%\‘\”\�'\’]"
    text = re.sub(chars_to_remove_regex, " ", text).lower().strip()
    text = re.sub(r"\s+", " ", text).strip()
    for tag in tags:
        text = text.replace(tag.lower(), "").strip()
    return text