speech-dataset-parser

Library to parse speech datasets stored in a generic format based on TextGrids. A tool (CLI) for converting common datasets like LJ Speech into a generic format is included. Speech datasets consists of pairs of .TextGrid and .wav files. The TextGrids need to contain a tier which has each symbol separated in an interval, e.g., T|h|i|s| |i|s| |a| |t|e|x|t|.

Generic Format

The format is as follows: {Dataset name}/{Speaker name};{Speaker gender};{Speaker language}[;{Speaker accent}]/[Subfolder(s)]/{Recordings as .wav- and .TextGrid-pairs}

Example: LJ Speech/Linda Johnson;2;eng;North American/wavs/...

Speaker names can be any string (excluding ; symbols). Genders are defined via their ISO/IEC 5218 Code. Languages are defined via their ISO 639-2 Code (bibliographic). Accents are optional and can be any string (excluding ; symbols).

Installation

pip install speech-dataset-parser --user

Library Usage

from speech_dataset_parser import parse_dataset

entries = list(parse_dataset({folder}, {grid-tier-name}))

The resulting entries list contains dataclass-instances with these properties:

• symbols: Tuple[str, ...]: contains the mark of each interval
• intervals: Tuple[float, ...]: contains the max-time of each interval
• symbols_language: str: contains the language
• speaker_name: str: contains the name of the speaker
• speaker_accent: str: contains the accent of the speaker
• speaker_gender: int: contains the gender of the speaker
• audio_file_abs: Path: contains the absolute path to the speech audio
• min_time: float: the min-time of the grid
• max_time: float: the max-time of the grid (equal to intervals[-1])

CLI Usage

usage: dataset-converter-cli [-h] [-v] {convert-ljs,convert-l2arctic,convert-thchs,convert-thchs-cslt,restore-structure} ...

This program converts common speech datasets into a generic representation.

positional arguments:
  {convert-ljs,convert-l2arctic,convert-thchs,convert-thchs-cslt,restore-structure}
                                        description
    convert-ljs                         convert LJ Speech dataset to a generic dataset
    convert-l2arctic                    convert L2-ARCTIC dataset to a generic dataset
    convert-thchs                       convert THCHS-30 (OpenSLR Version) dataset to a generic dataset
    convert-thchs-cslt                  convert THCHS-30 (CSLT Version) dataset to a generic dataset
    restore-structure                   restore original dataset structure of generic datasets

optional arguments:
  -h, --help                            show this help message and exit
  -v, --version                         show program's version number and exit

CLI Example

# Convert LJ Speech dataset with symbolic links to the audio files
dataset-converter-cli convert-ljs \
  "/data/datasets/LJSpeech-1.1" \
  "/tmp/ljs" \
  --tier "Symbols" \
  --symlink

Dependencies

• tqdm
• TextGrid>=1.5
• ordered_set>=4.1.0
• importlib_resources; python_version < '3.8'

Roadmap

• Supporting conversion of more datasets
• Adding more tests

Contributing

If you notice an error, please don't hesitate to open an issue.

Development setup

# update
sudo apt update
# install Python 3.7, 3.8, 3.9, 3.10 & 3.11 for ensuring that tests can be run
sudo apt install python3-pip \
  python3.7 python3.7-dev python3.7-distutils python3.7-venv \
  python3.8 python3.8-dev python3.8-distutils python3.8-venv \
  python3.9 python3.9-dev python3.9-distutils python3.9-venv \
  python3.10 python3.10-dev python3.10-distutils python3.10-venv \
  python3.11 python3.11-dev python3.11-distutils python3.11-venv
# install pipenv for creation of virtual environments
python3.8 -m pip install pipenv --user

# check out repo
git clone https://github.com/stefantaubert/speech-dataset-parser.git
cd speech-dataset-parser
# create virtual environment
python3.8 -m pipenv install --dev

Running the tests

# first install the tool like in "Development setup"
# then, navigate into the directory of the repo (if not already done)
cd speech-dataset-parser
# activate environment
python3.8 -m pipenv shell
# run tests
tox

Final lines of test result output:

py37: commands succeeded
py38: commands succeeded
py39: commands succeeded
py310: commands succeeded
py311: commands succeeded
congratulations :)

License

MIT License

Acknowledgments

Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project-ID 416228727 – CRC 1410

Citation

If you want to cite this repo, you can use this BibTeX-entry generated by GitHub (see About => Cite this repository).

Changelog

• v0.0.5 (unreleased)
  • Added:
    • Added option to parse LJ Speech --use-un-normalized-text
• v0.0.4 (2023-01-12)
  • Added:
    • Added support to parse OpenSLR THCHS-30 version
    • Added returning of an exit code
  • Changed:
    • Changed default command to be parsing the OpenSLR version for THCHS-30 by renaming the previous command to convert-thchs-cslt
• v0.0.3 (2023-01-02)
  • added option to restore original file structure
  • added option to THCHS-30 to opt in for adding of punctuation
  • change file naming format to numbers with preceding zeros
• v0.0.2 (2022-09-08)
  • added support for L2Arctic
  • added support for THCHS-30
• v0.0.1 (2022-06-03)
  • Initial release