Pre-trained Grapheme-to-Phoneme (G2P) models

This project is aiming to produce high quality phoneme transcriptions as the ground-truth for other projects.

Highlights

• If a word can be found in the given phonemic dictionary, it will use the phonemic transcriptions provided by the dictionary.
• If a word can not be found in the given phonemic dictionary, it will do the convertion with the pre-trained G2P model.
• You don't need to mind the speed of the pre-trained G2P model since all the convertion results of it will be cached.
• You can check the convertion results in the logs at the debug level.

Installation

• Install Sequitur G2P
• Clone the repository and update the submodules

git clone --recursive https://github.com/petronny/g2p

Usage

See demo.py