User:OrenBochman/bots/ipa
IPA-BOT
edit- A bot to automate IPA entry generation.
- the spelling.
- a phonemic model.
- all the existing IPA data.
Features
edit- knowledge based version (rule based).
- start with a languages that have simple spelling to sound maps like Hungarian and Swahili.
- add phonemic adjustment
- assimilation
- elision
- data base version (statistical).
- HMM based on input output data.
- use existing text to do.
- per language on/off flag
- check flag - add a template for human checking (for proper nouns).
- hybrid
- use both models and some discriminator
Issues
editQ.A. - train and test on 95% / 5% split of existing annotation per language.
Other Features
edit- poll:
- is there interest in generating TTS voice files for entries?
- is there interest in generating hyphenation as well?
Resources
edit- open source TTS projects with language models, scripts for tts.
- Mbrola
- Sphinx
- Hspell
- CMU dict for English.
- mallet to graphic models.