User Tools

Site Tools


private:pronunciations_for_every_language

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

private:pronunciations_for_every_language [2018/07/13 09:22] (current)
Line 1: Line 1:
 +====== Pronunciations/​ASR/​TTS for every language ======
 +
 +Goals:
 +  * About 128 phones shared across all languages
 +  * aid speech research (recognition and synthesis)
 +  * help under resourced languages survive
 +  * self-funding via selling commercial licenses
 +
 +Resources:
 +  * [[https://​github.com/​espeak-ng/​espeak-ng/​|espeak]]
 +  * [[https://​aclweb.org/​anthology/​P/​P16/​P16-1038.pdf|Grapheme-to-Phoneme Models for (Almost) Any Language]]
 +  * [[https://​bible.is|bible.is]]
 +  * openspeech.net - allow playback and correction
 +
 +Start with the Any Language resources and build acoustic models on bible data.   Use for speech synthesis with Idlak and for ASR with Kaldi. ​ Try to refine models with Bible data.  Then release as crowd sourcing, use and API so that others can pick up on the work.   Set challenges for native speakers.
 +
 +Maybe use for language identification,​ a small WFST for every language connected in parallel.
 +
 +Maybe release a single AM that can be used for any language. ​ Or better, a single ASR system that works in every language.
 +
  
private/pronunciations_for_every_language.txt ยท Last modified: 2018/07/13 09:22 (external edit)