French language support in Kaldi

Question

I am working on Kaldi but there is not info on its webpage about the language which it supports for conversion. Can I use Kaldi for French speech to text conversion.

I need to develop an offline French learning app. I tried PocketSphinx but the accuracy was very bad. Suggestion found was to train the model with your own data to improve it's accuracy. My assumption was that training a complete French language with it's grammer will be difficult and time taking.

-

score 5 · Accepted Answer · answered Aug 14 '19 at 23:44

5

Kaldi itself is only an engine, so its distribution does not include any acoustic or language model. Given enough data (at least 100 hours) you can train your own model with the scripts provided by Kaldi. Alternatively, you can look at the list of pre-built models, but it looks like there is no French model among them.

answered Aug 14 '19 at 23:44

Alexander Solovets

2,447
15
22

Thank you for your time. So is there any other offline ASR available which i can use to develop a french speaking app? – Sumit Vaise Aug 21 '19 at 16:34
There's also pocketsphinx project, but its capabilities are limited compared to Kaldi. – Alexander Solovets Aug 22 '19 at 23:20
Thanks Alexander, I tried PocketSphinx, and yes you are right its capabilities are limited. – Sumit Vaise Aug 23 '19 at 03:07

score 3 · Answer 2 · answered May 02 '20 at 07:43

I was in the same situation last year.

Recently I found : vosk-api

It's still under development but some modules are already available. I tested python with websockets (docker image), and the android version. Some one have even build a Wasm version

Last but not least, the repo links to French models provided by Paul Guyot

Enjoy.

French language support in Kaldi

2 Answers2