Other Acoustic Models

[fmobrj]

Hi,

I use custom acoustic (.am.bin) and language (.lm.bin) models in Portuguese for Julius. How can I change the voxforge default acoustic and language models in juliusjs so that I can use my own portuguese grammar? What is the role of the file recognizer.data in julius.js?

Best regards, Fabio.

[frequent]

@fmobrj if still relevant: you can build your acoustic model and grammar following the steps outlined in this tutorial. This will give you hmmdefs, tiedlist and you can generate sample.dfa and sample.dict, which you can already pass into juliusjs. The other two file paths are hardcoded in https://github.com/zzmp/juliusjs/blob/master/dist/worker.js#L86, so you can either hardcode your own paths there, or rewrite worker.js to pass in the other two paths, too.

[fmobrj]

Thank you, very much. Sorry for the delay, have been gone for a while!

[khoantv]

@fmobrj if still relevant: you can build your acoustic model and grammar following the steps outlined in this tutorial. This will give you hmmdefs, tiedlist and you can generate sample.dfa and sample.dict, which you can already pass into juliusjs. The other two file paths are hardcoded in https://github.com/zzmp/juliusjs/blob/master/dist/worker.js#L86, so you can either hardcode your own paths there, or rewrite worker.js to pass in the other two paths, too.

Could you show me how to use model of Julius Japanese Dictation-kit (https://github.com/julius-speech/dictation-kit). It uses bccwj.60k.bingram, bccwj.60k.htkdic instead of .dfa and .dic model. Can i convert them into .dfa and .dic model? Thanks a bunch.