-
Notifications
You must be signed in to change notification settings - Fork 358
Open
Labels
Description
Hi!
Many thanks for your amazing easy to use STT product!
I have yet to learn how to use your text models, but STT seems to work out-of-the-box really fine.
My language in Russian, and you may know that it features a great deal of obscene words, that people commonly use in some contexts.
In our use-case we have to recognize these words as well as ordinary words.
Looks like your language model on top of acoustic model does not know them.
We could add our own language model, but in this case we would need raw acoustic model outputs.
Is is somehow possible with the current API?
Looks like the pywit it just a requests wrapper and 99% of work is done on server-side.
Islanna and adamnsandleadamnsandle and snakers4Islanna, snakers4 and megz15