Open Noitidart opened 6 years ago
Custom language model is usually provided with paid services (e.g., IBM Watson). The default speech recognition, to my knowledge, does not provide this feature.
However, you might calculate the closest word of the recognized word using EDIT distance. Although this is a rough approximation, I guess it should work in a practical level.
I also think that this kind of custom tuning is something which would not be provided in this library level.
Does it answer your question?
On Mon, 1 Jan 2018 at 6:00 AM Noitidart notifications@github.com wrote:
I have two apps that use voice. The first one the users are saying one of 300 items. Is it possible to limit the voice recognition to only return closest elements to ones in an array?
In another application, the users are saying a single number, or multiple numbers. Anyyay to limit to just numbers?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/wenkesj/react-native-voice/issues/53, or mute the thread https://github.com/notifications/unsubscribe-auth/ABQzu2YocwiU78550-EUi9aol5BLPvzqks5tF_XwgaJpZM4RP7uR .
--
Jeungmin Oh (오정민) | CEO & Founder HayanMind Inc.
Try Mimiking app on iPhone https://goo.gl/Xrr3yz & Android https://goo.gl/k77XVW
Darn. Thnks very much for your quick reply. Yes it answers very well, I understand that it makes sense that this library doesnt support it because this feature is paid. If it ever comes out for free it would be awesome if this was available. Thanks for leaving this open for me to close.
@Noitidart For your information, for the paid services, you can even add more words on your own. For example, usually proper nouns like a person's name or brand name. I would love to help support those features if the default speech recognition library provides those features in the future.
Thank you sir very much for such a positive and optimistic reply! :) It would be awesome if we have offline support for that! :) IBM Watson is requiring internet connection. :(
@Noitidart Did you solved your problem?
I have two apps that use voice. The first one the users are saying one of 300 items. Is it possible to limit the voice recognition to only return closest elements to ones in an array?
In another application, the users are saying a single number, or multiple numbers. Anyway to limit to just numbers?