Speech Recognition - Trigger

Question

Hi,

I'm working on Speech Recognition IoT Framework (named SARAH) on top Microsoft Speech Recognition, Kinect1 and Kinect 2 APIs.

I'd like to include Project Oxford Speech Recognition in my framework because Microsoft Speech do not handle wild cards.

For now I'm using the "*" in grammar to query Google Speech. But the results are not very accurate. I assume DeepLearning of Google is more tied to mobile microphone than Project Oxford that gives better results.

Anyway,

- Is there a best pratice to mix grammar and Oxford ?
- Is there trick to always listening for a "trigger" (SARAH do something...) before spaming Oxford servers ?
- I didn't try LUIS (no access) but very curious about handling semantics in French ?

Until now grammar was great for semantics but my community asked for a Chatter bot to handle unknown sentences

Answer 1

Hi Jp

Thanks for asking about LUIS.  Currently, LUIS supports only English, but we're definitely looking to add additional language support to LUIS in the future; when ready, we'll announce new language support to current LUIS users.

Thanks,
Jason

Answer 2

Ok I thought I have seen other language,

I didn't understand does Cortana use Oxford ? Does windows 10 do Speech2Text localy or use a cloud base CNN ?

Anyway is there a magic C# code to detect a trigger like "Hello SARAH" before spaming Oxford server for Speech2Text ?