Using Azure SR with Custom Speech

[Eduardo Hermida]

Hi all

Has anybody used Azure SR plugin with custom speech? My main doubt is about how to set credentials to use an specific custom project. 

Thanks in advance

[Wilmar Pérez]

Hi Eduardo,

Thanks for bringing that topic up. I had no idea about custom speech. It seems to be quite useful!  That being said, I do not think you need any special set of credential. After you upload your data and you start testing, you call the standard speech to text API:

https://docs.microsoft.com/en-us/azure/cognitive-services/speech-service/how-to-custom-speech. I am sure you cheked that one up already, so, am I missing something?

Hasta pronto,

Wilmar

--
You received this message because you are subscribed to the Google Groups "UniMRCP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unimrcp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/unimrcp/212f7ba9-f333-40d9-83e1-b801569e77a9%40googlegroups.com.

--

--------------------------------------------------------

Wilmar Pérez 

[Eduardo Hermida]

Hi Wilmar

On point 6 of the link you've just sent it says: "Create a custom endpoint for your speech-to-text model and use it in your applications, tools, or products."

After creating this inside my custom speech project I can see the endpoint associated with my model and it is:

https://westeurope.stt.speech.microsoft.com/speech/recognition/conversation/cognitiveservices/v1?cid={endpoint id}

So my main doubt is how to indicate unimrcp that it has to use this endpoint. Inside Azure SR install guide, when talking about credentials it says...

Note that the field service-endpoint is supported since Azure SR 1.9.0 and must be explicitly specified when a custom speech endpoint is used.

So, my guess is that maybe I have to put the whole URL in this point inside credential json:

service-endpoint:"https://westeurope.stt.speech.microsoft.com/speech/recognition/conversation/cognitiveservices/v1?cid={endpoint id}"

Regards

El lunes, 13 de enero de 2020, 14:53:39 (UTC+1), Wilmar Pérez escribió:

Hi Eduardo,

Thanks for bringing that topic up. I had no idea about custom speech. It seems to be quite useful!  That being said, I do not think you need any special set of credential. After you upload your data and you start testing, you call the standard speech to text API:

https://docs.microsoft.com/en-us/azure/cognitive-services/speech-service/how-to-custom-speech. I am sure you cheked that one up already, so, am I missing something?

Hasta pronto,

Wilmar

Le lun. 13 janv. 2020 à 04:37, Eduardo Hermida <eduhe...@gmail.com> a écrit :

Hi all

Has anybody used Azure SR plugin with custom speech? My main doubt is about how to set credentials to use an specific custom project. 

Thanks in advance

--
You received this message because you are subscribed to the Google Groups "UniMRCP" group.

To unsubscribe from this group and stop receiving emails from it, send an email to uni...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/unimrcp/212f7ba9-f333-40d9-83e1-b801569e77a9%40googlegroups.com.

--

--------------------------------------------------------

Wilmar Pérez 

[Arsen Chaloyan]

Hi Eduardo,

Your guess is correct. A full URI, including the cid attribute, needs to be specified with the service endpoint. There is one issue, though, that is commonly disregarded.

The recognition mode is supposed to be "interactive" but not "conversation". So your service endpoint would look like

service-endpoint:"https://westeurope.stt.speech.microsoft.com/speech/interactive/conversation/cognitiveservices/v1?cid={endpoint id}"

To unsubscribe from this group and stop receiving emails from it, send an email to unimrcp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/unimrcp/0a94687f-5f1f-492b-b709-7c6dea1a704a%40googlegroups.com.

--

Arsen Chaloyan
Author of UniMRCP
http://www.unimrcp.org