question

SharvaniPopuriSai-8516 avatar image
0 Votes"
SharvaniPopuriSai-8516 asked YutongTie-MSFT answered

Automating Training for Custom Speech Model

I am working on a requirement where I have created a custom speech model for Azure Chatbot. I have uploaded some audio files in https://speech.microsoft.com , created a model and then deployed it manually for getting model id and speech recognition endpoint id. I have done all these steps manually in speech studio, but I want to know if there is any way to automate this process for updating the files.

I have used swagger API https://westus.dev.cognitive.microsoft.com/docs/services/speech-to-text-api-v3-0/operations/CreateModel but when I create datasets and models using this API, I will have to click on "Connect" in speech studio every time I create anything new as below. So, I want to know if adding new files, creating new models and deploying models can be automated with out manually clicking Connect as below, because unless I click Connect the new files are not reflecting in the portal.

112644-image.png


azure-speech
image.png (42.2 KiB)
· 2
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

Thanks for this feedback, I am reaching out to product group to see any automatic refreshing mechanism here to help.

Regards,
Yutong

0 Votes 0 ·

@SharvaniPopuriSai-8516
I have done some research about this issue. Assigning models, datasets. etc. to a project is optional. Customer is welcome to use Speech-to-Text REST API to completely automate the training process. This is API Reference https://centralus.dev.cognitive.microsoft.com/docs/services/speech-to-text-api-v3-0/operations/CopyModelToSubscription


If you would like to keep association with a Speech Studio project,. this is possible also. (I have confirmed that there is a backlog to update Swaggger doc on this).


This is how it's done:


an example of the request body for Create Dataset request (https://centralus.dev.cognitive.microsoft.com/docs/services/speech-to-text-api-v3-0/operations/CreateDataset) with the project reference.


{

            "kind": "Acoustic",

            "contentUrl": "https://contoso.com/location. May point to any publicly readable location: Azure blob, AWS S3, etc.",

            "locale": "en-US",

            "displayName": "My speech dataset name",

            "description": "My speech dataset description",

            "project": {

                           "self": "https://westeurope.api.cognitive.microsoft.com/speechtotext/v3.0/projects/c1c643ae-7da5-4e38-9853-e*********2"

            }

Hope this helps.

Regards,
Yutong
}

1 Vote 1 ·

1 Answer

YutongTie-MSFT avatar image
0 Votes"
YutongTie-MSFT answered

@SharvaniPopuriSai-8516
I have done some research about this issue. Assigning models, datasets. etc. to a project is optional. Customer is welcome to use Speech-to-Text REST API to completely automate the training process. This is API Reference https://centralus.dev.cognitive.microsoft.com/docs/services/speech-to-text-api-v3-0/operations/CopyModelToSubscription


If you would like to keep association with a Speech Studio project,. this is possible also. (I have confirmed that there is a backlog to update Swaggger doc on this).


This is how it's done:


an example of the request body for Create Dataset request (https://centralus.dev.cognitive.microsoft.com/docs/services/speech-to-text-api-v3-0/operations/CreateDataset) with the project reference.


{

            "kind": "Acoustic",

            "contentUrl": "https://contoso.com/location. May point to any publicly readable location: Azure blob, AWS S3, etc.",

            "locale": "en-US",

            "displayName": "My speech dataset name",

            "description": "My speech dataset description",

            "project": {

                           "self": "https://westeurope.api.cognitive.microsoft.com/speechtotext/v3.0/projects/c1c643ae-7da5-4e38-9853-e*********2"

            }

Hope this helps.

Regards,
Yutong
}

· 2
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

Thanks a lot @YutongTie-MSFT , adding project property in the request header works absolutely fine.

0 Votes 0 ·
YutongTie-MSFT avatar image YutongTie-MSFT SharvaniPopuriSai-8516 ·

@SharvaniPopuriSai-8516 Good to hear! It will be very helpful if you can accept this answer to help the community.

Regards,
Yutong

1 Vote 1 ·