question

RafaelCastelo-3180 avatar image
0 Votes"
RafaelCastelo-3180 asked GiftA-MSFT answered

Custom list phrase / vocabulary on batch transcriptions?

Hi,
I need the ability to provide a custom list of phrases for every transcription depending on the customer who will be transcribing a file.
Consequently, I need something like this
https://docs.microsoft.com/en-us/javascript/api/microsoft-cognitiveservices-speech-sdk/phraselistgrammar?view=azure-node-latest

But for batch transcriptions
https://centralus.dev.cognitive.microsoft.com/docs/services/speech-to-text-api-v3-0/operations/CreateTranscription

I can't see it in the documentation, so I would like to know if possible or maybe I am missing something.
Thanks in advance

azure-speech
· 4
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.

Hi, thanks for reaching out. Currently, custom phrase list is only supported via SDK.

0 Votes 0 ·

Hi thanks for your answer,
Then is there a possibility to transcribe long audio files via the sdk. Since I was experimenting issues doing that, and come to see this:
https://stackoverflow.com/questions/65748525/azure-cognitive-services-speech-to-text-large-long-audio-files-sample

So I switched to batch transcription and worked out of the box, but I can't use the custom phrase list feature.
So if there is possible to transcribe long files via sdk would solve my problem.

0 Votes 0 ·

@RafaelCastelo-3180 How long is your file? Batch transcription is indeed recommended for longer files, but PhraseList is currently only available for OnLine transcription (accessible via SDK)

0 Votes 0 ·
Show more comments

1 Answer

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

Hi, quick follow-up. We don't have any limitation of how much we can process via OnLine transcription, we care more about how fast you send the input to us. So, if for example, you send ~ 1x real speech (i.e. without any "fast-forwarding"), we can handle an hour or even more. However, If push too hard, the session may get disconnected due to "buffer overflow". We are working on improving buffer overflow by introducing adaptive throttling very soon. We recommend that you split big audio files to something like 10 min segments (standalone files). The quickstart only works for short utterances, please check out this sample which should support longer files. Hope this helps!


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.