Azure speech to text with Diarization receives stop event in case of long pause in audio file

Question

Hi

I am using continuous reading of a wav file and sending in frames to Azure cognitive services API to get the text. I use with Diarization. (

speechsdk.transcription.ConversationTranscriber

)

Everything works fine when audio file has speech in it, that means people speak continously. Otherwise, if the is a long pause, that is people dont speak for longer time, or there is a music in it. Session Stop event gets fired and my program terminates.

Example: Let us take a scenario, where there is a meeting for long 4 hours with short breaks of 15-30 mins in between. Now if I try to use this audio, when ever there is a break for 15 mins, "Session Stop" event gets triggered and my python script ends.

Is there a way to handle this. Obviously, what I am looking for is, irrespective of there is speech in audio, I should not receive Session Stop or Cancelled event.

Is there any such property that I can set?

I tried properties like "Conversation_Initial_Silence_Timeout", but when I read through the documentation, I dont think this serves my purpose.

Answer 1

Hi @Harish A ,

Thank you for using the Microsoft Q&A.

I can understand that you were looking for a method that can allow the maximum duration of silence, before the conversation is considered complete. I will be happy to assist you with this.

You can use the speech_config.set_service_property() method to set any one of the below 2 parameters:

conversationEndSilenceTimeoutMs

(OR)

Speech_SegmentationSilenceTimeoutMs

By setting any one of these 2 properties, you can handle silence and avoid the "session stop" event from getting triggered. Then the code can handle the conversation ending detection timeout.

# Set conversation ending detection timeout (4 hours in seconds)
conversation_ending_detection_timeout = 4 * 60 * 60
    speech_config.set_service_property("conversationEndSilenceTimeoutMs", str(conversation_ending_detection_timeout * 1000), speechsdk.ServicePropertyChannel.UriQueryParameter)

(OR)

# Set conversation ending detection timeout (4 hours in seconds)     conversation_ending_detection_timeout = 4 * 60 * 60     
    speech_config.set_service_property("speechsdk.PropertyId.Speech_SegmentationSilenceTimeoutMs", str(conversation_ending_detection_timeout * 1000), speechsdk.ServicePropertyChannel.UriQueryParameter)

Here is the link, where you can find more details.
How to recognize speech - Speech service - Azure AI services | Microsoft Learn

Hope this helps.

Thank you!

---

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.