Does Speech-to-text REST API do "at-start" or "continuous" language identification?

Evelyn Williams 0

Is Continuous Language Identification possible through the REST API? We're using the Batch transcription REST API and passing a list of candidateLocales using the languageIdentification parameter.

Does this method of LID only detect the language once at the start of the call? From our transcripts, this appears to be happening. I want to detect the language for every sentence/turn in the audio, as the speakers are code-switching. Is this possible through the REST API?

Thanks!

romungi-MSFT 43,696 Reputation points Microsoft Employee

2023-05-11T06:44:24.26+00:00
@Evelyn Williams Do you see the transcript output with the entire audio transcribed? The usage of languageIdentification parameter with candidateLocales is the correct way to detect multiple languages from the supported list.

"languageIdentification": { "candidateLocales": [ "en-US", "de-DE", "es-ES" ] }

Also, is the audio using multiple channels? For the batch transcription API if multiple speakers are speaking in a mono channel audio file, please set diarizationEnabled parameter to true to set the speaker count to ensure the transcript captures the transcripts of both the speakers.

"diarization": { "speakers": { "minCount": 0, "maxCount": 0 } },

I hope this helps!!
romungi-MSFT 43,696 Reputation points Microsoft Employee

2023-05-14T14:59:26.72+00:00

@Evelyn Williams Did the above suggestion help to get all identified languages in your transcript?
Evelyn Williams 0 Reputation points

2023-05-15T08:06:26.48+00:00

@romungi-MSFT This code block looks like what we're currently doing. What I'm wondering is whether it's possible to do Continuous Language ID via the REST API? Or only at-start Language ID?

Thanks!

Share via

Does Speech-to-text REST API do "at-start" or "continuous" language identification?