Continuous Speech Recognition

Questions about Thunkable
1

que4
que41990×532 90.7 KB

This is the only way I have been able to achieve continuous speech recognition in Drag and Drop mode. Speech recognition only works until the user goes silent - it captures one sentence at a time.
I have a couple of issues with this:

  1. There is no “partial result” result like in StP mode. So the display screen only updates the transcription once the user has gone quiet. The way I’ve set it up is that the text display updates with extra sentences as the user continues speaking. Is there a more elegant solution and is there a way to achieve instant transcription similar to “partial result”?
  2. When the app is updating the text in between sentences, sometimes the beginning of the next sentence isn’t picked up - is it because it’s not listening when it is combining the variables “recording”" and “more text”? is there a better way to set it up so words in between sentences don’t get missed?
2

Hello @joemoukarzel88skm4pi
Thank you for sharing many details and screenshot

Speech Recognizer’s Partial result is not available on DnD, but this is a good suggestion!
I’ll pass it on to the team.

To avoid creating an infinite loop I suggest the following blocks:

image
image1707×565 50.1 KB

1 Like
3

Thanks for that!
I did not think I was stuck in a loop. I set it up so when the user says “stop recording” or presses the stop button it breaks out of the loop. It’s a walk and journal app that uses questions as prompts, and the user can take as long as they need to expound on their thoughts.

I copied your code to test and I’m still facing the issue with some words in between recordings being missed. I think there’s no getting around the fact that it stops listening and has to be initiated again, so some syllables get missed.

It will be good to see “partial result” in the DnD version of Thunkable. Thanks for your help mate.