Audio transcription

Just found Prodigy and looks very nice. I’m wondering if it is possible to create an annotation task for audio transcription (for speech recognition). Looks like embedding audio is supported however editing the text shown to the annotator is not. In other words I’m looking to create gold-standard speech transcription data. This would require the annotator to be able to edit the text shown by prodigy.

Thanks a lot.

Hi! You could definitely build something like that using a custom HTML template plus a few lines of JavaScript. For example, a textarea that shows some text and an action to update the current task data (slightly simplified example that shows the idea):

<textarea class="textarea">{{text}}</textarea>
<button onClick="updateTask()">Done!</button>
function updateTask() {
    const text = document.querySelector('.textarea').value
    window.prodigy.update({ corrected: text })

The record saved in the database could then look something like this:

{"text": "MLP is cool", "corrected": "NLP is cool"}

That said, it’s definitely true that the Prodigy framework focuses a lot on automation and collecting very structured data for machine learning tasks. So it might not be the best fit if you’re looking for an editor for fully manual audio transcription or to collect mostly free-form written user input.

Thank you Ines.

1 Like