I'm a university professor, using Descript for research & educational purposes. I'm interested in repurposing the transcription function as a feedback tool to indicate to non-native English speakers the level of clarity of their English. I'm wondering whether Descript could be used as a way to reflect the level of comprehension likely to be achieved by listeners. In other words, where words are mispronounced, I'd like this reflected in the transcript to provide an estimate to students of their comprehensibility. The problem at the moment seems to be that Descript is too good at guessing meaning from context, even with students who have quite strong accents! Is there any way to instruct the AI to transcribe exactly what it 'hears'?