Use corrections as training data

Manual corrections would be valuable data for fine-tuning an ASR model. Ideally the existing model we be automatically optimized to the user based on any corrections they make. Failing that, they could be exported for the user to use as they wish.