Predicting mistakes made by humans during audio/video transcription, giving them auto-complete suggestions, classifying severity of changes, predicting the difficulty level based on audio characteristics.