Comment by michaelmrose

Can you get better performance similar to performance on exams by having the most pertinent data extracted from the dialogue and re-arranged more sensibly?