As some of the audio is read in different accents to the main accent used, ideally the different accent audio would have been removed. Doing so would be expected to help with voice quality, reducing the overall amount used and, as a bonus, cutting training time too.