Revisions for Speech Representation