We study edit intentions in text revisions by systematically analyzing prior edit intention research and developing UniT, a unified edit intention taxonomy that integrates heterogeneous taxonomies across domains to support consistent annotation, cross-domain analysis, and downstream revision understanding applications.