FASST Mining: Discovering Frequently Changing Semantic Structure from Versions of Unordered XML Documents