A preference-based semantics for CTD reasoning