The case for an interval-based representation of linguistic truth