For synonymous idiolects, a mapping table is built, while for superfluous idiolects, an insertion model is used to find suitable positions for insertion.