the paper mentions that CLP comes with a default set of schemas, but you can also provide your own rules for better compression and faster search