After a short review of sublanguages used by people communicating their ideas in restricted domains, the implementation of a spoken language database query system is described. The implementation is based on a commercial speech recognition system using a finite-state network language model.