I want to create a virtual "Command Mode" within WSR to provide improved recognition for specific commands.  Can this be done by creating a custom language model via the Windows Dictation Resource Kit?

It has been my experience with WSR that it works good for some people, and not so good for others.  Unlike DNS, there are not very many parameters to adjust to improve performance for persons having trouble.  One definite weakness of WSR is its lack of alternate recognition modes.  Would creating a custom language model allow a user to reduce the active vocabulary to a limited number of commands for performing specific tasks?  Or am I misunderstanding how these language models are implemented?

Darren