Speech input may allow more efficient expression of user goals,
which are best implemented through direct access to the functional
level of the application as opposed to the emulation of an existing
interface.
Tracking state is essential in cases where the user has alternate
means of changing application state (for example, input through a
non-speech mode).