. "For example, it is significantly easier to interpret an incoming voice command that is expected to consist only of numbers or of states of theUnited States compared to voice data that might consist of names or, particularly of street addresses." . . .