User's speech input will not be ambiguous but rather clear and relevant to the task in hand, although freedom allows many syntactic structures, and many words for the same desired system response. Therefore inability to handle variety will lead to low recognition rates rather than ambiguity [32].
There are therefore two ways to maximise the probability that a NLP system will understand each input; get the computer to understand every possible variety of input, or curtail the variety. Since the first is not feasible, although as time passes, more and more freedom is given to users, the second should be embraced. There are two ways of curtailing possible inputs.
It can be done overtly, where the user is given a list of phrases or words. This is frustrating for the user, makes words and phrases hard to remember and is far too cumbersome.
The second way in which to lessen the user's variety is to use covert methods. It has been said that when creating natural language communications, people simply ``stitch together swatches of text that [they] have heard before''.