Whenever possible, speech-recognition system designers try to keep prompt questions simple ”"Would you like the red, yellow, or green car?" or "What should we change ”the time, date, or amount?" However, there are times when the questions are unavoidably more complex. Often, the most complex questions ask the caller to choose between two options that need to be described in more than a few words. The following example shows one way to handle these situations. Like all financial services systems, the investment company's system enables the caller to identify and correct an error before a transaction is executed. After the user has notified the system that something is incorrect, the system could reply
It sounds like a reasonable and easy to understand question, doesn't it? The problem is that callers wouldn't know how to respond. Do they say, "The month that I want it to resume," "the resuming month," or perhaps just the word "resume?" The designer of the system solved this problem by collapsing the language structure of the question, making it easier for callers to know how to respond.
This works because the context of "automatic payment" has already been established in a prior statement (callers are, after all, changing something they have already identified as incorrect). |