Where We ve BeenWhere We re Going

Where We've Been ”Where We're Going

Understanding the problems that people have with a system and then being able to correct these problems is the only way to make a speech-recognition application that people can use and want to use. Being able to examine the application in a controlled laboratory environment is a good start, but there is no comparison to seeing how it performs in the real world, by motivated callers who need to use that system.