View article

A voice processing system receives spoken input from a user, which is then subjected to speech recognition to convert the spoken input into a text equivalent. At least two information elements are identified in the text equivalent, each having an associated uncertainty. The system selects a prompt according to which information element has the greatest uncertainty, and this is then played out to the user. Typically the outgoing prompt is arranged to emphasise, via word order, volume, or some other mechanism, the information element in greatest doubt, thereby bringing it to the particular attention of the user, should correction be required.