Each dialog step must be configured to perform the specific task required by the dialog flow. The
majority of dialog steps handle an interaction with the caller and typically perform the following tasks:
Play out an initial prompt; for example, Please select one of the following options: Red, Blue or
•
Yellow (this is configured on the Prompts page of the Edit Step dialog box, see Figure above)
Recognize the caller's utterance by matching it against a grammar. The grammar may be
•
supplied by the component (as in the case of Date, Time and Money recognitions) or it may be
supplied by the developer. Some components allow you to specify the grammar as a series of
'choices' (typically configured on the Choices and Results page, see the Figure above).
Other components allow you to specify an external grammar, written in GSL or JSGF, and
stored in a separate file (typically configured on the Grammar Mapping page).
(Optionally) repeat the result back to the caller and ask him or her to confirm it. (Both the
•
prompt used and the circumstances under which confirmation is required are configured on
the Confirmation page, see the Figure above).
Handle other behavior by the caller, for example:
•
disconfirmation, by playing out a prompt that repeats the initial request to the caller; for
o
example, Please repeat your selection: red, blue or yellow. (This is configured on the
Misrecognition page, see the Figure above)
silence, by playing out a prompt such as, I'm sorry I didn't hear that, then repeating the
o
request to the caller by playing out the Misrecognition prompt. (This is configured on
the No Input page, see the Figure above)
nonsense, by playing out a prompt such as, I'm sorry I didn't understand that, then
o
repeating the request to the caller by playing out the Misrecognition prompt. (This is
configured on the No Match page).
For each recognition it performs, the dialog step may also re-configure the speech recognizer.
•
(This is set up in the Voice Properties dialog box, see page 47).
A dialog step may also be set to repeat some of these tasks up to a specified maximum
•
number of times. This is configured on the Settings page.
The following examples illustrate the configuration of a typical component; other components may
differ in detail.
Speech Applications Builder Configurator User Guide • May 15, 2004 • page 43 of 116