|
VoiceXML provides shadow variables accessible within a page to
access recognition results such as application.lastresult$. While this includes a mechanism for presenting multiple possible
semantic matches (n-best results whose presence can be determined
by inspecting application.lastresult$.length), unfortunately the specification
does not describe how to present a result that contains multiple interpretations
for a given semantic match.
Avaya Experience Portal addresses
the issue of a result containing multiple interpretations for a given
semantic match by exposing the shadow variable application.lastresult$.interpretation$. To test for the presence of multiple interpretations in a result,
examine application.lastresult$.interpretation$.length.
If the VoiceXML page containing that grammar uses the maxnbest property
of 2, then the recognition results arising from a caller saying star
gazer
would be accessed as follows in the VoiceXML specification:
application.lastResult$.length 2
application.lastResult$[0].confidence 0.83
application.lastResult$[0].utterance Stargazer
application.lastResult$[0].interpretation eyes
application.lastResult$[1].confidence 0.12
application.lastResult$[1].utterance starchaser
application.lastResult$[1].interpretation legsEven though the recognition results contain the additional interpretations look
and run
, the VoiceXML specification makes no provisions
for making those additional interpretations available to the application.
To accomplish this, Avaya Experience Portal extends
this list of shadow variables as:
application.lastResult$[0].interpretation$.length 2
application.lastResult$[0].interpretation$[0] eyes
application.lastResult$[0].interpretation$[1] look
application.lastResult$[1].interpretation$.length 2
application.lastResult$[1].interpretation$[0] legs
application.lastResult$[1].interpretation$[1] run