Digital Signal Processing Reference
In-Depth Information
HTTP requests to the Dialog Manager (DM) and receiving VoiceXML
documents in return, it can also work in standalone mode by utilizing internal
applications. For example, an asynchronous signal from a GPS module can
interrupt an ongoing dialog (e.g., a query about restaurant information) when
the vehicle approaches an intersection where a turn is required. Grammars
and lexicons are stored in the client and the server, and the DM specifies the
location of the grammar and the lexicon to use, simply by including a specific
URI in the VoiceXML document. If a server-side grammar and lexicon are
required for the task, they are sent with the VoiceXML document. On the
server side, the DM provides centralized coordination and controls the dialog
according to pre-defined dialog scenarios. In order to create a VoiceXML
dialog that includes dynamic data (such as specific navigation directions), the
DM communicates with external databases through the Internet. External
databases provide various types of static and dynamic information, such as
traffic and parking conditions, nearby restaurant names, and the current
weather forecast. The VoiceXML compiler is a part of the DM and it
compiles the ScenarioXML into VoiceXML format for the VoiceXML
Interpreter. Grammars and lexicons are stored on the server side, and the DM
transmits them along with the VoiceXML output if the client side system does
not already contain them.
3.
EXTENSIONS OF VOICEXML
Layered extensions of VoiceXML have been proposed as a way to realize
straightforward development of VoiceXML applications with dynamic
content [4]. The DialogXML layer was proposed to enable the developer to
write any dialog flow using a state-and-transition model. Although it is
possible to write an equivalent form-filling and if-then style VoiceXML
document by hand, state-and-transition style dialog creation is more efficient
for developers, especially as dialogs grow beyond simple menu selection to
more complex dialogs. The DialogXML compiler translates DialogXML
dialogs into VoiceXML, which is generally much longer (by at least a factor
of two or three) and more difficult to read than the original DialogXML. The
ScenarioXML layer was proposed as a way to specify transitions between
active dialogs, and to support dynamic content retrieved from external
databases and other information sources. Since most information is dynamic
in real-life applications, it is necessary to generate the DialogXML output
with up-to-date information included at run-time. In our system, JSP
technologies are used for that purpose. The higher-level control of the JSP
engine is useful when writing dialog scenarios, since a dialog template can be
