SubDialogs   |  TOC   |   Nbest Lesson

Tutorial: Shadow Variables

• learn about VoiceXML shadow variables
• explore how to use shadow variables to tweak and tune our code
• learn the importance of using the <log> element

Step 1: creating our initial VoiceXML file

Step 2: Learning what shadow variables can do for us

• $.confidence
  This object will return the level of confidence that the interpreter gets upon a match, ranging from 0.0 to 1.0. Essentially, this shows how sure of the recognition the voice browser is. This can be extarordinarily handy to use when tweaking and tuning your big-daddy grammar for maximum recognition accuracy.


• $.utterance
  Returns the actual utterance the interpreter understood; in the above example, <value expr="F_1$.utterance.F_1"/> would return a value of "coka cola" rather than "coke."


• $.inputmode
  Returns whether the input given by the caller was 'dtmf', or 'voice'.


• $.interpretation.fieldname
  Returns the interpretaion of the matched grammar; that is, whatever your slotname and grammar return is defined as when you write your grammar. You'll note that we need to specify the 'fieldname' property after we specify the 'interpretation' object, which is unlike the other shadow variables. If we insert <log expr="F_1$.interpretation.F_1"/> into the above code, we would likely see:
  
VariableScope::eval: F_1$.interpretation.F_1 resulted in {F_1=coke}
VOICEXML - INFO: {F_1=coke}

  

Step 3. Adding the debug statements and shadow variables into the code

Step 4. Using the <record> shadow variables

• $.duration
  This variable will tell us just how long the recorded message is, in milliseconds:
  
  
RECORD DURATION = 2970
VOICEXML - INFO: RECORD DURATION = 2970



• $.size
  This will return to us the actual filesize of the recorded message, in bytes. This can be useful for determining whether or not to submit the message, thus conserving server resources:
  
  
<if cond="R_1$.size < 0">
  <!--dont do anything-->
<else/>
  <submit next="Myserver.com"/>

  
  An example of the logger output that we could expect to see from using this would look something like this:
  
  
RECORD FILESIZE = 23818
VOICEXML - INFO: RECORD FILESIZE = 23818



• $.termchar
  Okay, maybe not as useful as some of the other ones. If for some reason, we feel the burning need to find out if the caller used a termdigit to end a recording session, this object will return the digit value that was pressed by the caller.
  
  
- DTMF:6
- Ended recording
- { dtype=1 status="INTERRUPTED_BY_DTMF" fileName="C:\temp\vdk1\audiocache\b1\b1_2.wav" detectedTones="6" }
- RECORD TERMCHAR = 6

  

Download the Code!

What we covered:

• Syntax of the two Javascript host objects that define shadow variables
• The most commonly used shadow variables, and how to use them within <value> and <log> statements
• The common <record> shadow variables that define information about a caller's recording
• How to use shadow variables within <log> statements to gain additional information helpful to tuning an application

SubDialogs   |  TOC   |   Nbest Lesson