25 January 2002

Major New Features

• We provide support for voice enrollment through our new tag <bevocal:enroll>. You can use your voice to enroll phrases and later use your enrolled grammar to recognize the enrolled phrases.
• We now allow you to switch the TTS server for your application on the fly. You can switch the TTS voice for an entire application or dialog by setting the property bevocal.voice.name. Valid values are realspeak_us_female and articulator_uk_male. Based on the scope where the property is set, all the prompts in that scope are played with the chosen tts server. For finer granularity, you can use <voice name="voicename"> inside a <prompt> tag to switch tts server for specific prompts. The voice that our interpreter has always used is realspeak_us_female, and this will continue to be the default if you do not specify a particular voice. We plan to add more TTS servers in the future.
  
• We now provide beta-level support for Java Speech Grammar Format (JSGF) grammars, with several limitations. We do not yet support the import statement or semantic tags, and we do not currently handle inline grammar fragments. If you use inline JSGF grammars, you must use a complete, syntactically-correct JSGF grammar, e.g.
  <grammar>
  <![CDATA[
  #JSGF 1.0;
  grammar topping;
  public <topping> = ( pepperoni | sausage | mushrooms );
  ]]>
  </grammar>
  These limitations will be removed in a future release. For more information see Appendix D of the VoiceXML 1.0 specification.
• We now allow the audio data recorded by the <record> tag to be streamed out to a remote destination specfied by an HTTP url. The recorded data is streamed out as the recording is being done. The bevocal:dest and bevocal:destexpr attributes should be used to specify the HTTP url to which the recorded data should be streamed out. If there is an error streaming out data to the HTTP url specified, then an error.semantic is thrown.

• The User-Agent header in our HTTP request headers has changed to "BeVocal/2.0"
• In ABNF grammars, comments and white space are no longer allowed inside weights. For example, in
  [ /1.0/ apple | / .5/ banana ]
  the space before the ".5" is illegal. This was never allowed by the grammar working draft; our implementation allowed it by mistake.

Enhancements

VoiceXML Interpreter

BeVocal Extensions

• <bevocal:dial> now has a silent attribute. If you set silent="true", the call-progress sounds and all other sounds from the outbound call will be muted until the post-dial digits have finished playing. If silent="false", you'll get the old behavior where you can hear the call-progress sounds and a small snippet of the call after it's connected and before the post-dial digits are played. The default is false.
  
• <bevocal:dial ani="..."> now supports ANI's up to 32 digits long. You can specify the ANI either as a formatted phone number, with or without area code, e.g. (201) 555-1212, or as a string of ASCII digits, e.g. 012345647830193856.
  
• We now support a built-in stock index grammar that matches the names of the major US stock indices. The syntax is <grammar src="builtin:grammar/stockindex"/>.
  
• We now support a built-in grammar that recognizes all of the cities in the United States. The syntax is <grammar src="builtin:grammar/citystate"/>. This grammar sets four slots: city, county, state, and datacity. When used as a field grammar the field item variable will be a JavaScript object containing properties with these names. See the <grammar> documentation for details.

VoiceXML 2.0

• In XML and ABNF grammars we now support weighting of items in alternative lists. For example, in an ABNF grammar, you can have a rule like this:
  $fruit = /0.8/ apple | /1.2/ orange | /0.5/ banana;
  In XML, this looks like:
  <one-of><item weight="0.8">apple</item>
  <item weight="1.2">orange</item>
  <item weight="0.5">banana</item>
  </one-of>
  
• We have added a new DTD corresponding to the VoiceXML 2.0 Working Draft. This DTD does not contain the 1.0 features that are obsolete in VoiceXML 2.0. We have also removed BeVocal extensions which are now obsolete. You can access this DTD with:
  <!DOCTYPE vxml PUBLIC "-//BeVocal Inc//VoiceXML 2.0//EN"
  "http://cafe.bevocal.com/libraries/dtd/vxml2-0-bevocal.dtd">
  The changes in this DTD are:
  • The caching attribute on tags that fetch resources has been replaced with maxage and maxstale. See the Fetching Resources section of the BeVocal VXML Reference for details.
    
  • The obsolete BeVocal extended property expirationtime has been removed. Use maxstale instead.
    
  • The modal attribute of <subdialog> has been removed
    
  • The TTS markup tags <emp>, <div>, <pros>, and <sayas> have been replaced with <emphasis>, <sentence>, <paragraph>, <prosody>, and <say-as>.
    
  • The lang attribute of <vxml> has been replaced with the standard xml:lang attribute.
    
  • The <dtmf> tag has been replaced with <grammar mode="dtmf">.
    
  • The msecs attribute of <break> has been replaced with a new time attribute.
    
  • The BeVocal extension <debug> has been replaced with the new, standard <log> tag.
    
  • The BeVocal extension <send> has been replaced with another extension, <data>, which does everything <send> did and more.
    
  • The hotword attribute of <prompt> has been replaced with the new bargeintype attribute.
    
  • The type attribute of <value> has been removed. <say-as class="..."> instead, e.g.
    <say-as class="phone"><value expr="myphone"/></say-as>.
    
  • In this DTD we have removed the method and submit attributes of <goto>, which were a deprecated BeVocal extension.
    

• We have added a tech support request form to the Support section of the Cafe Web site.
  
• We have improved the Cafe Web site Look and Feel and Navigation for ease of use. For example, clicking on documents will not launch new windows, but rather will replace the content of the current window.
  

Bug Fixes

Interpreter

• 2955: removed the trailing ; at the end of the cookie string sent in HTTP headers
  
• 3376: When the DTD from the VoiceXML 1.0 specification was used, field-level <filled> blocks triggered an error saying that a mode attribute was not allowed, because the DTD was supplying a default value for this attribute. The interpreter no longer does this check and simply ignores the attribute for field-level <filled> tags.
  
• 3415: There was a null pointer exception if there was no data recorded while executing <record>. It does not throw the null pointer exception now.

• 3034 When using bevocal.cafe.SOStreet, when the system asked the caller for the street name, it just said: "cityName", it should ask: "Name a street in (cityName)".
  
• 3160 Vocal Scripter outputs log when there is a parse error in the application.
  
• 3506 File Manager now sorts URLs by name by default.