Voice Extensible Markup Language (VoiceXML) 2.1

W3C Working Draft 28 July 2004

This version:

http://www.w3.org/TR/2004/WD-voicexml21-20040728/

Latest version:

http://www.w3.org/TR/voicexml21/

Previous version:

http://www.w3.org/TR/2004/WD-voicexml21-20040323/

Editors:

Matt Oshry, Tellme Networks (Editor-in-Chief)

RJ Auburn, Voxeo Corporation

Paolo Baggia, Loquendo

Michael Bodell, Tellme Networks

David Burke, Voxpilot Ltd.

Daniel C. Burnett, Nuance Communications

Emily Candell, Comverse

Jim Ferrans, Motorola

Jeff Haynie, Vocalocity

Hakan Kilic, Scansoft

Jeff Kusnitz, IBM

Scott McGlashan, Hewlett-Packard

Rob Marchand, VoiceGenie

Brad Porter, Tellme Networks

Ken Rehor, Vocalocity

Copyright © 2004 W3C^® (MIT, ERCIM, Keio), All Rights Reserved. W3C liability, trademark, document use and software licensing rules apply.

Abstract

VoiceXML 2.1 specifies a set of features commonly implemented by Voice Extensible Markup Language platforms. This specification is designed to be fully backwards-compatible with VoiceXML 2.0 [VXML2].

Status of this Document

This section describes the status of this document at the time of its publication. Other documents may supersede this document. A list of current W3C publications and the latest revision of this technical report can be found in the W3C technical reports index at http://www.w3.org/TR/.

This is a W3C Last Call Working Draft for review by W3C Members and other interested parties. Comments on this document are welcome. Please send them before 1 September 2004. A Last Call Working Draft is a special instance of a Working Draft that is considered by the Working Group to meet the requirements of its charter, and that all known issues have been resolved. A positive outcome is precondition for advancement to Candidate Recommendation.

This specification describes version 2.1 of the Voice Extensible Markup Language. Its goal is to standardize a set of features commonly implemented by Voice Extensible Markup Language platforms. This specification is designed to be fully backwards-compatible with VoiceXML 2.0 [VXML2].

This document has been produced as part of the Voice Browser Activity (activity statement), following the procedures set out for the W3C Process. The authors of this document are members of the Voice Browser Working Group (W3C Members only).

This document was produced under the 24 January 2002 CPP as amended by the W3C Patent Policy Transition Procedure. The Working Group maintains a public list of patent disclosures relevant to this document; that page also includes instructions for disclosing a patent. An individual who has actual knowledge of a patent which the individual believes contains Essential Claim(s) with respect to this specification should disclose the information in accordance with section 6 of the W3C Patent Policy.

This document is for public review, and comments and discussion are welcomed on the (archived) public mailing list <www-voice@w3.org>. Publication as a Working Draft does not imply endorsement by the W3C Membership. This is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to cite this document as other than work in progress.

In this document, the key words "must", "must not", "required", "shall", "shall not", "should", "should not", "recommended", "may", and "optional" are to be interpreted as described in [RFC2119] and indicate requirement levels for compliant VoiceXML implementations.

1. Introduction

The popularity of VoiceXML 2.0 [VXML2] spurred the development of numerous voice browser implementations early in the specification process. [VXML2] has been phenomenally successful in enabling the rapid deployment of voice applications that handle millions of phone calls every day. This success has led to the development of additional, innovative features that help developers build even more powerful voice-activated services. While it was too late to incorporate these additional features into [VXML2], the purpose of VoiceXML 2.1 is to formally specify the most common features to ensure their portability between platforms and at the same time maintain complete backwards-compatibility with [VXML2].

This document defines a set of 8 commonly implemented additional features to VoiceXML 2.0 [VXML2].

2. Referencing Grammars Dynamically

As described in section 3.1 of [VXML2], the <grammar> element allows the specification of a speech recognition or DTMF grammar. VoiceXML 2.1 extends the <grammar> element to support the following additional attribute:

Exactly one of "src", "srcexpr", or an inline grammar must be specified; otherwise, an error.badfetch event is thrown.

The following example requests a city and state, and uses the result to select a street grammar dynamically for the second field.

<?xml version="1.0" encoding="UTF-8"?>
<vxml xmlns="http://www.w3.org/2001/vxml" 
  version="2.1">

  <form id="get_address">
    <field name="citystate">
      <grammar type="application/srgs+xml" src="citystate.grxml"/>
      <prompt>Say a city and state.</prompt>
    </field>

    <field name="street">
      <grammar type="application/srgs+xml" srcexpr="citystate + '.grxml'"/>
      <prompt> What street are you looking for? </prompt>
    </field>

    <filled>
      <prompt> 
       You chose 
       <value expr="street"/> 
       in 
       <value expr="citystate"/> 
      </prompt>
      <exit/>
    </filled>
  </form>
</vxml>

3. Referencing Scripts Dynamically

As described in section 5.3.12 of [VXML2], the <script> element allows the specification of a block of client-side scripting language code, and is analogous to the [HTML4] <SCRIPT> element. VoiceXML 2.1 extends the <script> element to support the following additional attribute:

Exactly one of "src", "srcexpr", or an inline script must be specified; otherwise, an error.badfetch event is thrown.

The following example retrieves a parameterized script based on the value of the variable user_id.

<?xml version="1.0" encoding="UTF-8"?>
<vxml xmlns="http://www.w3.org/2001/vxml" 
  version="2.1">
  <form>
    <var name="user_id" expr="12345"/>
    <script srcexpr="'http://example.org/passport?id=' + user_id"/>
  </form>
</vxml>

4. Using <mark> to Detect Barge-in During Prompt Playback

As described in section 3.3.2 of [SSML], the <mark> element places a marker into the text/tag sequence. An SSML processor must either allow the VoiceXML interpreter to retrieve or must inform the interpreter when a <mark> is executed during audio output.

[SSML] defines a single attribute, name, on the <mark> element, allowing the programmer to name the mark. VoiceXML 2.1 extends the <mark> element to support the following additional attribute:

Exactly one of "name" and "nameexpr" must be specified; otherwise, an error.badfetch event is thrown.

As described in section 4.1.1 of [VXML2], the <mark> element is permitted in conforming VoiceXML documents, but [VXML2] does not specify a standard way for VoiceXML processors to access <mark> element information. Processors of conforming VoiceXML 2.1 documents must set the following two properties on the application.lastresult$ object whenever the application.lastresult$ object is assigned (e.g. a <link> is matched) and a <mark> has been executed.

When a <mark> is executed during the processing of a form item, the interpreter sets shadow variables, the names of which correspond to the properties of the application.lastresult$ object. The value of each shadow variable must be identical to the value of the corresponding application.lastresult$ property.

The following example establishes marks at the beginning and at the end of an advertisement. In the <filled>, the code checks which mark, if any, was last executed when bargein occurred. If the "ad_start" mark is executed but "ad_end" is not, the code checks that at least 5 seconds of the advertisement has been played and sets the played_ad variable appropriately.

<?xml version="1.0" encoding="UTF-8"?>
<vxml xmlns="http://www.w3.org/2001/vxml" 
  version="2.1">

  <var name="played_ad" expr="false"/>

  <form>   
    <field name="team">
      <prompt>
        <mark name="ad_start"/>
        Baseball scores brought to you by Elephant Peanuts.
        There's nothing like the taste of fresh roasted peanuts.
        Elephant Peanuts. Ask for them by name.
        <mark name="ad_end"/>
        <break time="500ms"/>
        Say the name of a team. For example, say Boston Red Sox.
      </prompt>
  
      <grammar type="application/srgs+xml" src="teams.grxml"/>
  
      <filled>
        <if cond="typeof(team$.markname) == 'string' &amp;&amp;
           (team$.markname=='ad_end' || 
           (team$.markname=='ad_start' &amp;&amp;
              team$.marktime &gt;= 5000))">
          <assign name="played_ad" expr="true"/>
        <else/>
          <assign name="played_ad" expr="false"/>
        </if>
      </filled>
  
    </field>
  </form>   

</vxml>

5. Using <data> to Fetch XML Without Requiring a Dialog Transition

The <data> element allows a VoiceXML application to fetch arbitrary XML data from a document server without transitioning to a new VoiceXML document. The XML data fetched by the <data> element is bound to ECMAScript through the named variable that exposes a read-only subset of the W3C Document Object Model (DOM).

Attributes of <data> are:

Exactly one of "src" or "srcexpr" must be specified; otherwise, an error.badfetch event is thrown. If the content cannot be retrieved, the interpreter throws an error as specified for fetch failures in Section 5.2.6 of [VXML2].

Platforms should support parsing XML data into a DOM. If an implementation does not support DOM, the name attribute must not be set, and any retrieved content must be ignored by the interpreter. If the name attribute is present, these implementations will throw error.unsupported.data.name.

If the name attribute is present, and the returned document is XML, the VoiceXML interpreter must expose the retrieved content via a read-only subset of the DOM as specified in Appendix D. An interpreter may support additional data formats by recognizing additional media types. If an interpreter receives a document in a data format that it does not understand, or the data is not well-formed as defined by the specification of that format, the interpreter throws error.badfetch. If the media type of the retrieved content is "text/xml" but the content is not well-formed XML, the interpreter throws error.badfetch.

Like the <var> element, the <data> element can occur in executable content or as a child of <form> or <vxml>. In addition, it shares the same scoping rules as the <var> element. If a <data> element has the same name as a variable already declared in the same scope, the variable is assigned a reference to the DOM exposed by the <data> element.

If use of the DOM causes a DOMException to be thrown, but the DOMException is not caught by an ECMAScript catch handler, the VoiceXML interpreter throws error.semantic.

Like the <submit> element, when an ECMAScript variable is submitted to the server its value is first converted into a string before being submitted. If the variable is an ECMAScript Object the mechanism by which it is submitted is not currently defined. If a <data> element's namelist contains a variable which references recorded audio but does not contain an enctype of multipart/form-data, the behavior is not specified. It is probably inappropriate to attempt to URL-encode large quantities of data.

In the examples that follow, the XML document fetched by the <data> element is in the following format:

<?xml version="1.0" encoding="UTF-8"?>
<?access-control allow="*.roadrunner.edu *.acme.edu"?>
<quote>
  <ticker>F</ticker>
  <name>Ford Motor Company</name>
  <change>1.00</change>
  <last>30.00</last>
</quote>

The following example assigns the value of the "last" element to the ECMAScript variable "price":

<data name="quote" src="quote.xml"/> 
<script><![CDATA[
  var price = quote.documentElement.getElementsByTagName("last").item(0).firstChild.data;
]]></script>

The data is fetched when the <data> element is executed according to the caching rules established in Section 6.1 of [VXML2].

Before exposing the data in an XML document referenced by the <data> element via the DOM, the interpreter must check the <?access-control?> processing instruction included in the XML document indicating the domains allowed to access the data. If access is denied the interpreter must throw error.noauthorization. The "access-control" processing instruction algorithm is formally described in Appendix E.

The following example retrieves a stock quote in one dialog, caches the DOM in a variable at document scope, and uses the DOM to playback the quote in another dialog.

<?xml version="1.0" encoding="UTF-8"?>
<vxml xmlns="http://www.w3.org/2001/vxml" 
  version="2.1">
  <var name="quote"/>
  <var name="ticker" expr="'f'"/>

  <form id="get_quote">
     <block>
        <data name="quote" srcexpr="'http://example.org/getquote?ticker=' + ticker"/>
        <assign name="document.quote" expr="quote.documentElement"/>
        <goto next="#play_quote"/>         
     </block>
  </form>

  <form id="play_quote">

     <script><![CDATA[
     // retrieve the value contained in the node t from the DOM exposed by d
     function GetData(d, t, nodata)
     {
        try {
           return d.getElementsByTagName(t).item(0).firstChild.data;
        }
        catch(e)
        {
           // the value could not be retrieved, so return this instead
           return nodata;
        }
     }
     ]]></script>

     <block>
        <!-- retrieve the change in the stock's value -->
        <var name="change" expr="GetData(quote, 'change', 0)"/>

        <!--play the company name -->
        <audio expr="ticker + '.wav'"><value expr="GetData(quote, 'name', 'unknown')"/></audio>
        <!-- play 'unchanged, 'up', or 'down' based on zero, positive, or negative change -->
        <if cond="change == 0">
           <audio src="unchanged_at.wav"/>
        <else/>
           <if cond="change &gt; 0">
              <audio src="up.wav"/>
           <else/> <!-- negative -->
              <audio src="down.wav"/>
           </if>
           <audio src="by.wav"/>
           <!-- play change in value as positive number -->
           <audio expr="Math.abs(change) + '.wav'"><value expr="Math.abs(change)"/></audio>
           <audio src="to.wav"/>
        </if>
        <!-- play the current price per share -->
        <audio><value expr="GetData(quote, 'last', 0)"/></audio>
     </block>
  </form>
</vxml>

6. Concatenating Prompts Dynamically Using <foreach>

The <foreach> element allows a VoiceXML application to iterate through an ECMAScript array and to execute the content contained within the <foreach> element for each item in the array.

Attributes of <foreach> are:

Both "array" and "item" must be specified; otherwise, an error.badfetch event is thrown.

The <foreach> element can occur in executable content and as a child of <prompt>.

The following example calls a user-defined function GetMovieList that returns an ECMAScript array. The array is assigned to the variable named 'prompts'. Upon entering the <field>, if a noinput or a nomatch event occurs, the VoiceXML interpreter reprompts the user by executing the second <prompt>. The second <prompt> executes the <foreach> element by iterating through the ECMAScript array 'prompts' and assigning each array element to the variable 'thePrompt'. Upon each iteration of the <foreach>, the interpreter executes the contained <audio> and <break> elements.

<?xml version="1.0" encoding="UTF-8"?>
<vxml xmlns="http://www.w3.org/2001/vxml"
  version="2.1">

  <script src="movies.js"/>

  <form id="pick_movie">
  
    <!-- 
    GetMovieList returns an array of objects
      with properties audio and tts.
      The size of the array is undetermined until runtime.
    -->  
    <var name="prompts" expr="GetMovieList()"/>

    <field name="movie">
      <grammar type="application/srgs+xml" src="movie_names.grxml"/>
  
      <prompt>Say the name of the movie you want.</prompt>
  
      <prompt count="2">
        <audio src="prelist.wav">When you hear the name of the movie you want, just say it.</audio>
        <foreach item="thePrompt" array="prompts">
          <audio expr="thePrompt.audio"><value expr="thePrompt.tts"/></audio>
          <break time="300ms"/>
        </foreach>
      </prompt>
  
      <noinput>
        I'm sorry. I didn't hear you.
        <reprompt/>
      </noinput>
  
      <nomatch>
        I'm sorry. I didn't get that.
        <reprompt/>
      </nomatch>
  
    </field>
  </form>
</vxml>

The following is a contrived implementation of the user-defined GetMovieList function:

function GetMovieList()
{
  var movies = new Array(3);
  movies[0] = new Object();
  movies[0].audio = "godfather.wav"; movies[0].tts = "the godfather";
  movies[1] = new Object();
  movies[1].audio = "high_fidelity.wav"; movies[1].tts = "high fidelity";
  movies[2] = new Object();
  movies[2].audio = "raiders.wav"; movies[2].tts = "raiders of the lost ark";
  
  return movies;
}

When the interpreter queues the second <prompt>, it expands the <foreach> element in the previous example to the following:

<audio src="godfather.wav">the godfather</audio>
<break time="300ms"/>
<audio src="high_fidelity.wav">high fidelity</audio>
<break time="300ms"/>
<audio src="raiders.wav">raiders of the lost ark</audio>
<break time="300ms"/>

The following example combines the use of the <mark> and <foreach> elements to more precisely identify which item in a list of movies the user has selected. During each iteration of the <foreach>, the interpreter stores the current item in the array to the variable movie. Upon execution of the <mark> element, the name of the <mark> is set to the current value of the variable movie_idx which is incremented during each iteration of the loop using the <assign> element.

In the <filled>, if the form item variable mov is set to the value "more", indicating the user's desire to hear more detail about the movie name that was played when the user barged in, the code retrieves the index of the desired movie in the movies array from mov$.markname. If barge-in occurred within twenty milliseconds of the mark's execution, the code retrieves the index of the preceding movie under the assumption that the user was slow to react.

<?xml version="1.0" encoding="UTF-8"?>
<vxml xmlns="http://www.w3.org/2001/vxml"
  version="2.1">

  <script src="movies.js"/>

  <form id="list_movies">
    <!-- fetch the current list of movies, and transform them into an array -->
    <data name="domMovies" src="getmovies.cgi"/>
    <var name="movies" expr="GetMovieList(domMovies)"/>
    <var name="movie_idx" expr="0"/>
    <var name="movie_id"/>
  
    <field name="mov">
      <prompt>
        Say the name of the movie.
      </prompt>
  
      <prompt count="2">
        Here's the list of movies. 
        To hear more about a movie, say 'tell me more'.
        <break time="500ms"/>
        <foreach item="movie" array="movies">
          <mark nameexpr="movie_idx"/>
          <audio expr="movie.audio"><value expr="movie.tts"/></audio>
          <break time="500ms"/>
          <assign name="movie_idx" expr="movie_idx+1"/>
        </foreach>
      </prompt>
  
      <grammar type="application/srgs+xml" src="more.grxml"/>
      <grammar type="application/srgs+xml" src="movies.grxml"/>
  
      <catch event="nomatch">
        Sorry. I didn't get that.
        <reprompt/>
      </catch>
  
      <catch event="noinput">
        <reprompt/>
      </catch>
  
      <filled>
        <if cond="'more' == mov">
          <!-- user wants more detail -->
          <!-- assume no mark was executed -->
          <assign name="movie_idx" expr="0"/>
          <if cond="mov$.markname != undefined">
            <!-- alas, a mark was executed, so adjust movie_idx -->
            <if cond="mov$.marktime &lt;= 20">
              <!-- returns the id of the previous movie (or the first) -->
              <assign name="movie_idx" 
                expr="(mov$.markname &lt;= 1 ? 0 : mov$.markname-1)"/>
            <else/>
              <assign name="movie_idx" expr="mov$.markname"/>
            </if>
          </if>          
          <assign name="movie_id" expr="movies[movie_idx].id"/>
        <else/>
          <!-- user said a specific movie -->
          <assign name="movie_id" expr="mov"/>
        </if>
      </filled>    
    </field>

    <!-- 
     Given a movie id, fetch detail, and play it back.
     detail could be in a single wav file or in multiple files.
    -->  
    <block name="play_detail">
      <!-- GetMovieDetail returns an array of objects with properties audio and tts. -->
      <var name="details" expr="GetMovieDetail(domMovies, movie_id)"/>
      <foreach item="chunk" array="details">
        <audio expr="chunk.audio"><value expr="chunk.tts"/></audio>
      </foreach>
    </block>
  
  </form>
</vxml>

The following is a contrived XML document returned by getmovies.cgi. The user-defined GetMovieList and GetMovieDetail functions below are implemented to manipulate the DOM representation of its structure.

<?xml version="1.0" encoding="UTF-8"?>
<movies>
  <movie id="m0010">
    <title>
      <tts>the godfather</tts>
      <wav>godfather.wav</wav>
    </title>
    <details>
    <detail>
      <wav>directors/directed_by.wav</wav>
      <tts>directed by</tts>
    </detail>
    <detail>
      <wav>directors/coppola.wav</wav>
      <tts>francis ford coppola</tts>
    </detail>
    <detail>
      <wav>ratings/rated.wav</wav>
      <tts>rated</tts>
    </detail>
    <detail>
      <wav>ratings/r.wav</wav>
      <tts>r</tts></detail>
    <detail>
      <wav>pauses/300.wav</wav>
      <tts/>
    </detail>
    <detail>
      <wav>synopsis/m0010.wav</wav>
      <tts>a coming of age story about a nice italian boy</tts>
    </detail>
    </details>
  </movie>
  
  <movie id="m0052">
    <title>
      <tts>high fidelity</tts>
      <wav>high_fidelity.wav</wav>
    </title>
    <details>
      <!-- details omitted for the sake of brevity -->
    </details>
  </movie>
  
  <movie id="m0027">
   <title>
     <tts>raiders of the lost ark</tts>
     <wav>raiders.wav</wav>
   </title>
    <details>
      <!-- details omitted for the sake of brevity -->
    </details>
  </movie>
</movies>

The following is an implementation of the user-defined GetMovieList function:

// return an array of user-defined movie objects retrieved from the DOM
function GetMovieList(domMovies) {
  var movies = new Array();
  try {
    var nodeRoot = domMovies.documentElement;
    for (var i = 0; i < nodeRoot.childNodes.length; i++) {
      var nodeChild = nodeRoot.childNodes.item(i);
      if ("movie" != nodeChild.nodeName) {
        continue;
      }
      else {
        var objMovie = new Object(); 
        objMovie.id = nodeChild.getAttribute("id");
        var nodeTitle = GetTitle(nodeChild);
        objMovie.audio = GetWav(nodeTitle);
        objMovie.tts = GetTTS(nodeTitle);
        movies.push(objMovie);
      }
    }
  }
  catch(e) {
    // unable to build movie list
    var objError = new Object();
    objError.audio = "nomovies.wav";
    objError.tts = "sorry. no movies are available."
    details.push(objDetail);
  }
  return movies;
}

function GetTitle(nodeMovie) {
  return GetChild(nodeMovie, "title");
}

function GetTTS(node) {
  return GetInnerText(GetChild(node, "tts"));
}

function GetWav(node) {
  return GetInnerText(GetChild(node, "wav"));
}

// perform a shallow traversal of the node referenced by parent 
// looking for the named node
function GetChild(parent, name)
{
  var target = null;
  for (var i = 0; i < parent.childNodes.length; i++) {
    var child = parent.childNodes.item(i);
    if (child.nodeName == name) {
      target = child;
      break;
    }
  }
  return target;
}

// Given a node, dig the text nodes out and return them as a string
function GetInnerText(node) {
  var s = "";
  if (null == node) {
    Log("GetInnerText found null node");
    return s;
  }

  for (var i = 0; i < node.childNodes.length; i++) {
    var child = node.childNodes.item(i);
    if (Node.TEXT_NODE == child.nodeType || Node.CDATA_SECTION_NODE == child.nodeType) {
      s += child.data;
    }
    else if (Node.ELEMENT_NODE == child.nodeType) {
      s += GetInnerText(child);
    }
  }
  return s;
}

The following is an implementation of the user-defined GetMovieDetail function:

// get the details about the movie specified by id
function GetMovieDetail(domMovies, id) {
  var details = new Array();
  try {
    var nodeMovie = domMovies.getElementById(id);
    var nodeTitle = GetTitle(nodeMovie);
    var objTitle = new Object();
    objTitle.audio = GetWav(nodeTitle);
    objTitle.tts = GetTTS(nodeTitle);
    details.push(objTitle);
    var nodeDetails = GetChild(nodeMovie, "details");
    for (var i = 0; i < nodeDetails.childNodes.length; i++) {
      var nodeChild = nodeDetails.childNodes.item(i);
      if ("detail" == nodeChild.nodeName) {
        var objDetail = new Object();
        objDetail.audio = GetWav(nodeChild);
        objDetail.tts = GetTTS(nodeChild);
        details.push(objDetail);
      }
    }
  }
  catch(e) {
    // couldn't get movie details
    var objError = new Object();
    objError.audio = "nomovie.wav";
    objError.tts = "sorry. details for that movie are unavailable."
    details.push(objDetail);
  }

  return details;
}

When the interpreter queues the second <prompt> in the field named "mov", it expands the <foreach> element to the following:

<mark name="0"/>
<audio src="godfather.wav">the godfather</audio>
<break time="300ms"/>
<mark name="1"/>
<audio src="high_fidelity.wav">high fidelity</audio>
<break time="300ms"/>
<mark name="2"/>
<audio src="raiders.wav">raiders of the lost ark</audio>
<break time="300ms"/>

If the user chooses "The Godfather", when the interpreter executes the "play_detail" <block>, it expands the <foreach> element to the following:

<audio src="directors/directed_by.wav">directed by</audio>
<audio src="directors/coppola.wav">francis ford coppola</audio>
<audio src="ratings/rated.wav">rated</audio>
<audio src="ratings/r.wav">r</audio>
<audio src="pauses/300.wav"/>
<audio src="synopsis/m0010.wav">a coming of age story about a nice italian boy</audio>

7. Recording User Utterances While Attempting Recognition

Several elements defined in [VXML2] can instruct the interpreter to accept user input during execution. These elements include <field>, <initial>, <link>, <menu>, <record>, and <transfer>. VoiceXML 2.1 extends these elements to allow the interpreter to conditionally enable recording while simultaneously gathering input from the user.

To enable recording during recognition, set the value of the recordutterance property to true. If the recordutterance property is set to true in the current scope, the following three shadow variables are set on the appropriate form item variable:

The interpreter sets the corresponding properties on the application.lastresult$ object, and the value of each property must be identical to the value of the corresponding shadow variable. In the case of <link> and <menu>, the interpreter only sets the application.lastresult$ properties.

Support for this feature is optional on <record>, and <transfer>. Platforms that support it set the aforementioned shadow variables on the associated form item variable and the corresponding properties on the application.lastresult$ object when the recordutterance property is set to true in an encompassing scope.

Like recordings created using the <record> tag, utterance recordings can be submitted to a document server via HTTP POST using the namelist attribute of the <submit> and <subdialog> elements. The enctype attribute must be set to "multipart/form-data", and the method attribute must be set to "post". To provide flexibility in the naming of the variable that is submitted to the document server, the interpreter must allow the utterance recording to be assigned to and posted via any valid ECMAScript variable.

In the following example, the dialog requests a city and state from the user. On the third recognition failure, the recording of the user's utterance is submitted to a document server.

<?xml version="1.0" encoding="UTF-8"?>
<vxml xmlns="http://www.w3.org/2001/vxml" 
  version="2.1">
  <form>
    <property name="recordutterance" value="true"/>

    <field name="city_state">
     <prompt>
      Say a city and state.
     </prompt>

     <grammar type="application/srgs+xml" src="citystate.grxml"/>

     <nomatch>
      I'm sorry. I didn't get that.
      <reprompt/>
     </nomatch>

     <nomatch count="3">
        <var name="the_recording" expr="lastresult$.recording"/>
        <submit method="post" 
          enctype="multipart/form-data" 
          next="upload.cgi"
          namelist="the_recording"/>
     </nomatch>
    </field>  
  </form>
</vxml>

7.1 Specifying the Media Format of Utterance Recordings

To specify the media format of the resulting recording, set the recordutterancetype property. Platforms must support the audio file formats specified in Appendix E of [VXML2]. Other formats may also be supported. The recordutterancetype property defaults to a platform-specific format which should be one of the required formats. If an unsupported media format is encountered during recognition, the platform throws an error.unsupported.format event which specifies the unsupported media format in its message variable. Note that the recordutterancetype property does not affect the <record> element.

8. Adding namelist to <disconnect>

As described in section 5.3.11 of [VXML2], the <disconnect> element causes the interpreter context to disconnect from the user. VoiceXML 2.1 extends the <disconnect> element to support the following attribute:

The <disconnect> namelist and the <exit> namelist are processed independently. If the interpreter executes both a <disconnect> namelist and an <exit> namelist, the interpreter context receives both sets of variables.

9. Adding type to <transfer>

As described in section 2.3.7 of [VXML2], the <transfer> element directs the interpreter to connect the caller to another entity. VoiceXML 2.1 extends the <transfer> element to support the following additional attribute:

Exactly one of "bridge" or "type" may be specified; otherwise an error.badfetch event is thrown.

As specified in 2.3.7 of [VXML2], the <transfer> element is optional, though platforms should support it. Platforms that support <transfer> may support any combination of bridge, blind, or consultation transfer types.

If the value of the type attribute is set to "bridge", the interpreter's behavior must be identical to its behavior when the value of the bridge attribute is set to "true". If the value of the type attribute is set to "blind", the interpreter's behavior must be identical to its behavior when the bridge attribute is set to "false". The behavior of the bridge attribute is fully specified in section 2.3.7 of [VXML2]. If the type attribute is specified and the bridge attribute is absent, the value of the type attribute takes precedence over the default value of the bridge attribute.

The bridge attribute is maintained for backwards compatiblity with [VXML2]. Since all of the functionality of the bridge attribute has been incorporated into the type attribute, developers are encouraged to use the type attribute on platforms that implement it.

The connecttimeout attribute of <transfer> applies if the type attribute is set to "bridge" or "consultation".

The maxtime attribute of <transfer> applies if the type attribute is set to "bridge".

9.1 Consultation Transfer

The consultation transfer is similar to a blind transfer except that the outcome of the transfer call setup is known and the caller is not dropped as a result of an unsuccessful transfer attempt. When performing a consultation transfer, the platform monitors the progress of the transfer until the connection is established between caller and callee. If the connection cannot be established (e.g. no answer, line busy, etc.), the session remains active and returns control to the application. As in the case of a blind transfer, if the connection is established, the interpreter disconnects from the session, connection.disconnect.transfer is thrown, and document interpretation continues normally. Any connection between the caller and the callee remains in place regardless of document execution. For additional information on call transfers with consultation-like functionality see [ATT_50075], [MCI_ECR], and [ETSI300_369].

Figure 1: Audio Connections during a consultation transfer: <transfer type="consultation"/>

The possible outcomes for a consultation transfer before the connection to the callee is established are:

The possible outcomes for a consultation transfer once the connection to the callee is established are:

9.2 Consultation Transfer Errors and Events

One of the following events may be thrown during a consultation transfer:

If a consultation transfer could not be made, one of the following errors will be thrown:

9.3 Example of a Consultation Transfer

The following example attempts to perform a consultation transfer of the caller to a another party. Prompts may be included before or within the <transfer> element. This may be used to inform the caller of what is happening, with a notice such as "Please wait while we transfer your call." The <prompt> within the <block>, and the <prompt> within <transfer> are queued and played before actually performing the transfer. After the prompt queue is flushed, the outgoing call is initiated. The "transferaudio" attribute specifies an audio file to be played to the caller in place of audio from the far-end until the far-end answers. If the audio source is longer than the connect time, the audio will stop playing immediately upon far-end answer.

Figure 2: Sequence and timing during an example of a consultation transfer

<?xml version="1.0" encoding="UTF-8"?>
<vxml version="2.1" 
  xmlns="http://www.w3.org/2001/vxml" 
<catch event="connection.disconnect.transfer">
   <!-- far-end answered -->
   <log> Connection with the callee established: transfer executed.</log>
</catch>

<form id="consultation_xfer">
   <block>
     <!-- queued and played before starting the transfer -->
     <prompt>
        Calling Riley.
     </prompt>
   </block>
   <!-- Play music while attempting to connect to far-end -->
   <!-- Wait up to 60 seconds for the far end to answer  -->
   <transfer name="mycall" dest="tel:+1-555-123-4567"
      transferaudio="music.wav" connecttimeout="60s" type="consultation">
     <!-- queued and played before starting the transfer -->
     <prompt>
        Please wait...
     </prompt>
     <filled>
        <if cond="mycall == 'busy'">
           <prompt>
             Riley's line is busy. Please call again later.
           </prompt>
         <elseif cond="mycall == 'noanswer'"/>
           <prompt>
             Riley can't answer the phone now. Please call
             again later.
           </prompt>
         </if>
      </filled>
   </transfer>
   <!-- submit call statistics to server -->
   <block>
      <submit namelist="mycall" next="/cgi-bin/report"/>
   </block>
</form>
</vxml>

Appendix A: VoiceXML Document Type Definition

The VoiceXML DTD is located at http://www.w3.org/TR/voicexml21/vxml.dtd.

Due to DTD limitations, the VoiceXML DTD does not correctly express that the <metadata> element can contain elements from other XML namespaces.

Appendx B: VoiceXML Schema

This section is Normative.

The VoiceXML schema is located at http://www.w3.org/TR/voicexml21/vxml.xsd.

The VoiceXML schema depends upon other schema defined in the VoiceXML namespace:

• vxml-datatypes.xsd: definition of datatypes used in VoiceXML schema
• vxml-attribs.xsd: definition of attributes and attribute groups used in VoiceXML schema
• vxml-grammar-restriction.xsd: this schema references the no-namespace schema of the Speech Recognition Grammar Specification 1.0 [SRGS] and restricts some of its definitions for embedding in the VoiceXML namespace.
• vxml-grammar-extension.xsd: this schema references vxml-grammar-restriction.xsd and extends some of its definitions for VoiceXML.
• vxml-synthesis-restriction.xsd: this schema references the no-namespace schema of the Speech Synthesis Markup Language 1.0 [SSML] and extends as well as restricts some of its definitions for embedding in the VoiceXML namespace.
• vxml-synthesis-extension.xsd: this schema references vxml-synthesis-restriction.xsd and extends some of its definitions for VoiceXML.

The complete set of Speech Interface Framework schema required for VoiceXML 2.1 is available here.

Appendix C: Conformance

This section is normative.

Appendix C.1: Conforming VoiceXML Document

A conforming VoiceXML document is a well-formed [XML] document that requires only the facilities described as mandatory in this specification and in [VXML2]. Such a document must meet all of the following criteria:

1. The document must conform to the constraints expressed in the VoiceXML Schema
2. The root element of the document must be <vxml>.
3. The <vxml> element must include a "version" attribute with the value "2.1".
4. The <vxml> element must designate the VoiceXML namespace using the "xmlns" attribute [XMLNAMES]. The namespace for VoiceXML is defined to be http://www.w3.org/2001/vxml.

5. It is recommended that the <vxml> element also include "xmlns:xsi" and "xsi:schemaLocation" attributes to indicate the location of the schema for the VoiceXML namespace. If the "xsi:schemaLocation" attribute is present, it must include a reference to the VoiceXML Schema:

   xsi:schemaLocation="http://www.w3.org/2001/vxml 
     http://www.w3.org/TR/voicexml21/vxml.xsd"
   

6. There may be a DOCTYPE declaration in the document prior to the root element. If present, the public identifier included in the DOCTYPE declaration must reference the VoiceXML DTD using its Formal Public Identifier.

   <!DOCTYPE vxml 
     PUBLIC "-//W3C//DTD VOICEXML 2.1//EN" 
     "http://www.w3.org/TR/voicexml21/vxml.dtd">
   

   The system identifier may be modified appropriately. The DTD subset must not be used to override any parameter entities in the DTD.

Here is an example of a Conforming VoiceXML document:

<?xml version="1.0" encoding="UTF-8"?>
<vxml version="2.1" xmlns="http://www.w3.org/2001/vxml"
  xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" 
  xsi:schemaLocation="http://www.w3.org/2001/vxml 
  http://www.w3.org/TR/voicexml21/vxml.xsd">

  <form>
    <block>hello</block>
  </form>

</vxml>

In this example, the recommended "xmlns:xsi" and "xsi:schemaLocation" attributes are included as is an XML declaration. An XML declaration like the one above is not required in all XML documents. VoiceXML document authors are strongly encouraged to use XML declarations in all their documents. Such a declaration is required when the character encoding of the document is other than the default UTF-8 or UTF-16 and no encoding was determined by a higher-level protocol.

The VoiceXML language or these conformance criteria provide no designated size limits on any aspect of VoiceXML documents. There are no maximum values on the number of elements, the amount of character data, or the number of characters in attribute values.

Appendix C.2: Using VoiceXML with other namespaces

The VoiceXML namespace may be used with other XML namespaces as per [XMLNAMES], although such documents are not strictly conforming VoiceXML documents as defined above. Future work by W3C will address ways to specify conformance for documents involving multiple namespaces.

Appendix C.3: Conforming VoiceXML Processors

A VoiceXML processor is a user agent that can parse and process Conforming VoiceXML documents.

In a Conforming VoiceXML Processor, the XML parser must be able to parse and process all well-formed XML constructs defined within [XML] and [XMLNAMES]. It is not required that a Conforming VoiceXML processor use a validating parser.

A Conforming VoiceXML Processor must be a Conforming Speech Synthesis Markup Language Processor [SSML] and a Conforming XML Grammar Processor [SRGS] except for differences described in this document. If a syntax error is detected processing a grammar document, then an "error.badfetch" event must be thrown.

A Conforming VoiceXML Processor must support the syntax and semantics of all VoiceXML elements as described in this document and in [VXML2]. Consequently, a Conforming VoiceXML Processor must not throw an 'error.unsupported.<element>' for any VoiceXML element which must be supported when processing a Conforming VoiceXML Document.

When a Conforming VoiceXML Processor encounters a Conforming VoiceXML Document with non-VoiceXML elements or attributes which are proprietary, defined only in earlier versions of VoiceXML, or defined in a non-VoiceXML namespace, and which cannot be processed, then it must throw an "error.badfetch" event.

When a Conforming VoiceXML Processor encounters a document with a root element designating a namespace other than VoiceXML, its behavior is undefined.

There is, however, no conformance requirement with respect to performance characteristics of the VoiceXML Processor.

While the features of VoiceXML 2.1 are orthogonal additions to [VXML2], a VoiceXML application should not mix documents of both types. The handling of a single application that mixes VoiceXML 2.0 and VoiceXML 2.1 functionality is platform-specific. Reasonable behavior in this case ranges the gamut from successful execution to the throwing of an error.

Interpreters that support both VoiceXML 2.0 and VoiceXML 2.1 must support the ability to transition from an application of one version to an application of another version.

Appendix D: ECMAScript Language Binding for DOM

This appendix contains the ECMAScript binding for the subset of Level 2 of the Document Object Model exposed by the <data> element.

Appendix E: Securing access to <data>

Before exposing an XML document referenced by the <data> element via the DOM to a voice application, the interpreter must validate that the host requesting the document is allowed to access the data. This validation is performed by comparing the hostname and IP Address of the document server from which the document containing the <data> element was fetched to the list of hostnames, hostname suffixes, and IP addresses listed in the <?access control?> processing instruction included in the XML document referenced by the <data> element. When comparing hostnames, the interpreter must perform a case insensitive match as specified in [RFC2616]. If the XML document specifies an <?access-control?> processing instruction, access to the data is allowed based on the following algorithm:

1. If the IP address of the requested document matches a value in the deny attribute, access is refused, and the search algorithm is stopped.
2. If the IP address of the requested document matches a value in the allow attribute, access is allowed, and the search algorithm is stopped.
3. If the fully qualified domain name of the requested document exactly matches a value in the deny attribute, access is refused, and the search algorithm is stopped.
4. If the fully qualified domain name of the requested document exactly matches a value in the allow attribute, access is allowed, and the search algorithm is stopped.
5. The interpreter then searches for the best match using wildcards on the domain name. Best match is defined as the closest match using the wildcards (e.g. "bert.evil.example.com" matches "*.evil.example.com" more closely than "*.example.com").
6. If a best match occurs in the deny attribute, access is refused and the search algorithm is stopped.
7. If a best match occurs in the allow attribute, access is allowed and the search algorithm is stopped.
8. If there is no match on any of the <?access-control?> processing instructions then access is refused and the search algorithm is stopped.

If the XML document does not contain an <?access-control?> processing instruction, access to the XML content is dependent on the platform execution environment. If access is denied the interpreter throws error.noauthorization.

If the VoiceXML interpreter encounters multiple <?access-control?> processing instructions in the retrieved XML document, it combines them in document order.

The following grammar describes the syntax for the <?access control?> processing instruction used by the <data> element. The grammar is specified using Extended Backus-Naur Form (EBNF) notation. For more information on this syntax, see section 6, Notation, in [XML]. For definitions of the HostName and IPv4address, and IPv6address productions, see [RFC2732].

Access Control Processing Instruction

In the following example, the hosts named "voice.roadrunner.edu" and "voice.acme.edu" are allowed access to the data. A data request from a VoiceXML document located on all other hosts (e.g. "voice.coyote.net") will fail.

<?access-control allow="voice.roadrunner.edu voice.acme.edu"?>

Numerous hosts within a domain may require data access, and listing them all is impractical. For this reason, the VoiceXML interpreter supports wildcard matching through the use of an asterisk ('*') at the beginning of a domain name. In the following example, all hosts within the "roadrunner.edu" and "acme.net" domains are allowed access to the data in which the processing instruction is contained:

<?access-control allow="*.roadrunner.edu *.acme.edu"?>

To allow any host in any domain to access the data, set the value of allow to a single asterisk ('*') as shown in the following example:

<?access-control allow="*"?>

To allow any host in the "example.com" domain to access the data with the exception of hosts from within the "visitors.example.com" domain, set the value of allow to "*.example.com" and the value of deny to "*.visitors.example.com" as shown in the following example:

<?access-control allow="*.example.com" deny="*.visitors.example.com"?>

Appendix F: References

Appendix F.1: Normative References

DOM2

DOM Level 2, ed. Arnaud Le Hors et al. W3C Recommendation, November 2000. See http://www.w3.org/TR/DOM-Level-2-Core.

HTML4

HTML 4.01, ed. Dave Raggett et al. W3C Recommendation, December 1999. See http://www.w3.org/TR/1999/REC-html401-19991224/.

SRGS

Speech Recognition Grammar Specification Version 1.0, ed. Andrew Hunt and Scott McGlashan W3C Recommendation, March 2004. See http://www.w3.org/TR/2004/REC-speech-grammar-20040316/.

SSML

Speech Synthesis Markup Language, ed. Daniel C. Burnett et al. W3C Proposed Recommendation 15 July 2004. See http://www.w3.org/TR/2004/PR-speech-synthesis-20040715/.

RFC2119

Key words for use in RFCs to Indicate Requirement Levels, ed. S. Bradner, March 1997. See http://www.ietf.org/rfc/rfc2119.txt.

RFC2616

Hypertext Transfer Protocol -- HTTP/1.1, ed. R. Fielding et al. IETF RFC 2616, June 1999. See http://www.ietf.org/rfc/rfc2616.txt.

RFC2732

IPv6 Literal Addresses in URL's, ed. R. Hinden et al. IETF RFC 2732, December 1999. See http://www.ietf.org/rfc/rfc2732.txt.

VXML2

VoiceXML 2.0, ed. Scott McGlashan et al. W3C Recommendation, March 2004. See http://www.w3.org/TR/2004/REC-voicexml20-20040316/.

XML

Extensible Markup Language (XML) 1.0, ed. Tim Bray et al. W3C Recommendation, October 2000. See http://www.w3.org/TR/2004/REC-xml-20040204/.

XMLNAMES

Namespaces in XML, ed. Tim Bray et al. W3C Recommendation, January 1999. See http://www.w3.org/TR/1999/REC-xml-names-19990114/.

Appendix F.2: Other References

ATT_50075

AT&T Toll Free Transfer Connect (TR 50075), September 1999. See http://www.att.com/cpetesting/pdf/tr50075.pdf.

MCI_ECR

MCI Enhanced Call Routing, See http://global.mci.com/wholesale/us/voice/enhancedcallrouting/.

ETSI300_369

ETS 300 369-1: Integrated Services Digital Network (ISDN);Explicit Call Transfer (ECT) supplementary service;Digital Subscriber Signalling System No. one (DSS1) protocol;Part 1: Protocol specification, May 1995. See http://www.etsi.org/.

Appendix F.3: Acknowledgements

VoiceXML 2.1 was written by the participants in the W3C Voice Browser Working Group. The following have significantly contributed to writing this specification:

• Matt Oshry, Tellme Networks
• RJ Auburn, Voxeo Corporation
• Paolo Baggia, Loquendo
• Michael Bodell, Tellme Networks
• David Burke, Voxpilot Ltd.
• Daniel C. Burnett, Nuance Communications
• Emily Candell, Comverse
• Ken Davies, HeyAnita
• Jim Ferrans, Motorola
• Jeff Haynie, Vocalocity
• Matt Henry, Voxeo Corporation
• Hakan Kilic, Scansoft
• Jeff Kusnitz, IBM
• Scott McGlashan, Hewlett-Packard
• Brien Muschett, IBM
• Rob Marchand, VoiceGenie
• Brad Porter, Tellme Networks
• Ken Rehor, Vocalocity
• Laura Ricotti, Loquendo
• Davide Tosello, Loquendo