Prophecy 8.0 Porting Guide: VoiceXML

When switching from the Voice Center 5.5 software platform to the newer Prophecy software platform, there are a few things to bear in mind. As with any migration to a new software offering, there may well be some degree of application-scoped edits to apply on the developer side, and we will try to detail each and every possible impact to any existing code in the below porting guide.

Frequently Asked Questions

Why do I have to move to Prophecy 8 and what can I expect?
Voxeo is continuously making improvements to its platform, and the Voice Center 5.5 platform uses older technologies, and an antiquated code branch. As part of our efforts to continually enhance our IVR services, we have decided that our software platform from 5 years back was due for deprecation in favor of the Prophecy platform, which allows for even greater functionality, and much more reliability for your mission-critical applications.

When will I need to move to Prophecy 8?
The Voice Center 5.5 hosted platforms for CallXML, CCXML, and VoiceXML is slated for end-of-life as of November 1st, 2008. It is our hope that we can help all of our production customers and active developers get ready for this platform end-of-life well before this time comes.

What do I need to do in order to get ready?
For any existing applications that are currently deployed and active, you will need to secure resources for functional application testing on the new software platform. The first step to making sure that you are ready for this porting effort is to research what version of IVR software that you are currently using on Voxeo's hosted platform. You can verify this in several ways:

In the Application Manager interface of Voxeo's hosted evolution portal ("Account" --> "Application Manager"), you can view the details of each application mapping. After selecting the "Application Manager" link in question, you will see a listing of all applications in your account containing fields for "Application Name", "Application Type", and "Deployment Type".

If your application type(s) read as "VoiceCenter 5.5 - VoiceXML 2.0 - DEPRECATED", then this indicates that your application code is using the Voice Center 5.5 platform, which is slated for deprecation.

If the application type(s) list "Prophecy - VoiceXML 2.1", then this indicates that your application code is using the most recent version of the Prophecy VoiceXML platform, and that no porting will be required.

If there is any doubt at all as to what platform that you are using, we encourage you to ask us for confirmation: We are happy to provide assistance to our developers to ensure that no problems arise from the migration to the newer software platform.

The next thing thing you should do is to read through the porting guide relevant to the XML markup that you are using (CallXML, CCXML or VoiceXML), very carefully, and apply the suggested changes to a new instance of the IVR code. After auditing the XML applications and applying the suggested changes, then it is time to map the new instance of the application to the most recent version of Voxeo's Prophecy IVR software:

1) Login to your evolution account, and select "Application Manager" ("Account" --> "Application Manager")

2) Select "Add Application"

3) Choose the target platform relevant to your application porting:

Prophecy 8.0 - CallXML 3.0, Staging
Prophecy 8.0 - CCXML Voxeo, Staging
Prophecy 8.0 - CCXML W3C 1.0, Staging
Prophecy 8.0 - VoiceXML 2.1, Staging
*Prophecy 8.0 - VoiceXML 2.1, Nuance, Staging

* Note that the VoiceXML 2.1, Nuance, Staging platform is available only upon request.

4) Map your new instance of the XML code to the platform, provision a number or application token as needed

After testing, how do I migrate my production applications to the Prophecy platform?
When you are ready to switch your production applications to using the Prophecy platform, you may either create the new Production mappings yourself, and use the "move" button in the "phone numbers" or "call start tokens" tab of the application detail screen, or you may contact your Account Manager via our ticketing system, and list the application names denoted that you want switched over in order to migrate a large group of applications en masse.

How do I perform a functional test?
It is strongly advised that all call paths, event and error handlers, and text-to-speech formatting, audio files, and resource fetching is tested very thoroughly and completely to ensure that no unforseen problems occur after migrating to the new platform.

What if I have a problem during testing?
If you come across a problem in your porting efforts and testing that isn't covered in our porting guides, then the Voxeo Support team is standing by to assist 24 hours a day, 365 days a year.

Does this platform update incur any caller-facing changes to my applications?
The default TTS for deprecated Voice Center 5.5 platform is the Speechify 2.0 engine, with additional voices being from the Rhetorical 4.2 engine. The Prophecy platform uses it's own in-house TTS engine by default, with new voices from Speechify and Rhetorical being available as well (specific details are listed below).

Owners of VoiceXML applications will now have the choice between using two different ASR engines for voice recognition: Both the Nuance 8.5 ASR, and the in-house Prophecy ASR will be available to our developers. Unless specified otherwise, all Voice Center 5.5 VoiceXML applications will be switched over to Prophecy with Nuance ASR on November 1st 2008. However, some customers who have
may opt to use the Prophecy ASR to receive a lower per-minute rate for their VoiceXML call traffic. Contact your Voxeo Account Manager for specific pricing details, should you wish to take advantage of a lower recurring fee for your production applications.

Prophecy VoiceXML 2.1 features

The Prophecy VoiceXML browser supports a number of inclusions which are based on both the VoiceXML 2.1 specification, and from feedback that we have taken from our developer community over the years. The highlights of the features included on the Prophecy software platform are listed below:

While Voice Center 5.5 offered only the Nuance ASR for voice recognition, the migration to the Prophecy platform offers not only the Nuance 8.5 ASR, but also the in-house Prophecy ASR. Production customers have the option to use either on the Prophecy platform, and for those who choose to use our own Prophecy ASR, your pricing may well change to a less expensive per-minute rate: Contact your Account Manager for details.

Like the Voice Center 5.5 platform, Prophecy 8 is 100% compliant to the W3C specification, and has passed each and every test for certified platform compliance: No other voice browser in the world can make this claim.

Prophecy's VoiceXML browser also includes partial support for VXML 2.1-compliant utterance recording via application scoped shadow variables, with the only caveat being that this shadow variable is unavailable in nomatch scenarios, (but can easily be worked around by using conditional logic based on confidence scoring).

Another VoiceXML 2.1 feature available in Prophecy 8 software is the addition of the <foreach> element, which allows the developer to loop through array items without having to implement any additional client-side scripting.

Users of the Prophecy VoiceXML platform with the Nuance ASR can now access a wide array of ASR and TTS options for applications intended for the foreign market.

The Prophecy 8 VoiceXML browser now has full SRGS + SISR support for voice grammars, allowing developers to craft application content that is 100% compliant to W3C grammar specifications for longer IVR shelf-life. Note that SISR support applies both to the Prophecy ASR as well as to the Nuance 8.5 ASR.

Lastly, the Prophecy 8 VoiceXML browser includes support for the "marktime" shadow variable, which can assist in determining exactly when a caller has barged in on a given voice prompt.

VoiceXML 2.1 Application Porting

The Nuance extension grammars (referenced via the "x-nuance-" prefix) are deprecated entirely, and are unavailable for use on the Prophecy platform, including Prophecy + Nuance ASR platform.

Any user-defined Nuance Grammar Objects, or Statistical Language Model grammars will not function using the default Prophecy ASR. In the event that NGO/SLM files are required for your application, these must be used on the Prophecy - VoiceXML 2.1, Nuance platform.

In addition, NGO and SLM files will need to be recompiled due to the changes in language models used in the Prophecy - VoiceXML 2.1, Nuance platform. The language model that was used in 5.5 was "English.America.1.2.0", whereas the model that should be specified for Prophecy is "English.America.3.3.0"

When disabling all terminating characters via the "termchar" property setting, Prophecy requires that you specify a value of any alpha character as opposed to using the "@" character on Voice Center 5.5:

Voice Center 5.5: <property name="termchar" value="@"/>
Prophecy: <property name="termchar" value="abcd"/>

Outbound call transfer syntax via the <transfer>, element or programmatic outbound dialing via a request to http://api.voxeo.net should be specified using a strict e.164 format, as the Prophecy software is more strict in this regard: While Voice Center 5.5 allowed dial strings of "4072223333", or "tel:4072223333", outbound calls on Prophecy should use a syntax that specifies "tel:+[country code][number]", or "tel:+14072223333".

At present (6/08) there exists a platform bug where <prompt timeout="0s"> will not be honored; this bug is corrected internally, but not yet deployed. It is suggested that developers work around this by specifying <prompt timeout="0.1s"> when needed.

Changing TTS and ASR languages via a <property> setting is no longer required. These settings should be specified either within <vxml xml:lang="">, <prompt xml:lang="">, or <grammar xml:lang=""> as scoping constraints dictate within your application.

When using Prophecy ASR, note that any Nuance-specific <property> settings will not be available in any capacity. When using Prophecy with Nuance ASR some of these <property> settings may have been deprecated entirely, or will be mapped to W3c-compliant equivalents. If in doubt as to whether any particular Nuance-specific property is supported, please contact the support team and we would be happy to verify for you

In Prophecy 8, we no longer filter out disambiguous utterances that are received in the context of an application that employs n-best post-processing. For example, if we had the following grammar:

[(john) (john)(john) (john) (john) (john)] {<MySlot "John">}

And further assuming that our n-best length was set to 6, the ASR would still return back 6 results.

As the text-to-speech engines have changed from Voice Center 5.5 to Prophecy, there will be some changes to expect when invoking SSML code within your applications for the following elements:

* <phoneme>
* <say-as>
* <sub>
* <emphasis>
* <prosody>

As these changes are very much dependent upon the TTS engine used, the changes specific to each TTS engine are detailed in a listing that follows.

The below TTS property values previously invoked on Voice Center 5.5 are not available on the Prophecy platforms in any capacity. Application content should be edited to invoke TTS voices via the "xml:lang" attribute so as to be compliant with the W3C specification:

* <property name="nuance.core.tts.ResourceName" value="en-US.English-Female2" />
* <property name="nuance.core.tts.ResourceName" value="en-US.default" />
* <property name="nuance.core.tts.ResourceName" value="es-MX.default" />
* <property name="nuance.core.tts.ResourceName" value="fr-FR.default" />
* <property name="nuance.core.tts.ResourceName" value="en-US.English-Male2" />
* <property name="nuance.core.tts.ResourceName" value="en-US.M009" />
* <property name="nuance.core.tts.ResourceName" value="es-MX.F023" />
* <property name="nuance.core.tts.ResourceName" value="es-MX.M022" />

SSML Guide: Prophecy TTS

The following <say-as interpret-as> specifications are not available when using the default Prophecy TTS:

* <say-as interpret-as="duration">
* <say-as interpret-as="measure">
* <say-as interpret-as="name">
* <say-as interpret-as="net">
* <say-as interpret-as="address">

The "duration" and "volume", attributes of the <prosody> element are not available

The <emphasis> element is not available

SSML Guide: Rhetorical TTS engine

The Rhetorical TTS voices available for the production Prophecy ASR are listed as follows. Production customers should be very much aware of the fact that these TTS voices are not part of the standard offerings for our hosted platform, and usage TTS other than Prophecy Male or Prophecy Female does incur setup fees, and additional per-minute charges associated with them. Any questions regarding he pricing of these TTS alternatives should be directed at your dedicated Account Manager, (note that this listing is subject to change):

* English Female: <prompt xml:lang="en-us-fm3">
* English Female: <prompt xml:lang="en-us-fm19">
* English Male: <prompt xml:lang="en-us-m3">
* English Male: <prompt xml:lang="en-us-m09">
* Spanish Female: <prompt xml:lang="es-mx-fm3">
* Spanish Female: <prompt xml:lang="es-mx.fm23">
* Spanish Male: <prompt xml:lang="es-mx-m3">
* Spanish Male: <prompt xml:lang="es-mx-m22">
* UK Female: <prompt xml:lang="en-gb-fm1">

The Rhetorical TTS voices available for the production Nuance ASR are listed as follows (note that this listing is subject to change):

* English Female: <prompt xml:lang="en-us-fm3">
* English Female: <prompt xml:lang="en-us-fm19">
* English Male: <prompt xml:lang="en-us-m3">
* English Male: <prompt xml:lang="en-us-m09">
* Spanish Female: <prompt xml:lang="es-mx-fm3">
* Spanish Female: <prompt xml:lang="es-mx-fm23">
* Spanish Male: <prompt xml:lang="es-mx-m3">

SSML Guide: Speechify TTS engine (default)

The Speechify TTS voices available for the production Prophecy ASR are listed as follows (note that this listing is subject to change):

* French Female: <prompt xml:lang="fr-fr-fm2">
* Spanish Female: <prompt xml:lang="es-mx-fm1">
* Australian Female: <prompt xml:lang="en-au-fm1">

ANNOTATIONS: EXISTING POSTS

kettle

1/24/2007 2:53 AM (EST)

"Details: The <sayas phon> was equivalent to the w3c standard tag of <phoneme>. Of note is the fact that since neither Nuance or Motorola has SSML support enabled on the Voxeo network, neither incarnation of this element will have any effect on TTS output. For this reason, it is best practice not to include either of these speech markup tags within your code."

Is there an alternative then, which will allow me some fairly fine grained control over the way that my synthesized utterances are realized by the TTS engine? If so please describe it or link to it here.

Right now I am considering building my own utterances with my own Japanese TTS engine and then using the <audio> element to play them back on the fly. I'd prefer not to do this if possible.

MattHenry

1/24/2007 2:37 PM (EST)

Hi there,

I should be clear in the fact that the hosted Voxeo VXML platforms do not currently support <phoneme>. However, you may find a useful workaround by employing the <sub> element:

<prompt>
this message brough to you by the
<sub alias="World Wide Web Consortium">W3C</sub>
</prompt>

Hope this helps!

~Matthew henry

Subdialog Modifications

| TOC

| E: VoiceXML Variables