Limitations

This section lists current limitations and feature gaps to consider when using Mix for a VoiceXML application.

Escalations

IVR applications using Nuance Speech Suite with VoiceXML Connector require engine pack 2.4 (or later) to support specifying a transfer action type (such as blind transfer or route request) and destination as such, when configuring external actions nodes.
Workaround for earlier versions of VoiceXML Connector: Use Send data parameters to share this kind of information with your client application.

Channel control

IVR applications using Nuance Speech Suite with VoiceXML Connector require engine pack 2.4 (or later) to support using the predefined variable channel for channel control.

Messages/prompts

Latency messages used in a VoiceXML application cannot be dynamic—each latency message is limited to one audio file. See Play message on hold.

Global commands

• If your dialog design uses an external speech grammar for global commands, question and answer nodes that are configured to support NLU won’t recognize commands at runtime.
  Workaround: Specify a speech grammar for the entity in focus at every question and answer node that must support commands.
• The global command entity for your dialog design cannot be a rule-based entity.

External resources

Wordsets

IVR applications using Nuance Speech Suite with VoiceXML Connector require engine pack 2.4 (or later) to support external NLU and ASR precompiled resources (including compiled wordsets). Applications in self-hosted environments don’t support this feature. Projects using an earlier version of VoiceXML Connector only support inline wordsets (see Dynamic list entities).

Location of recorded audio files and external grammars

For a given language and channel combination, all required audio files, must be collocated in a single, flat folder. Likewise, all required external grammar files must be collocated in a single folder. IVR applications using Nuance Speech Suite with VoiceXML Connector require engine pack 2.4 (or later) to support base URLs specified in the Mix dashboard for external grammars.
Workaround: If your audio and grammar resources are organized in a hierarchical folder structure—for example, if you have global resources in one folder, and application-specific resources in various folders, move them into the expected folder structure, as described in Preparing audio and grammar files.

Recognition/interpretation

Regex-based entities

VoiceXML applications that use Speech Suite don’t support regex-based entities.

Freeform entities

VoiceXML applications that use Speech Suite can’t yet support collecting freeform entities.

Disambiguation

Mix doesn’t yet allow constraining recognition to a subset of intents, in open dialog.
Workaround: Use directed dialog (Nuance recognizer grammars) for disambiguation. (Note that directed dialog doesn’t support intent switching.)

Recognition parameters

Dragon voice doesn’t support hotword detection, speed vs. accuracy, and complete timeout. See Setting global properties and behaviors, for more information.

Confidence scores

Entity confidence scores from ASR/NLU models are not accurate.
Workaround: Use the corresponding intent confidence scores. Refer to Last interpretation specification for more information.

Formatted literal

VoiceXML applications that use Speech Suite for speech recognition (as opposed to ASR as a Service) require VoiceXML Connector 1.1 or later to support a formatted literal form of entities and intents from interpretation results.

DTMF-only input states

IVR applications using Nuance Speech Suite with VoiceXML Connector require engine pack 2.4 (or later) to support question and answer nodes configured to only accept DTMF input.
Workaround for earlier versions of VoiceXML Connector:

1. Create a variable with the name property_inputmodes.
2. Set its value to dtmf.
3. Pass it as a Send data parameter at the desired question and answer nodes.
   See Enabling DTMF input, and Send data to the client application.

Voices/languages

• Speech Suite 11 doesn’t support switching to another language after the first NLU/ASR recognition. Specific configurations might allow some language switching after the first recognition. Contact your Nuance representative for details.
  Workaround: In Mix.dialog, make sure the question and answer node that asks the user to choose a language is the first question and answer node in your dialog flow, and have it reference an external Nuance Recognizer grammar or a DTMF menu. Then, set the language variable to the appropriate language code. At runtime, the Dialog service uses this variable to return the appropriate recorded audio file paths, grammar file paths, and verbiage, for the specified language.
• A VoiceXML application that uses TTS to render messages doesn’t automatically switch to the appropriate TTS voice, when the user chooses to continue in another language. This is outside the scope of the Dialog service.
  Workaround: Pass the language variable as a Send data parameter, at the next question and answer nodes after the language change, and handle the logic on the client side so that VoiceXML Connector can use this to generate a vxml tag with the desired language code (assuming the voice platform supports a single voice per installed language). For example:
  <vxml version="2.1" xmlns="http://www.w3.org/2001/vxml" xml:lang="en-US">
• Mix doesn’t support using multiple TTS voices for the same language.
• Languages, TTS voices, and customizations are limited to those available in your Speech Suite installation.
• Neural voices and the Neural TTS service are only available for Nuance-hosted Mix projects.

Try mode

The Try mode supports rudimentary testing of IVR dialogs. For complete testing, you need to deploy your VoiceXML application, and test it with VoiceXML Connector.

• In Mix.nlu and Mix.dialog, the Try mode doesn’t support speech input. Interpretation results and confidence scores exposed in Try mode apply to text input, and are therefore not relevant.
• In Mix.dialog, the Try mode doesn’t support:
  • DTMF interaction
  • External grammar references
  • Audio messages (recorded or TTS)

Large applications

Using multiple projects to design a large application as separate modules (also called subdialogs) is not supported. Doing this would require workarounds at the logging level and to maintain context between the runtime applications generated from the different projects. Design your application as a single project, using generic components and intent components to achieve the desired level of modularity.