Using Mix resources in Speech Suite 11 application

This section describes key details of how to leverage built Mix speech resources with Dragon Voice engines (Krypton engine and Natural Language Engine (NLE)) in Speech Suite 11 applications, whether in a standalone installation or together with Dialog via Nuance VoiceXML Connector.

This section covers the key points at a high level. For full details, see the Speech Suite 11 documentation.

Dragon Voice artifacts

Dragon Voice makes use of various Dragon Voice artifacts as input resources for the engines along with user speech and text inputs.

The following artifacts can be included:

NLE: NLU model
Krypton: DLMs
Both: Inline wordsets

Creating and configuring Dragon Voice manifest file

You have to supply a manifest file to identify resources. The manifest names the Krypton and NLE resources and provides default configurations to be used when loading the resources.

The manifest file is a JSON file, and must be named nuance_package.json.

For more details on the structure of the manifest, see the Speech Suite 11 documentation.

For guidance on building a manifest file for your Mix-built resources, speak to your Nuance representative.

Deploying artifacts and manifest

The manifest must be deployed in a folder along with the resources named in the manifest at the same level in the folder.

If multiple manifests are used, they must each be in their own folder.

Limitations on artifacts that can be used

The following limitations are present on the resources that can be used:

NLU models: NLE uses one NLU model only
DLMs: Krypton allows up to five DLMs
Wordsets: Krypton allows up to five wordsets
Compiled wordsets: Not supported

Loading artifacts

Artifacts, as well as builtins included in the base language pack, can be loaded to support recognition and/or NLU interpretation.

Artifacts can be loaded in three different scopes:

Service: The artifact is loaded when the service starts up. The artifact is then available for all future sessions and recognition turns.
Session: A session consists of the time when a user is connected on a call.
Recognition: This is for a single turn of recognition.

A service scope can include several session scopes related to different users, and a session scope with a single user can include multiple recognition turns.

For full details of considerations for the different types of scope and on how to load artifacts, see the Speech Suite 11 documentation.

Weighting resources

There are different ways of weighting resources in Speech Suite 11.

Default weighting via manifest

Default weightings for resources can be set in the manifest file.

Note:

Note that weights cannot be set for wordsets.

Overriding default weighting for session or recognition

Default weightings can be overridden when loading resources in session or recognition scope within a VoiceXML <grammar> tag.

For more details about weighting in Speech Suite 11, see the Speech Suite 11 documentation.