Managing data packs

Dragon Voice requires one or more data packs for its Krypton recognition engine and Nuance Text Processing Engine (NTpE) components. It uses the data packs when building models for your applications

Later, when your applications invoke Dragon Voice, they specify the version of data pack used for those models. See Triggering the Dragon Voice recognizer.

Installing data packs

Install data packs on every Krypton, NLE, and NTpE host during the Speech Suite installation process or at any time in the future.

Data pack installation packages are separate from Nuance Speech Suite packages. Download the data packs you need from Nuance Network and install them using the Windows installer (.msi file) or the RPM package on Linux.

Notes:
  • Install data packs on Krypton and NTpE hosts. Do not attempt to share a central repository across multiple hosts..

  • If you install Speech Suite on a Krypton/NTpE host without a data pack, you must stop the watcher service before installing the first data pack. (Otherwise, the required NDP_HOME environment variable will not be set.) For instructions on manually starting and stopping services, see Starting and stopping the watcher service manually.

  • At runtime, Krypton and NTpE loads all installed data packs by default. To load a subset see Configuring data packs.

Installation path and the NDP_HOME environment variable

The data pack installer does not prompt for an installation path. When you install the first data pack on a host, the installer creates the NDP_HOME environment variable, and subsequent installations use the environment variable to construct the installation path.

The default data pack home is:

  • Linux: /usr/local/Nuance/Data_Packs
  • Windows: C:\Program Files\Nuance\Data Packs

The individual data packs reside in subdirectories beneath NDP_HOME. The installer creates the directory name with the specific version of the data pack. For example:

  • Linux : $NDP_HOME/ndp-eng-USA-GEN-version/
  • Windows: %NDP_HOME%\ndp-eng-USA-GEN-version\

README file

The installation includes a README text file containing relevant information about the data pack, including:

  • Name of the data pack including version number
  • Language code (lang-LOCALE), such as eng-USA (American English)
  • Topic, such as GEN (Generic)
  • Builtins associated with the data pack's base language model, which are used to improve recognition for specific common tasks such as recognition of numbers, dates, and so on
  • Supported sampling rates, such as 8000 (8 kHz) and 16000 (16 kHz)
  • Supported audio encoding, such as linear pcm 16 bit

Replacing data packs

You cannot upgrade existing data packs. If you have a new version, you install it and both versions are co-installed.

Every data pack has a unique filename and version that cannot be re-used or substituted in any way. For example, if you need to install a new data pack you cannot rename it to match an existing filename and version.

Never replace an installed data pack with a different pack of the same (identical) name. Doing this causes unpredictable conflicts between the new pack and existing data in the memory cache.

Never rename or move data packs after installation. Doing this causes the Krypton engine capabilities to disappear in the resource manager.

Configuring data packs

Dragon Voice engines register their specific capabilities with the Nuance Resource Manager. For example, they register the language, topic, and version of loaded data packs. This registration enables the Resource Manager to allocate the engine best equipped to handle a specific application request. For more information on resource allocation rules, see Resource manager features.

By default, the Dragon Voice engines load all data packs installed under Installation path and the NDP_HOME environment variable. To load a subset, configure these properties:

Tip: By default these property values are empty. When a property includes one or more entries, only the specified entries are loaded. Therefore, to exclude a data pack, you need to include the list of data packs that you do want to load.

Details for configuring the properties:

  • Specify the exact name of the data pack; for example, [ndp-eng-USA-GEN-3.7.1]. To specify multiple data packs, use a comma-separated list, as follows: [ndp-eng-USA-GEN-3.7.1,ndp-cmn-CHN-GEN-3.7.1].
  • Entries must match the corresponding existing directories under NDP_HOME. Data pack names (and directory names) that do not comply with the above naming convention are ignored.
  • You can install different versions of the same data pack, but you cannot use them in the same Krypton or NTpE instance. For example, the engines do not support starting and loading both ndp-eng-USA-eng-USA-3.1.1 and ndp-eng-USA-eng-USA-3.7.2. Therefore, to install multiple versions on a host, you must specify the explicitly version to load.

Removing data packs

To uninstall data packs that you no longer need, use the standard removal mechanism for your operating system (on Windows the Programs and Features Control Panel, and on Linux the yum erase package_name command).

When you uninstall a data pack: