Configuring applications to use VoCon 4.2

Last reviewed: 12/15/2011

HOW Article ID: H121103

The information in this article applies to:

  • GrammarKit 4
  • LexiconKit 4
  • ProfileKit 4
  • SpeechKit 7

Summary

VoCon 4.2 introduces new features and deprecates others. In addition there are new acoustic models that are not downward compatible. Applications need to recompile grammars with the new acoustic models.

More Information

For the application to select VoCon 4.2 as the recognizer, it must set the EnginePath and AcousticModel properties for proper initialization. The following example illustrates initializing VoCon 4.2 with the SpeechKit ChantSR class. 

// Nuance VoCon 3200 V4 Reco
NChantSR1.SetNumberProperty(ChantNumberProperty.CNPHighConfidenceThreshold, (int)ChantEngineAPI.CEVoCon4);
NChantSR1.SetStringProperty(ChantStringProperty.CSPEnginePath, "C:\\Program Files\\Nuance\\vocon3200\\EDS_v4_2\\bin\\winx86rls");
// Set the models to the ones you used to compile grammars
NChantSR1.SetStringProperty(ChantStringProperty.CSPAcousticModel, "C:\\Program Files\\Nuance\\vocon3200\\EDS_v4_2\\models\\acmod4_900_enu_gen_car_f16_v1_0_0.dat");
// Set either the common linguistic component or data driven G2P model files
NChantSR1.SetStringProperty(ChantStringProperty.CSPCLCModel, "C:\\Program Files\\Nuance\\vocon3200\\EDS_v4_2\\models\\clc_enu_cfg3_v1_0_3.dat");
//NChantSR1.SetStringProperty(ChantStringProperty.CSPDDG2PModel, "C:\\Program Files\\Nuance\\vocon3200\\EDS_v4_2\\models\\ddg2p_enu_vadvde_ttsasr_large_v2_1_3.dat");

The same initiliziation is used for the GrammarKit ChantGM class, LexiconKit ChantLM class, and ProfileKit ChantPM class.

VoCon 4.2 offers new context tuning parameters:

  • CNPLowConfidenceThreshold - For VoCon 3200 V4.2+, specifies the maximum amount of confidence level that a spoken utterance is out of grammar. If confidence level is equal or below this parameter value, it means that the spoken utterance is out of grammar. If confidence level is above value of this parameter, the spoken utterance can be probably in grammar. The value range is 0 to 10000 with the default value 5000.
  • CNPHighConfidenceThreshold - For VoCon 3200 V4.2+, specifies the minimum amount of confidence level that a spoken utterance is in grammar. If confidence level is equal or above this parameter value, it means that the spoken utterance is in grammar (even through it is the same as CNPLowConfidenceThreshold value). If confidence level is below this parameter value, the spoken utterance can be probably in grammar or out of grammar. The value range is 0 to 10000 with the default value 5000.

Applications can set these properties before enabling grammars with the ChantSR SetNumberProperty method: 

NChantSR1.SetNumberProperty(ChantNumberProperty.CNPHighConfidenceThreshold, 7500);