Configuring Recognizer with the browser

To set Recognizer parameters with the voice browser, use MRCP standard headers or Vendor-Specific-Parameters header in one of these requests sent to Speech Server:

  • SET-PARAMS
  • GET-PARAMS
  • DEFINE-GRAMMAR
  • INTERPRET
  • RECOGNIZE (MRCPv2 only)

(Some parameters get set implicitly as part of the standard MRCP messaging. When the browser requests features defined in the MRCP specification, Speech Server automatically sets the corresponding parameters to appropriate values.)

Setting values when activating a grammar

The browser uses the DEFINE-GRAMMAR method with the MRCP Vendor-Specific-Parameters header to set parameters when activating a grammar. The format is: 

Vendor-Specific-Parameters: parameterName=parameterValue

Setting values for the duration of a recognition

The browser can use the RECOGNIZE method to set Recognizer parameters.

The value remains in effect for the duration of the request. It reverts to the previous setting when the recognition is complete.

Setting values for the duration of a session

The browser can also use the SET-PARAMS method to set most Recognizer parameters.

The value remains in effect for the duration of the session.

This sequence changes the value for some recognitions during the session:

  1. Use GET-PARAMS to get current value.
  2. Set the parameter. For example: 
    MRCP/2.0 96 SET-PARAMS 543257
    Confidence-Threshold: 0.2
  3. Perform recognitions with the new value.
  4. Reset the original value, and perform more recognitions.