server.mrcp2.osrspeechrecog.mrcpdefaults.
VSP.server.osrspeechrecog.result.mediatype

Specifies the media type of the recognition result returned to the application.

Value

MIME media type supported.

DEFAULT: application/x-vnd.speechworks.emma+xml;strictconfidencelevel=1;mrcpv=2.06

Note: strictconfidencelevel is not supported for Dragon Voice recognition.

How to set

If using the Management Station, set on the Speech Server service. If not using Management Station, set in an NSS configuration file (user-NSSxx.txt).

Usage

Rarely used.

The default value returns NLSML (Natural Language Semantic Markup Language) results. In addition to the default, this parameter accepts the following values:

Accepted value Description

application/x-vnd.nuance.emma+xml

 Uses the EMMA format.

application/x-vnd.speechworks.emma+xml

 Uses the NLSML format.

application/x-vnd.speechworks.recresult+xml

 Uses the SpeechWorks format.

application/x-vnd.speechworks.wordlattice+xml

 Uses the wordlattice format for getting raw recognition results.

Note: When you use the flag mrcpv=2.06 the returned confidence scores are floating point numbers between 0 and 1 (up to 2 decimal values). Without the flag, the scores are between 0 and 100.

The strictconfidencelevel flag always has the value "1". When you specify it, Recognizer returns only results with a confidence level that is higher than Confidence-Threshold. When you remove the flag (by removing the string "strictconfidencelevel=1"), Recognizer returns all recognition results.

For example, let’s say our nbest list + confidence is:

  1. .5
  2. .4
  3. .3

If strictconfidencelevel is set and minconf is .45, then our nbest list will be:

  1. .5

Whereas if strictconfidencelevel is not set, the nbest list will be:

  1. .5
  2. .4
  3. .3

Supported for MRCPv2 browsers. (Ignored for MRCPv1 browsers.)

Related parameters