Minimum data for updating acoustic models.
|
Value |
Integer: 500–100000 (utterances). |
|
Default |
(depends on language) The typical default is 1000. |
|
How to set |
Shown in precedence order (see Rules of parameter precedence): Set in the browser, using the Vendor-Specific-Parameters header in a RECOGNIZE or SET-PARAMS method. In Management Station set on the Nuance recognition service. If not using Management Station, set in a Recognizer configuration file (User-nrsxx.xml). |
|
Usage |
Seldom used. Most Recognizer installations do not change this parameter. If the minimum is not reached at the time of an update, the data remains in the statistics files and the update is not performed. |
|
Performance |
Lesser values speed the Recognizer’s self-learning because fewer utterances are needed for each cycle of adaptation. However, lesser values can also cause a (typically undesired) recognition bias towards those utterances. Larger values have the opposite effect: self-learning is slowed, and the adaptations are more generalized and desirable. |
Defines a minimum amount of data (number of utterances) in the statistics files used by the Recognizer to update its acoustic models. This feature is part of the Recognizer’s self-learning technology.
This parameter interacts closely with swirec_acoustic_adapt_rate. When making changes to their values, it is best to change one parameter or the other. (Typically, you would change swirec_acoustic_adapt_rate.) Avoid changing both parameters at the same time.
For an overview of adaptation, see Controlling the self-learning recognizer.
Related parameters:
