Specifies the maximum duration of an utterance collected from users.
|
Value |
Integer: -1, 1–INT_MAX (milliseconds) DEFAULT: 22000 (22 seconds) |
|
How to set |
Shown in precedence order (see Rules of parameter precedence): In a VoiceXML application. However, if set in VoiceXML, the system ignores values longer than the recognizer default. Set in the browser, using the Vendor-Specific-Parameters header in a SET-PARAMS method. In a parameter grammar. In Management Station set on the Nuance recognition service. If not using Management Station, set in a Recognizer configuration file (User-nrsxx.xml). |
|
Usage |
Frequently used to tune individual collections. It protects recognizer resources when a call fails to disconnect, or when there is endless noise on the line (that the recognizer might interpret as speech). |
|
Performance |
Longer values use more CPU and memory, but ensure the caller’s entire speech is captured. Shorter values save resources, but risk interrupting the caller’s speech. |
Note: The Nuance Recognizer version of this parameter is a vendor-specific parameter. It is different from the VoiceXML maxspeechtimeout parameter, and cannot be set directly in a VoiceXML application.
The recognizer begins measuring the utterance length from the begin-of-speech point in the audio stream.
An MRCP setting overrides any <meta> setting, and there are various implications depending on the endpointer mode:
- In begin-only mode, this parameter is ignored.
- In magic-word or selective-bargein mode, this timer is reset for each new audio chunk. If swiep_magic_word_max_msec is set, and is smaller than maxspeechtimeout, then Recognizer uses the lesser setting.
Note: The Krypton recognition engine does not support hotword mode recognition, including magic word and selective barge-in.
Applications can set this parameter via a VoiceXML property. However, the value is ignored if it is higher than the Nuance Recognizer default.
Note the difference in the value format (the voice browser would convert this to "7000"):
<property name="maxspeechtimeout" value="7s"/>
This property generates a Recognition-Timeout MRCP header that takes the value and converts it to milliseconds. The voice browser may then multiply the value by a safety factor to provide a margin of error. This example takes the 7000 milliseconds and multiplies that value by 1.25:
Recognition-Timeout: 8750
The example below shows the feature being set to 25 seconds in a grammar file. The setting overrides the default value, but only for this grammar.
<meta name="maxspeechtimeout" content="25000"/>
Related parameters:
