![]() |
In Experience Portal 7.1 and later, the default values of MPP timers that control speech server events and protect speech server failures have been re-calibrated. Depending upon the application deployed, you can modify these values.
| Timer Name | Previous Release | Experience Portal 7.1 |
|---|---|---|
| mpp.mrcpsessionrefresh.timer | 40 | 40 |
| mpp.mrcppoststartofspeechevent.timer | 20 | 80 |
| mpp.mrcppostrecogstartedevent.timer | 20 | 20 + Dynamic value |
Use this timer as a guard against a speech server failure while playing prompts or using ASR. You can also use this timer to generate some traffic when speech is not being used to prevent a speech server session timeout.
mpp.mrcpsessionrefresh.timer value should always be set to 20 seconds lesser than the speech server session timeout value. This value must be larger than the largest prompt to be played and must be longer than the longest expected ASR response.
The session timeout value is 60 seconds by default for Nuance. On each Nuance speech server, modify the value in the server.mrcp2.sip.sessionTimeout for MRCP V2 and in server.mrcp1.rtsp.sessionTimeout for MRCP V1 in the configuration file. Other vendors may have a different way to update this value.
For example, if you need to play a prompt that is 180 seconds long, then set the value of mpp.mrcpsessionrefresh.timer to 190 seconds on each MPP server and set the speech server session timeout to 210.
Use this timer as a guard for cases where the MPP has received a Start of Speech Event, but has not received Recognition done or an error response because the speech server has gone out of service.
The value should always be 20 seconds more than the maximum speech value set for the speech server.
The value on the speech server can be updated in the following ways in the order of precedence:
As a VXML property: This method is the preferred method as it is specific to Individual Grammar. property <name="maxspeechtimeout" value="10s"/>
By logging into the EPM and changing the value of Recognition Timeout:
Log on to EPM > Applications > Change Application page.
In change the value of Recognition Timeout.
This is a parameter that applies to all the recognition requests for the particular application.
By changing the maxspeechtimeout property: On the Nuance Speech Server, change the maxspeechtimeout property in the Recognizer’s Baseline.xml file.
(Default is 22s for V1 and 10s for V2)
Repeat Step 3 on all Speech Servers.
Other vendors may have a different way to specify this value.
If the maximum value set for speech for all applications is less than 60 seconds, then there is no need to modify the MPP configuration. For values greater than 60 seconds, modify the MPP configuration value for mpp.mrcppoststartofspeechevent.timer to the maximum speech value + 20 seconds on each MPP server.
Use this timer as a guard against a speech server failure while waiting for recognition-complete. The guard timeout value is dynamically calculated by adding the configured value (mpp.mrcppostrecogstartedevent.timer) to the no-input timeout value passed to the speech server.
![]() | Note: |
|
As the guard value is dynamically calculated, the MPP configured value should not be modified. |
A no-input timeout value is passed to the speech server as RECOGNITON-START-TIMERS in case of MRCP V1 or START-INPUT-TIMERS in case of MRCPv2.
This MPP guard value starts with mpp.mrcppostrecogstartedevent.timer that is set to 20 seconds, and then adds the no-input timeout value set through one of the following method in order of precedence:
Change the value of VXI: The default value of VXI for timeout is 7 seconds. You can change this value per recognition request as follows:
<property name="timeout" value="10s"/>
For example, the guard’s value will be 20 + 10 = 30 seconds.
Globally updating the no-input timeout value: This method updates the no-input timeout value globally in the Application, unless each recognition request has a property tag used for setting the timeout value as mentioned in method 1.
Log on to EPM and navigate to Applications page > Speech Parameters > ASR.
Update the value of No Input Timeout.
For example: No Input Timeout = 45000 milliseconds
This will set the value of the guard to 20 + 45 = 65 seconds.
![]() | Note: |
|
The no-input timeout value passed to the speech server should not be greater than the value set for the speech server session timeout. |