|
In previous Experience Portal releases, the ASR resources were always acquired at the start of the call and retained until the call was disconnected. This ensured that the required ASR resources were ready at the start of the call to perform speech recognition at any time during the call.
From Experience Portal 7.2.2, for any application configured with the Acquire and release as needed option, latencies can be observed in the speech recognition process. Acquisition of the ASR resources includes establishing the communication channel to the speech recognition server and sending any default recognition parameters. This must occur before the grammars are loaded, and speech recognition is requested. Any network operations such as this can take time to complete and hence increases the time required for the speech recognition process.
The application structure and design should be considered when deciding on whether to enable the Acquire and release as needed option for a speech application. If an application frequently opens and closes speech recognition ports, the application may require a redesign, or one of the other ASR resource control options is more suitable.