Genesys CTI User Forum
Genesys CTI User Forum => Genesys CTI Technical Discussion => Topic started by: nonny on July 13, 2015, 07:21:25 AM
-
Hi guys
For a few reasons that have me frustrated, my customer is wanting to stick with Predictive Dialing Mode for their OCS Campaign but want their customers to have messages left on their answer machines. Because it's Predictive Mode, the am-beep-detect option can't be used. I haven't really done anything in Nuance ASR before but was looking to do something like the following:
1 - Have answer machine detected treatment route to a RP that in turn calls a Composer app. In that app I'm looking at the following logic:
2 - Play a silence prompt as part of a Grammar Menu and then listen for a garbage word ("tiddlywinks"). If there is no voice input the assumption is that the greeting has finished. If the voice input doesn't match "tiddlywinks" then it's assumed the greeting is still playing.
The problem is I've set the Grammar Menu to retry 2-3 times but those retries are so fast that essentially the greeting is still playing when this logic has completed. So how do I tell it to essentially either wait until it hears silence, or retry for what could be a high number of loops until it gets a noinput? It doesn't seem good coding practice to do the high count loop option. Should I use the loop block as opposed to the retry options on the Grammar Menu block?
Has anybody come up with something similar that works?
-
If your VAD already marked it as AM, why do another detection? The message was already played while was being analyzed on the AM detection, so maybe you have one to three seconds as much of message remaining or none at all. So after that just play the message.
Enviado de meu C6602 usando Tapatalk
-
From a customer experience viewpoint getting a message left halfway through, and then repeated again isn't the best. Simple from a technical standpoint, not good from a service viewpoint.
-
I don't see how to do that. ASR searches for voice patterns, not silence.
The beep of an AM is not always there tho.
What if it is an Automated Attendant?
VAD/AM analyze wave lengths of sound durations. That is why is able to detect AM and difference from human voice. You are asking to analyze that, wave lengths on app level, and I don't know any VAD that provides that resource for an application level.
Enviado de meu C6602 usando Tapatalk