Quick bit of background, we have configured an Email Channel that is used to determine the type of case to create based off a selection of 50 different Intents.
With the large number of Intents, Emails are being miscategorized and IVA Intent sometimes just outputs the completely wrong Intent.
These Emails can be large in detail which again isn't helpful for Intent Model.
We have configured keywords for each Intent & various phrases from the Email bodies.
Any recommendations as to how best to Model the Intent with the variety we have?
***Edited by Moderator: Pallavi to update platform capability tags***
You should start using machine learning models. Also, while training you should remove any extra text which may confuse the model. Good quality of training data results in higher accuracy of detection.