In our project, we have 250+ intents stored in one flow, with 3 levels of hierarchy. Its working nicely but some first-level intents are confused with eachother, according to the mouse-over feedback. However, we inherit training phrases, sometimes from 20+ sub intents for each of these 1st level intents.
It seems the inherited training phrases are causing the confusion on the 1st level. We had to inherit phrases to improve performance, so just switching those off is not really an option. Also I think this issue still occurs when there is no inheriting.
Request is for 2nd/3rd level intents to report confusion with specific other intents outside of their family.
ah I should mention that in cross-validation tests we almost never see a customer question end up in the wrong family. In other words, even though Cognigy tells us there is confusion on 1st level, the fact that we are not getting more specific feedback from the system is not a big pain point right now.
I can imagine that over the next few years, as Cognigy bots are getting bigger and bigger, this issue becomes more important
Hi Willem Don, can you submit the feature request to firstname.lastname@example.org? Then our support team will prioritize it and track back to you.