Virtual Agent Triggers Incorrect Topics

[RenatoMendoza]

Hi everyone,

I have a list of utterances that are triggering incorrect topics, I have used the feedback feature to test the model and improve the prediction. Train and published it and it still triggers an incorrect topic.

How can I avoid this behavior?

I should add that the Expert Feedback Loop is not working on our instance (there is a known bug and SNOW is working on it)

Thanks in advance.

[Chris D]

97% is very very high - that's implying that you have an utterance that nearly identically matches the phrase "Download Now Mobile".

Can you share the list of utterances you have in the InTuneEnrolling Intent? (export from the sys_nlu_utterance table for easiest access)

[Tricia Cornish]

@RenatoMendoza  - Do you have a intent for downloading apps where Now Mobile is an option and is that intent in the same model as Intune enrollment? 
Another option is to create a pattern vocabulary entity that "Now" is not a point in time, the caps version indicates Servicenow. 

One other suggestion that I also try when testing doesn't seem to pick up my changes ( for anything) is clearing the cache and / or trying this in an incognito window.