When running your bot in the shell have you tried using rasa shell --debug ? This gives you a lot more information than just running it normally.
Another suggestion is that each bot response has 'intent_ranking" which similar to your screenshot contains all the probabilities of each intent for a user’s message. Perhaps you could create a new custom action to test a bot’s reponse, so you could just do:
query = tracker.latest_message
print(query)
This will just print the bot response to the terminal in which you’re running your actions and you will be able to see the intent rankings within that.
P.s. tracker.latest_message[‘intent_ranking’] should return just the probability scores if you want that
Had forgotton about the split train test, which is producing an intent histogram with confidence.
Or the alternative cross validation test.
I was little bit confused, because the rasa test on end-to-end stories reported an intent histogram and intent config with no errors. Reason is, end-to-end-test doesn’t split intents, I guess