Hello, I’m using my model now and discovered rasa x, I wanted to use it for interactive learning. Right now I have a fallback policy, so when no intent is guessed, action_fallback is used. However I don’t see any way to specify that there is no intent in the message. One way to fix it is to add chitchat intent, but it looks like some hack, also I think it would downgrade my model because chitchat will contain everything except actual intent, but for example if I add one more intent, core will be confused, because chitchat will also be recognised as such intent. What to do in that case?
Hi @gsp0din, every processed message should go into an intent. The bot only knows what it has training data for, it doesn’t know how to determine if something “doesn’t belong” to any of your given categories. For this use case, we suggest adding an
out_of_scope intent to classify any user input your bot cannot handle. This intent can directly trigger the fallback action if you want it to.