DSTC2 Intent classifier

I was reading the documentation about the dstc2 go bot framework and i manage to get how the classifier depends on the dialogue acts to predict the probabilites of the intents labels but i didn’t get how does the intent iterator extracts artificial intents? what techniques are applied here? and where do we set our intents so that the model can look for them

Hey! Typically you should use RASA configs to configure intents etc