Response Selector


I was listening to Rasa Whiteboard videos on Response Selector. I could understand how Response Selector concentrates on a specific intent. But what I couldn’t understand is

  • I checked Rasa Github Demo where faq and chitchat where present as separate files in data folder apart from usual nlu.yml file. Is this a specific requirement to exclude DIET from considering this intent while training?

  • Similarly faqs.yml file was present inside Stories folder which defines the appropriate action for each intent. Is this also a required to not mingle this with usual stories.yml file?

  • Response Selector uses the same Architecture as DIET, then why do we need a separate parameter? why can’t we just include all the data in regular nlu and stories and train the DIET with it?

Can someone please help me to understand this?