Dynamic Stories and role of policies, trackers

Hi everyone. I have been working on rasa core for the last two months. Basically, I am trying to figure out a way to dynamically write stories because they seem to be very inflexible and hence cannot hold for a longer conversation. The model actually has learnt when the conversation needs even in the absence of intent(something that signal that it is the end of the conversation). I need some ideas in this regard.

Also, I am having a hard time understanding the in-depth role of policies and tracker. I am aware of the flow but questions like what exactly be the individual input/output of these systems, is bothering me.

I’d really appreciate any help in that matter as well.

Could you elaborate a bit on what kind of stories you’re talking about?

I am trying to make a retail store bot that could one through the super store. The bot needs to be conversationalist. So I have written trying to accommodate different scenarios. But an actual conversation can be limitless. For example ,

  • greet
  • utter_greet
  • utter_ask_howcanhelp
  • inform{“object”: “TV”}
  • utter_agree
  • utter_ask_size
  • inform{“size”: “30 inch”}
  • utter_ask_brand
  • inform{“brand”: “sony”}
  • action_basic_info
  • inform{“price”: “10000”}
  • action_basic_info
  • inform{“display_features”: “LED”}
  • action_basic_info
  • inform{“connectivity_features”: “bluetooth”}
  • action_basic_info
  • final_decision
  • utter_awesome
  • utter_end_conversation

This is a glimpse of one of my story. There are lot of story similar to this with little tweaking in the flow.But how do i make stories that can last for any duration. My current bot has learnt to end conversation after couple of Q & A (which makes sense) but how to do i scale it. Do i need accommodate conversation for long range of duration to achieve my goal ?

1 Like