Utilize the Rasa framework to build a goal oriented bot


(Vassilis) #1

Hello everyone,

I would kindly like to ask for your input/expertise on how to utilize the RASA framework to make it a goal/task oriented bot.

More specifically, I am seeking for any examples within the RASA framework to create an “intelligent” agent which is capable to use Reinforcement Learning type of algorithms in order to :

  • Negotiate a with the end user
  • Make recommendations under constraints (like the cuisine, price range, location, etc.)

I have been reading on this topic, from various sources like:

  1. A New Approach to Conversational Software – Rasa Blog – Medium
  2. https://arxiv.org/pdf/1706.05125.pdf

I am particularly interested in examples illustrating how RASA can incorporate such paradigms.

RASA community, it would be a pleasure to hear back from you, and thanks in advance for your feedback/recommendations.

Thanks in advance


(Remy Reurling) #2

This is a rather broad question to ask for a forum format :slight_smile: What exactly are you looking for? Rasa can be fully customised to your wishes and use case. The only limiter is your own expertise. I’d say, go try it out and see if you can make a mock-up of the bot you want to deploy later. You’ll run into more concrete questions and problems that’ll help drive you and this community to building better bots.

gl!


(Vassilis) #3

Cheers Remy,

I would be grateful if you could provide examples/chatbot mock-ups that:

  • Use Rasa and RL type of algorithms for negotiating with end-users

Best Regards, Vassilis


(Remy Reurling) #4

Personally, I can’t. I’ve never done a project that has the bot negotiate with a user to that end. You’d have to see if someone else could. Searching these forums for keywords might help in identifying someone that is working on a project like yours.