Categorisation of Rasa and other CA

Hi all

I am reading this old but widely cited paper: McTear, M. F. (2002). Spoken dialogue technology: enabling the conversational user interface. ACM Computing Surveys (CSUR) , 34 (1), 90-169.

It has this classification that many authors still use:

Where would you classify Rasa?

  • It has finite-state elements (e.g forms and rules)
  • It has frame-based slots filling
  • But there is adaptation of the stories (agent-based)

Or is this classification too outdated and should not be used any longer? thanks

rafa

Hi @rafa21. It’s a great question. I think this categorization is valid, but at Rasa we look at this through the lens of 5 levels of conversational AI assistants.

Based on the framework you shared, Rasa would be positioned amongst the agent-based dialogue management systems. While it has the features of Finite-state and Frame-based systems as well, the main point that distinguishes Rasa is the fact that Rasa’s dialogue management models (like TED, for example) take the context of the conversation in account when predicting the next action.

Thank you @Juste that is useful.

cheers

rafa