DialogflowCX - Making entity types mandatory when detecting intent routes

Google Cloud Build with AI Conversational Agents (Dialogflow CX)
1

Hi,

In Dialogflow CX, I am facing an issue with the below scenario, wanted to understand are there any best practices to overcome this.

  • I have a page which has multiple routes and each route expecting an intent to be matched which will trigger the specific webhook endpoint.
  • Even though I have given training phases for each intent tagging along their specific entity types as parameters, the Dialogflow agent fails to detect the right intent & route and I feel this happens because the parameter(entity type) is not being mandated in any of the intent routes and also sometimes if there are keywords which are present in both the intent training phases, it gets confused to detect the right intent.
  • I would like to understand are there ways to mandate the parameter(entity type) for each route within the same page and How to ensure even if the same keyword is present in two intent training phases their entity should always be matched in order to detect and route it.
2

Hi!

What I would recommend is to check your intent to see if there are intent conflicts. By intent conflicts, I mean that you have similar intents causing the issue while routing. If that is the case it could be worth it to re-think about another strategy.

For the transitions itself I will recommend to add more conditions to the transitions to check the parameters that are required to transition: https://cloud.google.com/dialogflow/cx/docs/reference/condition#param
You can also mark a parameter as required.

3

To improve intent matching and routing in Dialogflow CX:

  1. Explicitly specify parameters in intents.
  2. Use slot filling to gather missing information.
  3. Set priority and conditions for routes.
  4. Utilize contexts to maintain conversation flow.
  5. Provide diverse training phrase variations.
  6. Test thoroughly and iterate based on feedback.