Rasa entity extraction like siri,google and alexa

[htayyab63 (Hamza)]

Hello everyone, Newbie here asking for your attention and help.

Lets assume a scenario:

the NLU content has the intent so called rasa bird but dont have penguin bird

The user asks “hey penguin bird”

So here i want the ai to reply as “who is penguin bird, did you mean rasa bird” ?

then user replies “i mean penguin bird!”

rasa replies: “I DONT KNOW ANY PENGUIN BIRD, you might have mistaken me for someone else, i am RASA BIRD"

So does anyone know how to make this scenario work like how should i make rasa find the relevant context that is “bird” and do the action like find is bird similar to any other context in the system then replies back or can anyone point me to the documentation side which contains information related to my question or anyone have did it earlier who can share their codes right here?

[anca (Anca Lita)]

Hi @htayyab63 In order to enable a story example as you shared, you might try the following:

• defining a slot and an entity with the same name (bird) will enable the slot auto-fill functionality;

• your nlu intent example can then be implemented with an entity annotation; look-up tables could be useful too here

• implement a custom action that checks if the slot value is rasa by using tracker.get_slot and dispatch a certain message, otherwise dispatch a different message

• add your action to the domain and then use the action in your story paths or rules

Alternatively, you could also read about conditional response variations based on certain slot values or about fallback and out-of-scope messages.