Entity extraction priority

Rasa Open Source
1

Hi, I’m facing a big issues when it comes to entity extraction, let me give an example of what’s happening.

Imagine your asking for a product and the user says chocolate, the chatbot is able to understand that chocolate is an entity of type product and everything works fine, but, things get a little more crazy when the entities are of multiple words, for example: Chocolate cake. In many other cases were that “multiple word entity” has no other “subentity” inside, the nlu extracts it fine, but in this other case ( Chocolate cake ), instead of extracting the whole entity it gathers chocolate.

My question is: Is there any way to prioritize the entity extraction, like adding a number on a lookup table?

2

@palagonKevo can you please share the nlu.yml file with the training examples where you have mentioned chocolate and chocolate cake as entities? Please share the rasa --version also. -Thanks Nik

3

The nlu.yml is pretty heavy, and in spanish btw, but I can show you the intent and a tiny part of the lookup table.