RASA NLU: Multiple entity extraction from Single intent

Rasa Open Source
1

I am trying to retrieve different entities from a single intent using rasa nlu below is the nlu part of training data

##intent communicate
 - communication address of [JhonDoe](name)
 - communication address of [Engineer](designation)

When I try this approach, I correctly get intent as communicate but even a spelling mistake of the entity value(like engineer) will result in entity list as empty. So for the above scenario how to tackle the problem to detect the different entity from same intent for same question?

1 Like
2

Hi @sameesh-s

Just to be sure: your example does work when the user does not make any spelling mistakes, right? So if the issue is about typos, you have to train rasa with misspelled words. That is, you have to give more examples. Otherwise, how could a computer know if two strings mean the same thing or not? For example, you could write

## intent:communicate
  - communication address of [John Doe](name)
  - communication address of [John Do](name:John Doe)
  - communication address of [Jon Doe](name:John Doe)
  - communication address of [JD](name:John Doe)

and so on. You can also use synonym lists (see Training Data Format).

1 Like
3

Thanks