Training NLU Models with only Lookups and Synonyms

Rasa Open Source
1

I currently have a use case where i need to extract entities. The entities are country names and its synonyms. In order to extract them, i decided to use Look Up Tables and synoyms. Ideally, i do not need a custom model trained liek the DIET Classifier or the CRF entity extractor. Simple lookup and synoym would be sufficient. I don’t need any intent to be identified as well but looks like that is mandatory, so we, added a sample intent and some example sentences as well.

However, when i try to train is this model, synoyms are not being picked up. So, it looks like i have to put every value under the synonym section under look up as well?

This is how I have the nlu.yml file now

version: "3.1"
nlu:
- synonym: United States Of America
  examples: |
     - U.S.A.
     - America
     - U.S.
     - US
- lookup: country
  examples: |
     - United States of America
     - India
     - United Kingdom

So, in this, it looks like the EntitySynonymExtractor is not working at all. It doesn’t matter if i place is before or after RegexEntityMapper, it just doesn’t identify the country if i use any of the synonym words. What can be done here?