NLU training data - json vs markdown

rasa-nlu

(sfinnie) #1

I’m looking at how to embed entities in my NLU training data. The quickstart uses markdown for mapping utterances to intent. However, it doesn’t include any examples with embedded entities. The examples on github use json which does show entity identification. So, two questions please:

  1. Is one format recommended over the other, or are both equally supported?
  2. Is there any documentation on specifying entities in markdown?

Thanks.


(Akela Drissner) #2

Hmm, have you looked at this documentation? there’s examples of the markdown format at the very start http://rasa.com/docs/nlu/dataformat/ They’re both equally supported, markdown is the recommended option because it’s a lot easier to read


(sfinnie) #3

Thanks @akelad. Hiding in plain view :(. I was looking for details under the “entity extraction” section. Didn’t spot the separate section. Thanks for the pointer.