I am several days testing Rasa-NLU, which internally uses spaCy. I had a great disappointment about the Portuguese language. Trying to figure out how to improve the training data, I found an excellent script comparing spaCy with udpipe that can be checked in this link and also in the image below.
I would like to know if I can continue using Rasa-NLU, but replace the engine spaCy for udpipe?