How can I consider the punctuation at the intent classifier?


I watched some Rasa Masterclass videos and I noticed in one of the videos that, by default, Rasa doesn’t use the punctuation.

In my pipeline, I am using spacyTokenzier, which by itself (when string is tokenized using only this lib) doesn’t remove punctuations and keep them as separated tokens. As my custom word embedding have word vectors for some of them and as I want to consider the punctuation information as an additional criteria to differ questions from orders, I want to keep the punctuations through the entire pipeline. How can I do this or is there any documentation that I should look in particular?