I use from nltk.tokenize import word_tokenize which needs punkt. In code you can download it with nltk.download('punkt').
I do have nltk as a requirement, but there is no target nltk[punkt]. Is there another way I set this in my setup.py as a requirement? What is the recommended way of dealing with this data dependency of nltk?
Current "solution"
Currently, I just call nltk.download('punkt') within the function ... hence every single time I call this function, it is slowed down.
