Different behaviors when tokenizing apostrophe with nltk in French

Question

I have a French text with two apostrophes. I would like to split all the apostrophes in the same way. For instance:

>