I think the reason is that pickled Punkt tokenizer available in nltk_data was trained on byte strings, and implicit byte strings fail under Python 3.x. Other pickled data installable with ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results