i'm using sklearn's CountVectorizer, which allows bigrams. I didn't like the results with full bigrams, so i need to figure out how to make "stacker news" the only bigram in the vocabulary