from sklearn.feature_extraction.text import CountVectorizer

bigram_vectorizer = CountVectorizer(
    ngram_range=(2, 2),               # only bigrams
    max_features=1000                 # top 1000 bigrams by frequency
)