I'm using python and I want to get the TFIDF representation for a large corpus of data, I'm using the following code to convert the docs into their TFIDF form.

from sklearn.feature_extraction.text import TfidfVectorizer
tfidf_vectorizer = TfidfVectorizer(
    min_df=1,  # min count for relevant vocabulary
    max_features=4000,  # maximum number of features
    strip_accents='unicode',  # replace all accented unicode char 
    # by their corresponding  ASCII char
    analyzer='word',  # features made of words
    token_pattern=r'\w{1,}',  # tokenize only words of 4+ chars
    ngram_range=(1, 1),  # features made of a single tokens
    use_idf=True,  # enable inverse-document-frequency reweighting
    smooth_idf=True,  # prevents zero division for unseen words
    sublinear_tf=False)

tfidf_df = tfidf_vectorizer.fit_transform(df['text'])

Here I pass a parameter max_features. The vectorizer will select the best features and return a scipy sparse matrix. Problem is I dont know which features are getting selected and how do I map those feature names back to the scipy matrix I get? Basically for the n selected features from the m number of documents, I want a m x n matrix with the selected features as the column names instead of their integer ids. How do I accomplish this?