CODEWITHSUNDEEP
pythonscikit-learntf-idftfidfvectorizer
Jane Sully
Jane Sully

Reputation: 3337

Inputting document-term frequency matrix in TfidfVectorizer()?

I am coming up with a made-up example of bag of words from three documents (I am demonstrating how tf-idf works given a document-term frequency matrix), and I want to transform my bow matrix into a tf-idf matrix. I don't actually have text data, just the number I made up in my example? How can I use that to produce tf-idf output? I am getting the error message "'numpy.ndarray' object has no attribute 'lower'" on the last line (and I am assuming it is because fit_transform is expecting text data. Is it possible to specify or override this somehow?

bow = np.array([[15,0,5,0,20], [20,30,0,25,0], [15,10,10,20,15]])
vectorizer = TfidfVectorizer()
vectorizer.fit_transform(bow)

Upvotes: 1

Views: 410

Answers (1)

KRKirov
KRKirov

Reputation: 4004

You can use TfidfTransformer instead of TfidfVectorizer.

from sklearn.feature_extraction.text import TfidfTransformer
import numpy as np

bow = np.array([[15,0,5,0,20], [20,30,0,25,0], [15,10,10,20,15]])
transformer = TfidfTransformer()
transformer.fit_transform(bow)

Upvotes: 2

Related Questions