Skip to Content
Learn
Term Frequency–Inverse Document Frequency
What is Tf-idf?

Term frequency-inverse document frequency is a numerical statistic used to indicate how important a word is to each document in a collection of documents, or a corpus.

When applying tf-idf to a corpus, each word is given a tf-idf score for each document, representing the relevance of that word to the particular document. A higher tf-idf score indicates a term is more important to the corresponding document.

Tf-idf has many similarities with the bag-of-words language model, which if you recall is concerned with word count — how many times each word appears in a document.

While tf-idf can be used in any situation bag-of-words can be used, there is a key difference in how it is calculated.

Tf-idf relies on two different metrics in order to come up with an overall score:

  • term frequency, or how often a word appears in a document. This is the same as bag-of-words’ word count.
  • inverse document frequency, which is a measure of how often a word appears in the overall corpus. By penalizing the score of words that appear throughout a corpus, tf-idf can give better insight into how important a word is to a particular document of a corpus.

We will dig into each component of tf-idf in the next two exercises.

Instructions

1.

The code in script.py is used to apply tf-idf to a corpus of documents and display the scores as a DataFrame in the web browser component. The current code is incomplete, so the DataFrame is not yet appearing!

Like with many natural language processing tasks, we must preprocess our text data before applying a technique. Paste the below line of code into the “preprocess documents” section of script.py to preprocess each document in the corpus. Then run the code.

processed_corpus = [preprocess_text(doc) for doc in corpus]
2.

You should now see what is known as a term-document matrix in the browser. Each row of the table represents a term, and each column of the table represents a document. The value in each cell indicates the tf-idf score for each term-document pair.

Try changing the text in document_1, document_2 and document_3 and re-running the code. How do the tf-idf values change?

Folder Icon

Sign up to start coding

Already have an account?