Text feature basics (CountVectorizer, TF-IDF) in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of CountVectorizer on simple text

What is the output of the following code snippet using CountVectorizer?

ML Python

from sklearn.feature_extraction.text import CountVectorizer
corpus = ['apple orange apple', 'orange banana orange']
vectorizer = CountVectorizer()
X = vectorizer.fit_transform(corpus)
result = X.toarray()
feature_names = vectorizer.get_feature_names_out()
print(feature_names)
print(result)

A ['apple' 'banana' 'orange']\n[[2 0 1]\n [0 1 2]]

B ['apple' 'banana' 'orange']\n[[1 0 2]\n [0 2 1]]

C ['apple' 'banana' 'orange']\n[[2 1 0]\n [1 0 2]]

D ['banana' 'apple' 'orange']\n[[2 0 1]\n [0 1 2]]

💡 Hint

CountVectorizer sorts features alphabetically and counts word occurrences per document.

The CountVectorizer extracts unique words sorted alphabetically: 'apple', 'banana', 'orange'. It counts how many times each word appears in each document. The first document has 2 'apple' and 1 'orange'. The second has 1 'banana' and 2 'orange'.

🧠 Conceptual

intermediate

Understanding TF-IDF importance

Which statement best describes why TF-IDF is useful compared to simple word counts?

A TF-IDF counts the total number of words in a document without weighting.

B TF-IDF reduces the weight of common words and highlights rare but important words.

C TF-IDF only counts words that appear in all documents equally.

D TF-IDF ignores word frequency and only uses document length.

💡 Hint

Think about how common words like 'the' or 'and' should be treated in text analysis.

TF-IDF stands for Term Frequency-Inverse Document Frequency. It lowers the importance of words that appear in many documents (common words) and raises the importance of words that are rare but may carry more meaning.

❓ Metrics

advanced

Comparing vector lengths from CountVectorizer and TF-IDF

Given the same text corpus, which statement about the vector lengths produced by CountVectorizer and TfidfVectorizer is true?

A Vectors from TfidfVectorizer usually have smaller values but the same length as CountVectorizer vectors.

B Vectors from CountVectorizer always have larger length because they count words multiple times.

C Vectors from TfidfVectorizer are always longer because they add extra features.

D Vectors from CountVectorizer and TfidfVectorizer have different lengths because they use different vocabularies.

💡 Hint

Both vectorizers use the same vocabulary by default.

Both CountVectorizer and TfidfVectorizer create vectors with the same number of features (words). TF-IDF scales the counts to weights between 0 and 1, so values are usually smaller but vector length (number of features) stays the same.

🔧 Debug

advanced

Identifying error in TF-IDF code snippet

What error will this code raise and why? from sklearn.feature_extraction.text import TfidfVectorizer corpus = ['cat dog', 'dog mouse'] vectorizer = TfidfVectorizer(stop_words='english') X = vectorizer.fit_transform(corpus) print(vectorizer.get_feature_names_out())

A ValueError because 'english' stop words remove all words leaving empty vocabulary

B AttributeError because get_feature_names_out() does not exist

C TypeError because stop_words must be a list, not a string

D No error; output is ['cat' 'dog' 'mouse']

💡 Hint

Check what words remain after removing English stop words from the corpus.

The corpus words 'cat', 'dog', 'mouse' are not stop words, so they remain. However, if the corpus had only stop words, the vocabulary would be empty causing ValueError. Here, no error occurs actually because words remain.

But in this exact corpus, 'dog' is not a stop word, so no error. So the correct answer is no error.

Re-examining: 'dog' is not in English stop words, so no error.

Therefore, option D is correct, others are wrong.

❓ Model Choice

expert

Choosing the best vectorizer for short text classification

You want to classify very short text messages (like tweets) where common words appear frequently but are not useful. Which vectorizer choice is best and why?

A Use CountVectorizer with max_features=10 to limit vocabulary size.

B Use CountVectorizer without stop words because raw counts capture all info.

C Use TfidfVectorizer with English stop words to reduce common word impact and highlight rare words.

D Use TfidfVectorizer without stop words to keep all words weighted equally.

💡 Hint

Think about how to reduce noise from common words in short texts.

Short texts often have many common words that do not help classification. TfidfVectorizer with stop words removes common words and weights rare words higher, improving model focus on meaningful features.