Naive Bayes for text in NLP - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

How does Naive Bayes handle word independence in text classification?

Naive Bayes assumes that words in a text are independent of each other when calculating probabilities. What does this assumption mean in simple terms?

A Words are grouped into pairs and their combined frequency is used for classification.

B The order of words in the text is the most important factor for classification.

C Each word's presence or absence does not affect the presence of any other word in the text.

D Words are ignored and only sentence length is used for classification.

💡 Hint

Think about how the model treats each word when calculating probabilities.

Naive Bayes treats each word as if it appears independently from others, simplifying calculations by ignoring word order or combinations.

❓ Predict Output

intermediate

Output of Naive Bayes prediction probabilities

Given the following Python code using sklearn's MultinomialNB for text classification, what is the output of the prediction probabilities?

NLP

from sklearn.feature_extraction.text import CountVectorizer
from sklearn.naive_bayes import MultinomialNB

texts = ['I love apples', 'I hate bananas']
labels = [1, 0]

vectorizer = CountVectorizer()
X = vectorizer.fit_transform(texts)

model = MultinomialNB()
model.fit(X, labels)

new_text = ['I love bananas']
X_new = vectorizer.transform(new_text)
probs = model.predict_proba(X_new)
print(probs)

A [[0.5 0.5]]

B [[0.75 0.25]]

C [[0.25 0.75]]

D [[1.0 0.0]]

💡 Hint

Look at how the model was trained and the new text features.

The new text 'I love bananas' contains 'love' from class 1 and 'bananas' from class 0. Both classes have 'I', and with Laplace smoothing, the likelihoods balance exactly, giving equal probabilities [[0.5 0.5]].

❓ Model Choice

advanced

Choosing the best Naive Bayes variant for text with word counts

You have a dataset of text documents represented as word counts. Which Naive Bayes variant is most suitable for this data?

A Multinomial Naive Bayes

B Complement Naive Bayes

C Bernoulli Naive Bayes

D Gaussian Naive Bayes

💡 Hint

Consider which variant models count data well.

Multinomial Naive Bayes is designed for count data like word frequencies, making it ideal for text classification with word counts.

❓ Hyperparameter

advanced

Effect of smoothing parameter alpha in Naive Bayes

In Multinomial Naive Bayes, what is the effect of increasing the smoothing parameter alpha?

A It removes stop words automatically from the text data.

B It increases the influence of rare words by adding more weight to zero counts.

C It decreases the model's ability to handle unseen words by ignoring them.

D It prevents zero probabilities by adding a small count to all features, smoothing the model.

💡 Hint

Think about how smoothing helps with words not seen in training.

Alpha adds a small positive value to word counts to avoid zero probabilities for unseen words during prediction.

❓ Metrics

expert

Interpreting Naive Bayes classification report metrics

A Naive Bayes text classifier reports the following metrics on test data: precision=0.8, recall=0.5, accuracy=0.7. What does this tell you about the model's performance?

A The model perfectly balances false positives and false negatives.

B The model is good at finding positive examples but misses many actual positives.

C The model predicts positives too often, causing many false positives.

D The model has high recall but low precision.

💡 Hint

Recall measures how many actual positives are found, precision measures correctness of positive predictions.

Precision 0.8 means most predicted positives are correct, but recall 0.5 means it misses half of the actual positives, so it is conservative in positive predictions.