LDA with scikit-learn in NLP - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of LDA topic distribution prediction

Given the following code snippet using scikit-learn's LatentDirichletAllocation, what is the shape of the output variable topic_distribution after calling lda.transform(doc_term_matrix)?

NLP

from sklearn.decomposition import LatentDirichletAllocation
from sklearn.feature_extraction.text import CountVectorizer

docs = ["apple banana apple", "banana orange banana", "apple orange orange"]
vectorizer = CountVectorizer()
doc_term_matrix = vectorizer.fit_transform(docs)
lda = LatentDirichletAllocation(n_components=2, random_state=0)
lda.fit(doc_term_matrix)
topic_distribution = lda.transform(doc_term_matrix)
print(topic_distribution.shape)

A (3, 2)

B (2, 3)

C (3, 3)

D (2, 2)

💡 Hint

The output shape corresponds to number of documents and number of topics.

The transform method returns the topic distribution for each document. Since there are 3 documents and 2 topics, the shape is (3, 2).

❓ Model Choice

intermediate

Choosing the correct model for topic modeling

Which scikit-learn model is specifically designed for discovering topics in a collection of documents by modeling word distributions per topic?

A PCA

B KMeans

C RandomForestClassifier

D LatentDirichletAllocation

💡 Hint

This model uses a probabilistic approach to find topics.

LatentDirichletAllocation (LDA) is a probabilistic model that finds topics by modeling word distributions per topic in documents.

❓ Hyperparameter

advanced

Effect of n_components in LDA

In scikit-learn's LatentDirichletAllocation, what does the hyperparameter n_components control?

A The number of topics to find in the documents

B The maximum number of iterations during training

C The learning rate for the optimizer

D The minimum document frequency for words

💡 Hint

Think about how many groups of words the model tries to discover.

The n_components parameter sets how many topics the LDA model will find.

❓ Metrics

advanced

Evaluating LDA model quality

Which metric is commonly used to evaluate the quality of topics generated by an LDA model in scikit-learn?

A F1 Score

B Accuracy

C Perplexity

D Mean Squared Error

💡 Hint

This metric measures how well the model predicts a sample.

Perplexity measures how well the LDA model predicts unseen data; lower perplexity indicates better generalization.

🔧 Debug

expert

Identifying error in LDA input data

What error will occur if you pass a dense numpy array instead of a sparse matrix to scikit-learn's LatentDirichletAllocation fit method?

NLP

import numpy as np
from sklearn.decomposition import LatentDirichletAllocation

X = np.array([[1, 2, 3], [4, 5, 6]])
lda = LatentDirichletAllocation(n_components=2)
lda.fit(X)

A ValueError: Input must be a sparse matrix or a non-negative array

B No error, code runs successfully

C AttributeError: 'numpy.ndarray' object has no attribute 'tocsc'

D TypeError: Expected sparse matrix but got dense array

💡 Hint

LDA accepts both dense arrays and sparse matrices as input.

LatentDirichletAllocation accepts both dense numpy arrays (non-negative) and sparse matrices. No error occurs.