Word embeddings concept (Word2Vec) in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Word2Vec Training Objective

Which of the following best describes the training objective of the Word2Vec Skip-gram model?

A Predict the center word given the context words.

B Cluster words into groups based on their frequency.

C Reduce the dimensionality of word vectors using PCA.

D Predict the context words given the center word.

💡 Hint

Think about what the model tries to guess during training.

The Skip-gram model learns to predict surrounding context words given a center word, helping it capture word relationships.

❓ Predict Output

intermediate

Output of Word2Vec Vector Similarity Calculation

Given the following Python code using Gensim's Word2Vec, what is the output of the similarity calculation?

ML Python

from gensim.models import Word2Vec
sentences = [['cat', 'sat', 'on', 'the', 'mat'], ['dog', 'barked', 'at', 'the', 'cat']]
model = Word2Vec(sentences, vector_size=10, window=2, min_count=1, epochs=10)
sim = model.wv.similarity('cat', 'dog')
print(round(sim, 2))

A A float value close to 0.5

B A float value close to 1.0

C A float value close to -1.0

D Raises a KeyError because 'dog' is not in vocabulary

💡 Hint

Both 'cat' and 'dog' appear in the sentences, so similarity is computed.

The similarity between 'cat' and 'dog' is positive but less than 1 because they appear in related contexts but are different words.

❓ Model Choice

advanced

Choosing the Right Word2Vec Architecture

You want to train word embeddings that capture rare word relationships well in a small dataset. Which Word2Vec architecture is best?

A Hierarchical clustering model

B Skip-gram model

C Principal Component Analysis (PCA) on one-hot vectors

D Continuous Bag of Words (CBOW) model

💡 Hint

One architecture is better at learning from rare words.

The Skip-gram model performs better on rare words because it predicts context from a single word, allowing it to learn better representations for infrequent words.

❓ Hyperparameter

advanced

Effect of Window Size in Word2Vec

What is the effect of increasing the window size parameter in Word2Vec training?

A The model considers a larger context around each word, capturing broader semantic relationships.

B The model trains faster because it uses fewer context words.

C The model ignores stop words during training.

D The model reduces the vector size to speed up training.

💡 Hint

Window size controls how many words around the center word are used.

A larger window size means the model looks at more surrounding words, helping it learn broader meanings but possibly less syntactic detail.

❓ Metrics

expert

Evaluating Word2Vec Embeddings Quality

Which metric is commonly used to evaluate the quality of trained Word2Vec embeddings on a word analogy task?

A Silhouette score of word clusters

B Cross-entropy loss on training data

C Accuracy of correctly answered analogy questions

D Mean Squared Error between embeddings

💡 Hint

Think about how well embeddings solve analogy puzzles like 'king - man + woman = ?'

Accuracy on analogy tasks measures how often the embeddings correctly solve word relationship puzzles, reflecting semantic quality.