Objective 2.4 Apply Problem-Specific Transformations to Data Sets CERTIFIED-DATA-SCIENCE-PRACTITIONER Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A data scientist is building a topic modeling system for news articles. They want to preprocess the text so that common words like 'the', 'is', and 'and' do not dominate the topic distributions. They also want to split hyphenated words (e.g., 'state-of-the-art') into separate tokens. Which preprocessing steps should they apply before tokenization?

   • AUse TF-IDF weighting without stop word removal.
   • BRemove stop words and split on hyphens.
   • CApply stemming and remove punctuation.
   • DTokenize using whitespace and then apply lemmatization.
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A data science team is building a word analogy solver (e.g., 'king - man + woman = queen'). They have a large text corpus and want to use embeddings that are trained on global co-occurrence statistics. They also want the embeddings to be computationally efficient to train. Which method should they use?

   • AGloVe
   • BTF-IDF
   • COne-hot encoding
   • DWord2Vec (skip-gram)
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   What is the main advantage of Word2Vec embeddings over one-hot encoded vectors?

   • AWord2Vec assigns a unique index to each word
   • BWord2Vec does not require training on a corpus
   • CWord2Vec vectors are sparse and easier to store
   • DWord2Vec captures semantic relationships between words
   Reveal Solution Discussion Select an answer first
5. 9expert · hard

   A data science team is building a text classification model for a legal document review system. They have a large corpus of legal documents and need to represent each document as a feature vector. They want to capture the presence of specific legal terms (e.g., 'breach', 'indemnity') and also the semantic relationships between terms (e.g., 'breach' and 'violation' are similar). They have a limited budget for computational resources. Which approach should they use?

   • AUse averaged Word2Vec embeddings.
   • BUse TF-IDF vectorization.
   • CConcatenate TF-IDF vectors with averaged Word2Vec embeddings.
   • DUse one-hot encoding of all legal terms.
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   What is the key difference between GloVe and Word2Vec in how they learn word embeddings?

   • AGloVe uses global word co-occurrence statistics, while Word2Vec uses local context windows
   • BGloVe produces sparse vectors, while Word2Vec produces dense vectors
   • CGloVe requires labeled data, while Word2Vec is unsupervised
   • DGloVe uses a neural network, while Word2Vec uses matrix factorization
   Reveal Solution Discussion Select an answer first