Hybrid search (semantic + keyword) in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why combine semantic and keyword search in hybrid search?

Hybrid search uses both semantic and keyword search methods. What is the main benefit of combining these two?

A It reduces the time needed to index documents by skipping keyword extraction.

B It improves search accuracy by capturing both exact matches and related meanings.

C It eliminates the need for any text preprocessing before searching.

D It guarantees that only exact keyword matches are returned.

💡 Hint

Think about how semantic search understands meaning and keyword search looks for exact words.

Semantic search finds related meanings, while keyword search finds exact words. Combining both helps find more relevant results.

❓ Predict Output

intermediate

Output of hybrid search scoring function

Given the following Python code snippet for a hybrid search score, what is the printed output?

Prompt Engineering / GenAI

semantic_score = 0.7
keyword_score = 0.5
alpha = 0.6
hybrid_score = alpha * semantic_score + (1 - alpha) * keyword_score
print(round(hybrid_score, 2))

A 0.65

B 0.58

C 0.60

D 0.62

💡 Hint

Calculate weighted average: multiply semantic score by alpha, keyword score by (1 - alpha), then add.

Hybrid score = 0.6*0.7 + 0.4*0.5 = 0.42 + 0.20 = 0.62

❓ Model Choice

advanced

Choosing a model for semantic search in hybrid search

You want to build a hybrid search system combining keyword and semantic search. Which model is best suited for the semantic search part?

A A pretrained transformer-based sentence embedding model like Sentence-BERT

B A simple TF-IDF vectorizer without embeddings

C A rule-based keyword matching system

D A clustering algorithm like K-Means

💡 Hint

Semantic search needs to understand meaning, not just word counts.

Transformer-based models like Sentence-BERT create embeddings that capture sentence meaning, ideal for semantic search.

❓ Hyperparameter

advanced

Effect of alpha in hybrid search scoring

In hybrid search, the combined score is calculated as:
hybrid_score = alpha * semantic_score + (1 - alpha) * keyword_score.
What happens if alpha is set to 1?

A Only semantic search scores are used, ignoring keyword matches.

B The scores are averaged equally between semantic and keyword search.

C Only keyword search scores are used, ignoring semantic matches.

D The hybrid score becomes zero regardless of input scores.

💡 Hint

Consider what multiplying by 1 or 0 does to each score.

When alpha=1, hybrid_score = 1*semantic_score + 0*keyword_score = semantic_score only.

❓ Metrics

expert

Evaluating hybrid search effectiveness

You run a hybrid search system and want to measure how well it balances semantic and keyword search results. Which metric best captures both relevance and ranking quality?

A Root Mean Squared Error (RMSE)

B Precision at K (P@K)

C Normalized Discounted Cumulative Gain (NDCG)

D Mean Reciprocal Rank (MRR)

💡 Hint

Look for a metric that considers position of relevant results and graded relevance.

NDCG measures ranking quality by considering relevance scores and their positions, ideal for hybrid search evaluation.