Design and Implement Intelligent Search DP-800 Practice Questions (Page 6)

Questions 26–30

1. 26application · medium

   A company wants to build a recommendation engine that finds similar products based on user reviews. They plan to use semantic vector search. What is the first step they should take?

   • AUse VECTOR_SEARCH directly on the raw review text.
   • BDefine a VECTOR column and immediately create an ANN index.
   • CGenerate embeddings for each review using a pre-trained language model.
   • DCreate a full-text index on the review column.
   Reveal Solution Discussion Select an answer first
2. 27application · medium

   A medical research database stores 100,000 gene expression profiles as vectors. Researchers need exact nearest neighbors for a small set of critical queries, but the overall system must also support fast approximate search for exploratory queries. What is the best way to support both requirements?

   • AUse only an HNSW index and set ef_search to maximum for all queries.
   • BUse only an IVFFlat index with a very large number of lists.
   • CUse only a brute-force index and accept slower performance for all queries.
   • DCreate two vector indexes on the same column: one brute-force for exact queries and one HNSW for approximate queries.
   Reveal Solution Discussion Select an answer first
4. 28expert · hard

   A social media platform stores user-generated content embeddings in Azure SQL Database. The dataset has 50 million vectors, and the search must return results in under 100 milliseconds. The team is deciding between HNSW and IVFFlat indexes. They need high recall but can accept some approximation. Which index should they choose and why?

   • AHNSW, because it provides better recall-quality trade-offs and faster query times at high scale compared to IVFFlat.
   • BIVFFlat, because it requires less memory and is always faster than HNSW for large datasets.
   • CIVFFlat, because it does not require any training or build time.
   • DHNSW, because it guarantees exact nearest neighbors for all queries.
   Reveal Solution Discussion Select an answer first
5. 29expert · hard

   A team has implemented a hybrid search system. They are evaluating performance and notice that the vector search component is the bottleneck, taking 2 seconds per query. The full-text component takes 50 ms. They need to reduce overall latency to under 500 ms. What should they do?

   • ASwitch the vector search from ANN to ENN to improve speed.
   • BRemove the full-text component to reduce total latency.
   • CReduce the number of vectors returned by the vector search component.
   • DUse an ANN index with a lower recall setting to speed up the vector search.
   Reveal Solution Discussion Select an answer first
6. 30foundation · easy

   Which technique is commonly used to combine the results of full-text and vector search into a single ranked list?

   • AFull outer join
   • BReciprocal Rank Fusion (RRF)
   • CUnion all
   • DPivot
   Reveal Solution Discussion Select an answer first