Design and Implement Intelligent Search DP-800 Practice Questions (Page 9)

Questions 41–45

1. 41application · medium

   A legal research platform stores case documents in Azure SQL Database. Users search for 'precedent on data breach liability' and expect results that match both the exact phrase and conceptually similar cases. The solution must combine keyword matching with semantic relevance, and the development team wants a single query that returns a unified ranked result set. What should you implement?

   • AUse a single VECTOR_SEARCH query with a text-embedding model that also encodes lexical features, eliminating the need for full-text search.
   • BUse CONTAINS for full-text search and UNION the results with a VECTOR_SEARCH query, then manually merge scores in the application layer.
   • CRun a full-text search with FREETEXT and a vector search separately, then present the results in two separate tabs in the application UI.
   • DCreate a full-text index and a vector index, then use a hybrid query with the RRF function to combine the relevance scores from both search methods.
   Reveal Solution Discussion Select an answer first
2. 42application · medium

   An e-commerce site searches product descriptions in Azure SQL Database. Users often search for brand names (exact match) and product categories (semantic match). The team wants a single search that handles both. Which approach should they use?

   • AUse only vector search with embeddings that include brand names in the text.
   • BRun full-text and vector search separately and show results in two tabs.
   • CUse only full-text search with a thesaurus that maps categories to keywords.
   • DRun a full-text search for brand names and a vector search for categories, then combine with RRF.
   Reveal Solution Discussion Select an answer first
4. 43application · medium

   A developer is implementing vector search in a SQL database. They have a table with a VECTOR column and need to return the 3 most similar items to a given vector. They also need to ensure the query uses the ANN index. Which query should they use?

   • ASELECT TOP 3 * FROM Items WHERE VECTOR_DISTANCE(embedding, @input) < 0.5
   • BSELECT * FROM VECTOR_SEARCH(Items, embedding, @input, 3)
   • CSELECT TOP 3 * FROM Items ORDER BY VECTOR_DISTANCE(embedding, @input) ASC
   • DSELECT TOP 3 * FROM Items ORDER BY VECTOR_NORMALIZE(embedding) ASC
   Reveal Solution Discussion Select an answer first
5. 44expert · hard

   A healthcare analytics company stores millions of clinical notes as vectors in Azure SQL Database. A compliance audit requires that every search returns the mathematically true nearest neighbors with no approximation. The search workload is low-concurrency, and the team can tolerate slower query times. Which vector index configuration should you choose?

   • ACreate a brute-force index that scans all vectors and computes exact distances for every query.
   • BCreate a HNSW index with a high ef_search value to get results close to exact but not guaranteed.
   • CCreate a HNSW index and run a post-query verification pass in the application to confirm results are exact.
   • DCreate an IVFFlat index with a large number of lists to reduce approximation error.
   Reveal Solution Discussion Select an answer first
6. 45expert · hard

   A medical research database stores 2 million patient records with embeddings. Researchers require exact nearest neighbor results for a study, but the current ANN index returns results in 300 ms. They need to switch to ENN. What is the most significant impact they should expect?

   • ARecall will decrease because ENN is less accurate than ANN.
   • BQuery latency will increase significantly because ENN requires scanning all vectors.
   • CQuery latency will decrease because ENN uses a more efficient index.
   • DStorage requirements will increase because ENN requires additional indexes.
   Reveal Solution Discussion Select an answer first