Develop AI Solutions by Using Azure Database for PostgreSQL AI-200 Practice Questions (Page 8)

Questions 36–40

1. 36foundation · easy

   What is the primary purpose of semantic retrieval in an AI application?

   • ATo retrieve documents by their exact ID
   • BTo encrypt documents at rest
   • CTo find documents based on meaning rather than exact keyword matches
   • DTo compress documents for storage
   Reveal Solution Discussion Select an answer first
2. 37expert · hard

   A search application stores embeddings in Azure Database for PostgreSQL with pgvector. The table has 50 million rows, and the team needs to perform exact nearest-neighbor searches for a small subset of high-priority queries, while using approximate search for the rest. Which strategy should they implement?

   • ACreate a separate table for the high-priority embeddings and use exact search only on that table.
   • BCreate an HNSW index for approximate searches and use a sequential scan for exact searches on the small subset.
   • CCreate an IVFFlat index and use it for both exact and approximate searches.
   • DUse a single HNSW index for all searches and accept approximate results for the high-priority queries.
   Reveal Solution Discussion Select an answer first
4. 38expert · hard

   A database administrator for an online marketplace is tuning a PostgreSQL database. A query that filters on a low-cardinality status column and a high-cardinality created_at column is running slowly. The administrator creates a composite index on (status, created_at), but the query planner still chooses a sequential scan. What is the most likely reason, and what should the administrator do?

   • AThe index is corrupted; the administrator should drop and recreate the index.
   • BThe planner estimates that the sequential scan is cheaper because the status filter is not selective enough; the administrator should analyze the table to update statistics and consider a partial index if the status values are skewed.
   • CThe composite index is not being used because the columns are in the wrong order; the administrator should create the index on (created_at, status) instead.
   • DThe query uses a function on the created_at column, preventing index usage; the administrator should rewrite the query to avoid the function.
   Reveal Solution Discussion Select an answer first
5. 39application · medium

   A legal research platform uses Azure Database for PostgreSQL with pgvector to perform semantic retrieval of case documents. Users enter natural-language queries, and the system must return relevant documents even when the query does not contain exact keywords. Which approach should the platform use?

   • AUse a LIKE pattern with wildcards to match partial keywords.
   • BUse a full-text search query with tsvector and tsquery on the document text.
   • CConvert the user query into an embedding using a language model and search for the nearest embeddings in the database.
   • DUse a trigram similarity index on the document text.
   Reveal Solution Discussion Select an answer first
6. 40foundation · easy

   In a retrieval-augmented generation (RAG) pattern, what is the purpose of using metadata filters?

   • ATo narrow the vector search to a relevant subset of documents
   • BTo encrypt the documents
   • CTo store the generated response
   • DTo generate embeddings for the query
   Reveal Solution Discussion Select an answer first