Design and Implement Models and Embeddings DP-800 Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   Which embedding maintenance method uses an Azure service that can react to database changes and call a function to regenerate embeddings?

   • AMicrosoft Foundry
   • BChange Tracking
   • CAzure Functions with SQL trigger binding
   • DTable triggers
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   When designing embeddings for a product catalog, which column is most important to include to capture the product's semantic meaning?

   • ACreation date
   • BProduct ID
   • CInventory count
   • DProduct description
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A healthcare analytics company is building a multilingual patient-support chatbot. The database stores patient education documents in English, Spanish, and French. The team needs to generate embeddings for semantic search. They are evaluating two external embedding models: Model A supports 50 languages and outputs 384-dimensional vectors; Model B supports 10 languages and outputs 1536-dimensional vectors. The team's primary requirement is accurate cross-lingual retrieval, and they have ample storage and compute budget. Which model should they select?

   • AModel A, because smaller vectors reduce storage costs, which is the primary constraint.
   • BModel A, because broader language coverage improves retrieval for all three languages.
   • CModel B, because higher-dimensional embeddings always produce more accurate semantic search.
   • DModel B, because it is more efficient for structured output generation.
   Reveal Solution Discussion Select an answer first
5. 9expert · hard

   A company uses an external embedding model for a production search feature. The model vendor announces a new version that improves accuracy but changes the vector dimensionality from 768 to 1024. The team must update without downtime. They have a large corpus of existing embeddings. Which approach best balances accuracy improvement and operational risk?

   • ARegenerate embeddings directly in production without testing, to minimize effort.
   • BRegenerate all embeddings in a staging environment, validate accuracy, then switch the search index to the new embeddings in a single cutover.
   • CKeep the old model and only use the new model for new documents.
   • DGradually replace old embeddings with new ones as documents are accessed, keeping both in the index.
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A customer support portal stores tickets with the following columns: TicketID, Subject, Description, Status, Priority, CreatedDate, and ResolvedDate. The team wants to generate embeddings to enable semantic search over past tickets. Which columns should be included in the embedding text?

   • AAll columns, including TicketID and dates
   • BSubject, Description, Status, and Priority
   • CStatus and Priority only
   • DSubject and Description only
   Reveal Solution Discussion Select an answer first