Plan and manage an Azure AI solution 1.1 Choose the appropriate Foundry services for generative AI and agents

A parts catalog index in Azure AI Search contains text fields and embeddings.

Study guide skill: Choose an appropriate method for retrieval and indexing

Question B8-002

A parts catalog index in Azure AI Search contains text fields and embeddings. Users search with exact part numbers as well as natural-language descriptions, and the app should get one ranked result list. Which two statements about the right retrieval approach are correct? (Choose two.)

Select 2 answers.

  • A Send separate keyword and vector requests and merge the results in the app
  • B Hybrid search averages the BM25 and vector similarity scores of each document
  • C Use a hybrid query: one request that includes both the search and vectorQueries parameters ✓ Correct answer
  • D Use vector search only, because it matches exact part numbers more precisely than keyword search
  • E Keep the full-text (keyword) part of the query, because scenarios such as product codes perform better with keyword search, which can identify exact matches ✓ Correct answer

Answer: C, E

Hybrid search is a single query request that includes both the search and vectorQueries parameters, so the app gets one ranked list without merging anything itself. Keep the keyword part: scenarios such as product codes, jargon, dates, and names perform better with keyword search because it identifies exact matches, while vector search finds conceptually similar content even without keyword matches, so vector-only search is the weaker choice for part numbers. The service merges the full-text and vector results with Reciprocal Rank Fusion (RRF), not by averaging BM25 and similarity scores.

Deep dive

How it works

Hybrid search in Azure AI Search is a single query request that includes both the search (full-text) and vectorQueries parameters. The service merges them with Reciprocal Rank Fusion (RRF), so the response is one result set ranked by RRF. The app has nothing to merge.

Why keep the keyword half for a parts catalog? Microsoft calls it out: scenarios such as product codes, highly specialized jargon, dates and people's names perform better with keyword search, because it can identify exact matches. Vector search's strength is the other half of this workload: it finds content that's conceptually similar to the query even when no keywords match, which handles the natural-language descriptions.

The other options, and when you'd use them

  • Separate requests merged in the app: Unnecessary. Hybrid search is one request, and RRF already returns one unified result set.
  • Average BM25 and similarity scores: That's not how the merge works. Results from each query are merged by Reciprocal Rank Fusion (RRF).
  • Vector search only: Exact part numbers are the case where keyword search does better. Vector search finds conceptually similar content, which is the wrong strength for exact codes.

Exam tip

Codes, IDs, names, dates = keep keyword. Meaning = vectors. Both in one request = hybrid, merged with RRF.

References

Last reviewed:

Ready to test yourself? The quiz asks this and every other AI-103 question, with instant feedback. Practice this in a quiz →

All Plan and manage questions · AI-103 study guide

Thanks for your feedback!