Retrieval misses exact product codes, while meaning-based queries work well. What helps?
Why D is the answer
Hybrid search that adds keyword matching
Keyword search catches exact terms that semantic search can miss.
In detail
- Keyword search finds exact codes.
- Hybrid keeps the meaning-based matches.
- Hybrid search combines both methods.
Use case
An illustrative scenario, not a real incident.
Users search for "XR-200" and get unrelated products. Adding keyword matching finds exact codes as well.
See it in different ways
mindmap
root((RAG))
Ingest
Sync data source
Detect personal data
Chunk
By structure
Hierarchical
Search
Vector
Keyword
Rank
Rerankerflowchart TD
Q{"What does the query need?"}
Q -->|"Meaning"| A0["Semantic search"]
Q -->|"Exact code"| A1["Keyword search"]
Q -->|"Both"| A2["Hybrid search"]Key point: Exact terms need keyword search.
Why the other options are wrong
- A. Longer chunks, so that each product code has more text around it
- Longer chunks dilute the match.
- B. A larger embedding dimension, with no change to the query approach
- Dimensions do not fix exact matching.
- C. A higher temperature on the model that writes the final answer
- Temperature has no effect on retrieval.