SAG / ARCHITECTURE NOTE
Query Rewriting and Decomposition: Breaking Complex Questions into Searchable Units
Explains the definition and need for query decomposition, how it works, criteria for applying it to the SAG architecture, and a practical checklist, with support from research and official documentation.
Definition in One Sentence
Query decomposition is the process of separating the subjects, conditions, and subproblems in a complex question and expanding them into multiple searches.
Key answer: “Is it secure, low-cost, and compatible with existing systems?” is difficult to answer with sufficient evidence from a single search. The evidence may differ for each condition.
Why Is This Technology Needed?
“Is it secure, low-cost, and compatible with existing systems?” is difficult to answer with sufficient evidence from a single search. The evidence may differ for each condition.
How It Works
Create subquestions while preserving entities and constraints, then combine the results under the original question. Check that the rewriting process has not changed the meaning.
When designing the system, do not consider accuracy alone. Latency, cost, data boundaries, update frequency, and failure behavior must also be defined to make results reproducible in operation. Values that automation cannot determine with confidence should be left as unmeasured or requiring review, rather than changed to zero or marked as successful; this is the safer approach.
Connection to SAG Technology
SAG’s Goal and question snapshot serve as reference points for preserving the original question and connecting subsearches and evidence under the same provenance.
Practical Checklist
- Preserve proper nouns and numerical constraints
- Indicate whether the evidence requirements for each subquestion are met
- Keep a record that makes it possible to audit differences in meaning before and after rewriting
- Distinguish the status of failures, empty results, and permission errors from success
- Revalidate before and after changes under the same conditions
Research and Official Documentation
The references support the principles and recommendations. They do not guarantee search visibility, AI mentions, rankings, or revenue. The effects of actual implementation must be verified through observations of service data under equivalent conditions.
Selection Criteria and a Practical Example
“Under 200,000 won, at least 10 hours of battery life, and suitable for video conferences” can be searched by splitting it into price, usage time, and microphone requirements. If the price cap is omitted during rewriting, the result does not answer the original question. The relationship between the original question’s constraints and those of its subquestions must be stored as well.
Scope of Application at SAG
This article covers research principles and extension design for search AI. Read it in connection with SAG’s page collection, evidence recording, and report verification structures, but do not interpret it to mean that all search algorithms from the papers are deployed in the production pipeline. Whether a given approach has been applied should be verified using the search module, evaluation data, and execution records.
Further Reading on This Technology
Compare RAG, GraphRAG, and Self-RAG papers and their conditions for application.
SAG / KNOWLEDGE LINKS