Query Reformulation (Query Rewriting)
Definition
Query Reformulation, also called Query Rewriting, is the process of transforming an original query into a different query representation intended to improve retrieval or better express the underlying information need.
Why it matters
Users do not always phrase queries in the form best suited to an index or retriever. Reformulation can correct ambiguity, add missing context, change wording or restructure a request before retrieval.
How it works
Input interpretation
- The system begins with the original query and its interpreted intent, entities and constraints.
Rewrite generation
- A rewritten query is produced that preserves the core information need while changing its wording or structure.
- The rewrite may be lexical, semantic, structured or model-generated.
Retrieval execution
- The rewritten query is sent to one or more retrieval systems.
- Some systems may use both the original and rewritten forms.
How Netsleek uses the term
Netsleek distinguishes Query Reformulation from Query Expansion and Query Fan-Out. Reformulation changes a query representation, expansion adds related terms or concepts, and fan-out creates multiple retrieval queries or pathways.
Comparisons
- Query Reformulation vs Query Expansion: Reformulation rewrites the query. Expansion broadens it with additional related terms or concepts.
- Query Reformulation vs Query Fan-Out: Reformulation may produce one revised query. Fan-out produces multiple queries or retrieval pathways.
- Query Reformulation vs Prompt: The prompt is the original input to the AI system. A reformulated query is a retrieval-facing transformation of that information need.
How Query Reformulation Connects Across AI Systems
Query Reformulation (Query Rewriting) does not operate in isolation. It connects to other established concepts across the AI search and retrieval process:
- Depends on understanding: Query Understanding can identify what the rewrite must preserve.
- Can precede expansion: A rewritten query may then be expanded or decomposed.
- Can improve lexical matching: Rewriting can introduce terminology better aligned with an index or source corpus.
- Can affect retrieval outcomes: Different rewrites can surface different candidates and sources.
Related glossary concepts
- Query Understanding
- Query Expansion
- Query Fan-Out
- Multi-Query Decomposition
- Lexical Retrieval
- Semantic Retrieval
- Prompt
Common misinterpretations
- A query rewrite is not automatically visible to the user.
- A rewrite should not be inferred unless the system exposes it or the methodology explicitly simulates it.
- Reformulation is not the same as generating unrelated alternative queries.
Summary
Query Reformulation transforms a query into a different retrieval representation while preserving the underlying information need. It is an established retrieval mechanism distinct from expansion, decomposition and fan-out.