Query Expansion with Synonyms
What you'll learn: How to broaden your RAG search by automatically generating alternative phrasings and synonyms to capture more relevant documents.
The Core Problem
When users search with phrases like "fix a bug," your vector database might miss relevant documents that use "resolve a defect" or "troubleshoot an issue." Even with semantic embeddings, exact phrasing variations can reduce retrieval quality. Query expansion solves this by transforming a single query into multiple semantically equivalent versions before searching.
How It Works
Think of query expansion like casting a wider net. Instead of searching with just the original query, you:
- Generate alternatives: Use an LLM or synonym dictionary to create variations of the user's query
- Execute multiple searches: Run each expanded query against your vector database
- Merge results: Combine and deduplicate the retrieved chunks, often using score fusion techniques you learned earlier
For example, the query "automobile maintenance" expands to:
- "car maintenance"
- "vehicle servicing"
- "automotive care"
Each variation retrieves slightly different documents, increasing the chance you'll find all relevant context.
When to Apply It
Query expansion works best when:
- Users phrase questions in narrow, specific ways
- Your knowledge base uses varied terminology
- Initial retrieval returns too few results
- Domain-specific jargon has multiple common expressions
It pairs naturally with the hybrid search and score fusion techniques you've already learned, since you're now managing multiple result sets.
Key Takeaway: Query expansion with synonyms broadens search coverage by automatically generating semantically equivalent query variations, helping your RAG system capture relevant documents even when terminology varies across your knowledge base.