query expansion

**Query expansion** is the **retrieval technique that augments the original query with related terms, synonyms, or concepts to improve recall** - expansion helps recover relevant documents that do not share exact wording. **What Is Query expansion?** - **Definition**: Automatic generation of additional query terms or phrases preserving original intent. - **Expansion Sources**: Thesaurus terms, embedding neighbors, knowledge graphs, or LLM-generated variants. - **Primary Goal**: Increase retrieval coverage for semantically related but lexically different content. - **Risk Tradeoff**: Excessive expansion can introduce topic drift and noise. **Why Query expansion Matters** - **Recall Boost**: Finds documents using alternate terminology and paraphrases. - **Domain Robustness**: Helps with acronym, jargon, and synonym variation. - **Long-Tail Support**: Improves retrieval on sparse or underspecified user phrasing. - **RAG Quality**: Better evidence coverage improves grounded answer completeness. - **Adaptive Search**: Useful when initial retrieval underperforms. **How It Is Used in Practice** - **Controlled Expansion**: Add limited high-confidence terms with weighting and filters. - **Hybrid Integration**: Use expansion differently for sparse and dense retrieval stages. - **Quality Monitoring**: Track recall gains versus precision loss to tune expansion aggressiveness. Query expansion is **a powerful recall-enhancement tool in retrieval systems** - when carefully constrained, expanded queries significantly improve evidence coverage without overwhelming ranking quality.