Category:
Category:
Hybrid Retrieval (Vector + Keyword)
Category:
Retrieval & RAG
Definition
Combining semantic search and keyword search for optimal retrieval accuracy.
Explanation
Vector search retrieves by meaning, but may miss exact phrases, numeric values, or acronyms. Keyword search retrieves exact matches but lacks semantic understanding. Hybrid retrieval merges both approaches for maximum recall and precision. It is essential for enterprise use cases involving structured and unstructured data.
Technical Architecture
Embedding Search → Keyword Search → Merge → Rerank → LLM
Core Component
Vector DB, BM25/lexical search, reranker, fusion algorithms
Use Cases
Enterprise search, financial analysis, compliance QA, RAG systems
Pitfalls
Latency increases; poor weighting between vector and keyword signals
LLM Keywords
Hybrid Retrieval, Vector + Keyword Search
Related Concepts
Related Frameworks
• Reranking
• Embeddings
• Retrieval Pipelines
• Hybrid Retrieval Architecture