← Back to Glossary
Dense Retrieval
Retrieving documents based on semantic meaning using vector embeddings rather than matching exact words.
Why it exists
Catches synonyms and conceptual matches (e.g. matching 'doctor' with 'physician') even when no exact words overlap.
How it works
Query and documents are passed through embedding models; vector distance search returns closest concept matches.
Related terms
Where you'll use this
roadmap