Skip to content
Cosdata Documentation

Search Relevance

Cosdata provides advanced search relevance capabilities to ensure your users find exactly what they’re looking for.

Enhance search precision and recall with Cosdata’s hybrid search, leveraging the power of combined dense, sparse, and full-text (BM25) searches to deliver highly relevant, context-rich results for complex queries.

How It Works

Hybrid search combines multiple search techniques:

  1. Dense Vector Search: Uses embeddings to capture semantic meaning
  2. Sparse Vector Search: Captures keyword importance similar to traditional search
  3. Full-Text Search (BM25): Provides fast, scalable keyword and phrase queries with industry-leading performance
  4. Combined Scoring: Intelligently merges results for optimal relevance and ranking

Benefits

  • Enhanced contextual understanding
  • Superior relevance for both semantic and keyword queries
  • Lower latency and higher throughput than traditional search engines
  • Flexible scoring for a wide range of use cases