Knovvu AI Agent uses Hybrid Search to deliver the most accurate and relevant answers to your customers. This advanced retrieval method combines the power of semantic search with precise keyword matching to provide superior performance across different query types and content structures. Whether your users ask conceptual questions or search for specific technical terms, Knovvu AI Agent consistently finds and delivers the most appropriate information from your Knowledge Base.
How it works
Hybrid Search operates by running two complementary retrieval methods in parallel and intelligently combining their results to maximize accuracy and relevance.
Semantic Search uses dense vector embeddings to understand the meaning and intent behind queries. This approach excels at capturing conceptual similarity, making it particularly effective when users phrase questions in different ways or use synonyms. For example, a query about "car maintenance" can successfully retrieve documents discussing "vehicle upkeep" or "automobile service" because the system understands these concepts are semantically related.
Keyword Search leverages sparse vector representations to perform precise lexical matching. This method is invaluable for queries containing specific identifiers, technical terminology, product codes, or acronyms that must be matched exactly. When a user searches for "T-123456" or "TLS 1.3", keyword search ensures these exact terms are found.
Hybrid Search combines both methods through a sophisticated algorithm that fuses them. When a query is submitted, the system simultaneously retrieves candidate documents from both the semantic and keyword search paths. Each method scores its results based on relevance, and these scores are then merged using a ranking algorithm that considers both semantic similarity and lexical precision. This ensures that the final results benefit from the strengths of both approaches.
Hybrid Search Architecture
