Combines keyword precision with vector recall.
Hybrid retrieval uses a combination of retrieval methods – typically dense vector search and traditional keyword (sparse) search – to find relevant documents. Instead of relying on just one approach, it runs both methods and then merges the results, leveraging the strengths of each system.
Hybrid retrieval aims to get the best of both worlds. Sparse (term-based) search excels at exact matches – if the user's query uses the same terminology as a document, a keyword search will precisely find that document. Dense search, on the other hand, excels at finding conceptually related info (synonyms, paraphrases) but might also return something contextually similar yet not actually relevant. By combining them, the pipeline can catch relevant documents that one method alone might miss.
hybridRetrieval.searchThroughSampleCollection
hybridRetrieval.aboutDescription
Building a robust Hybrid Retrieval solution is challenging. Respeak's Enterprise RAG Platform handles this complexity for you.