Combining Vector and Full-Text Search with Reciprocal Rank Fusion
Every time I build retrieval for a RAG system, I run into the same wall. Vector search is wonderful at understanding meaning. Ask it "what is our policy on working from home" and it happily finds the paragraph titled "Remote Work Guidelines" even though the words do not match. But ask it for "ERR_4021" or "Policy 7.3" and it flounders, because an error code has no meaningful embedding…
Combining vector and full-text search with reciprocal rank fusion enables RAG systems to effectively handle both meaningful embeddings and exact token matches. The method works by ignoring the raw scores from each search and focusing solely on the rank of each document within its respective list. The fused score for a document is calculated as the sum of 1 / (k + rank) for each list it appears in, where k is a damping term that defaults to 60.
This formula ensures that the top result from any single list does not dominate the overall score, while still rewarding agreement between the two searches. The technique can be implemented using just a few lines of code, making it versatile for various rankers, including lexical indexes, dense retrievers, and rerankers.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
