Hybrid Search: Keywords Plus Vectors
Combine keyword and vector search so you find both exact codes and loosely worded questions.
A taste of a lesson
How does reciprocal rank fusion handle a document that only one system found?
It still gets a score, just from one list. With k equal to 60, a document ranked first by keyword search alone scores 1/61, about 0.0164. A document ranked fifth by both systems scores 2 times 1/65, about 0.0308, so it comes out ahead. That is the point: agreement between systems is rewarded, but a strong single system hit is not lost. Try it: a document is third in vector results and tenth in keyword results. What is its RRF score?
Written by the teacher as an example. In your lesson the tutor answers your own questions, and like any AI it can be wrong.
What you will be able to do
- Explain the opposite failure modes of keyword and vector search
- Merge rankings with reciprocal rank fusion and weighted score fusion
- Tune the keyword and vector balance for different query types
- Avoid fusion mistakes such as duplicates and mismatched score scales
- Measure hybrid search gains per query type
Lesson plan
- 1 Two ways to fail See where keyword and vector search each break, using real queries. Start
- 2 Reciprocal rank fusion Merge two ranked lists with RRF and understand why it is robust. Start
- 3 Weighted score fusion Combine normalised scores with a tunable weight. Start
- 4 Practical merging details Implement fusion without duplicates or missing candidates. Start
- 5 Query dependent balance Adjust the mix based on what kind of query arrives. Start
- 6 Measuring the gain Show where hybrid search helps with recall per query type. Start
Try asking
About this tutor
For developers whose vector search misses product codes, names and exact phrases, or whose keyword search misses questions phrased in different words. You learn why the two approaches fail in opposite ways, how to run both and merge their rankings with reciprocal rank fusion or weighted scores, how to tune the balance for different query types, and where learned sparse methods sit between the two. You set up a query test set split by type, so you can see exactly where hybrid search helps and where it does not, and you avoid common merging mistakes such as duplicate chunks and mismatched score scales.
Reviews
Students can review a tutor after a paid lesson. Nobody has yet.
About the teacher
Search engineer teaching embeddings, chunking, vector and keyword search, and reranking from first principles
9 tutors 374 lessons taught Sample
I come from search: indexes, ranking and the long tail of queries that make a search box look foolish. When retrieval augmented generation arrived, most of what mattered turned out to be old search problems in new clothes, so that is how I teach it. We start with how text becomes something you can compare, then how documents are split,...
See Emeka's profile and tutorsMore like this
Other tutors on the same or nearby topics.