Glossary
A plain-language definition from the Curriculo hiring glossary.
Hybrid search combines exact keyword matching with semantic search, so it finds text with the same words and text with the same meaning.
Why it matters for hiring
Candidates describe the same skill in different ways. One writes PostgreSQL, another writes Postgres, and a third writes that she ran the company's relational database. Keyword search alone finds the first. Semantic search alone may miss exact terms like a license number. Hybrid search finds all three and keeps exact matches near the top.
How it works
- Keyword search scores documents by shared terms, often with a method called BM25.
- Semantic search turns text into vectors, lists of numbers that capture meaning, and finds documents with nearby vectors.
- The two result lists are merged into one ranking.
A common merging formula
Reciprocal rank fusion gives each document a score of 1 / (k + rank) in each list, then adds the scores. The constant k is often set to 60.
Example: a resume ranks 2nd in keyword results and 5th in semantic results. Its score is 1 / 62 + 1 / 65, about 0.0315. A resume that ranks 1st in keyword results but doesn't appear in semantic results scores 1 / 61, about 0.0164. The first resume ranks higher because both methods agree on it.
Common mistakes
- Assuming semantic search understands every niche term without testing.
- Dropping keyword search entirely and losing exact matches like license numbers.
- Searching on job titles only instead of the work people describe.
- Never checking results against a few known good candidates.
See how AI candidate search finds people, compare it with keyword matching, and learn how resume parsing prepares the data.