Get the latest tech news

Hybrid Search in CrateDB - ranking and scoring calculations in pure SQL


Learn how to enhance search relevancy with Hybrid Search in CrateDB, combining BM25 and Vector Search for better accuracy and results.

CrateDB is a great database for creating products with this technique due to its strong resiliency, fast queries thanks to its unique index-all, SQL language, and its rich and complete data types. BM25 or full-text search is a bag-of-words algorithm that ranks an unordered set of documents based on the keyword of a query. There are machine learning models that can transform data like text or images into dense vectors (list of non-zero floating point numbers).The closer two vectors are in space, the more similar they are, we can use this property tocalculate how similar a query is to our data.

Get the Android app

Or read this on Hacker News

Read more on:

Photo of SQL

SQL

Photo of calculations

calculations

Photo of scoring

scoring

Related news:

News photo

Show HN: Briefer – Multiplayer notebooks with schedules, SQL, and built-in LLMs

News photo

Redis Alternative at Apache Software Foundation Now Supports RediSearch and SQL

News photo

Show HN: zfind – CLI like find but using SQL syntax and also does tar/zip