BM25 in PostgreSQL

VectorChord-BM25 is a new PostgreSQL extension offering advanced BM25 scoring, optimized indexing, and tokenization for faster, more relevant text search results. This open-source tool outperforms ElasticSearch, delivering queries thrice as quickly. The article presents BM25's fundamentals, PostgreSQL's current text search capabilities, and the gains of using VectorChord-BM25.

• BM25 ranking is 3x faster than ElasticSearch.
• Includes optimized indexing and enhanced tokenization.
• Implemented as a native PostgreSQL extension.
• Higher Top1000 QPS than ElasticSearch.
• Comparable NDCG@10 ranking accuracy.