Shypd.aiThe largest open index of AI toolsBM25S is an Open Source tool that provides a fast and memory-efficient implementation of BM25 lexical search algorithms in Python. It is built on Numpy and Scipy, offering speeds comparable to or exceeding ElasticSearch.
Trending
About
BM25S (or BM25-Sparse) is an open-source Python library offering fast and efficient implementations of BM25 algorithms, built on top of Numpy and Scipy. It is designed for low-dependency and low-memory usage, allowing everything to happen within Python without the need for external servers like ElasticSearch. Thanks to a novel sparse eager computation strategy, bm25s achieves high retrieval speeds, often outperforming ElasticSearch in benchmarks. It includes a tokenizer and supports various BM25 variants like Robertson, ATIRE, BM25L, BM25+, and Lucene. The tool also features memory-mapping for optimized RAM usage, making it suitable for large datasets. While not a full replacement for ElasticSearch, bm25s is ideal for prototyping and implementing BM25-based ranking algorithms where precise control over parameters and tokenization is needed.
Capabilities
Pricing & Plans
Open Source
Open Source
FAQs
Trending
