BM25.
also: TF-IDF · IDF · inverse document frequency · term frequency · relevance score · relevance scoringThe standard relevance score: rewards terms frequent in a document (with diminishing returns) and rare across the corpus, normalising for length.
the two-minute refresher
- Term frequency: more occurrences = more relevant, but 10 isn't 10× better than 1 (saturation, k1).
- IDF: rare words matter more than common ones.
- Length normalisation (b): long documents don't win just by containing more words.
- The default scorer in Lucene and Elasticsearch.