BM25

I’ve been using Lunr.js to implement browser-side freetext search for Paradicms statically-generated sites. I was curious how Lunr scores documents, and read that it uses the BM25 algorithm developed by Stephen E. Robertson, Karen Spärck Jones and others in the 1970s and 1980s.

I never met Professor Spärck Jones, but after she passed away in 2007 I realized I’d seen her in the halls of the Computer Laboratory several times. She had an impressive career, and had also been married to Roger Needham, a former head of the department and the co-author of one of my favorite systems papers, “On the duality of operating system structures”.