Reciprocal Rank Fusion - NISHIO Hirokazu's Scrapbox (Auto-translated from Japanese)

Reciprocal Rank Fusion

Reciprocal rank fusion | Elasticsearch Guide 8.13 | Elastic

$ score_i = {1\over a_i} + {1\over b_i}

Same as harmonic mean for large and small relationships

$ k may be added

https://gyazo.com/7fd3a17c5cc500f57a9a380e421765a1

Reciprocal Rank Fusion outperforms Condorcet and individual Rank Learning Methods(PDF)

Here k=60

Increasing k reduces the difference between first and second place.

---

This page is auto-translated from /nishio/Reciprocal Rank Fusion using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.