Talk:Ranking (information retrieval)

Latest comment: 12 years ago by Elvenhack in topic This is not about ranking functions

Other Meaning(s)

edit

Ranking functions also are a tool for proving termination of programs. Here the ranking function maps configurations to a well-founded ordering in such a way that any step of the program will make the value decrease. See Colón & Sipma, Proc. TACAS 2001. PhS (talk) 12:26, 14 July 2010 (UTC)Reply

This is not about ranking functions

edit

This page isn't about ranking functions at all; it's about weighting schemes. Cosine similarity is a ranking function. Qwertyus (talk) 09:33, 12 September 2012 (UTC)Reply

Exactly what made you think that? Article is exteremely light on material, but I think it's in the right general direction. I don't get your "at all" remark. I've always thought of cosine similarity, as more of a distance metric between documents, better suited for things like clustering. Just summing up tf-idfs scores for each query term is as good (or, rather, as bad) a scoring functing as cosine similarity between tf-idf vectors of query and document IMHO. -- X7q (talk) 22:54, 12 September 2012 (UTC)Reply
This is definitely about information retrieval ranking functions (aka relevance scoring algorithms.) See Manning's text in Information Retrieval "Scoring, term weighting and the vector space model" Elvenhack (talk) 13:46, 19 September 2012 (UTC)Reply