tag:blogger.com,1999:blog-6569681.post1399960430146499809..comments2024-03-24T10:38:16.997-07:00Comments on Geeking with Greg: Blending the link and query-click graphsGreg Lindenhttp://www.blogger.com/profile/09216403000599463072noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-6569681.post-76615143870161022092009-01-07T02:18:00.000-08:002009-01-07T02:18:00.000-08:00The paper mentions in the conclusions that click-b...The paper mentions in the conclusions that click-based ranking would be biased and should be used with other ranking methods because searchers would obviously click on the top results, which would lead to self-enforced ranking.<BR/><BR/>But I think the top results are self-enforcing (to a lesser extent though) even in link-based models and others. There are a large number of bloggers who, if they want to link to some source, would typically do a search for their keywords and link to the top results.<BR/><BR/>This happens so much when people write those (still) famous '10 ways to do blah-blah-blah'. And these same posts then get promoted on social media, twitter, etc. So the combined effect would still be biased even if you use social media popularity, bookmarks, links, etc. for ranking purposes. <BR/><BR/>It would be interesting if there can be a research on measuring the average boost (per niche & ranking position) in additional links, click-throughs, bookmarking, etc. after a result comes to the first page, and then discounting that boost to find new results. <BR/><BR/>As a sidenote, I think algos can also use (if not using already) the lack of boost in links, bookmarks for first page results to demote them. Like if result #2 & #3 are linked, bookmarked, twittered more than #1 in a given time, that could mean #1 should be demoted.Anonymousnoreply@blogger.com