Comments on: SIGIR 2010: Day 1 Technical Sessions http://thenoisychannel.com/2010/07/21/sigir-2010-day-1-technical-sessions/ Mon, 21 May 2012 05:21:00 +0000 hourly 1 http://wordpress.org/?v=3.3.2 By: Daniel Tunkelang http://thenoisychannel.com/2010/07/21/sigir-2010-day-1-technical-sessions/comment-page-1/#comment-6523 Daniel Tunkelang Sun, 25 Jul 2010 13:25:02 +0000 http://thenoisychannel.com/?p=3204#comment-6523 Dinesh, content-based recommendations of movies based on similarity to a single movie are probably intuitive enough for users not to require a reductionist explanation of the similarity factors. That's much less true for recommendations based on a set or sequence of movies, especially when the recommendations aren't just content-based (e.g., collaborative filtering). Andy, you can read the full paper <a href="http://portal.acm.org/citation.cfm?id=1835449.1835484" rel="nofollow">here</a> if you have access to the ACM Digital Library. I wish authors would be more be more consistent about posting their papers on their own web sites! Dinesh, content-based recommendations of movies based on similarity to a single movie are probably intuitive enough for users not to require a reductionist explanation of the similarity factors. That’s much less true for recommendations based on a set or sequence of movies, especially when the recommendations aren’t just content-based (e.g., collaborative filtering).

Andy, you can read the full paper here if you have access to the ACM Digital Library. I wish authors would be more be more consistent about posting their papers on their own web sites!

]]> By: Andy http://thenoisychannel.com/2010/07/21/sigir-2010-day-1-technical-sessions/comment-page-1/#comment-6511 Andy Thu, 22 Jul 2010 13:58:57 +0000 http://thenoisychannel.com/?p=3204#comment-6511 How did the IBM researchers combine and rank the results from the two recommender models (tag-based and people-based)? How did the IBM researchers combine and rank the results from the two recommender models (tag-based and people-based)?

]]> By: Dinesh Vadhia http://thenoisychannel.com/2010/07/21/sigir-2010-day-1-technical-sessions/comment-page-1/#comment-6509 Dinesh Vadhia Thu, 22 Jul 2010 09:10:53 +0000 http://thenoisychannel.com/?p=3204#comment-6509 @ daniel, re: Recommendation engines cry out for transparency Re-read the original post on this and still not sure that I understand where you're coming from. For example, if the last movie I watched (with the kids!) was "lilo and stitch" and we enter it into a movie recommendation engine I'd expect to find similar movies to that one. Are you saying that this is not what today's recommendation engines do? @ daniel, re: Recommendation engines cry out for transparency

Re-read the original post on this and still not sure that I understand where you’re coming from. For example, if the last movie I watched (with the kids!) was “lilo and stitch” and we enter it into a movie recommendation engine I’d expect to find similar movies to that one.

Are you saying that this is not what today’s recommendation engines do?

]]>