Comments on: More on Pandora: genres, genomes, and musical taste… http://www.scholarslab.org/digital-humanities/more-on-pandora-genres/ Works in Progress Thu, 02 Feb 2012 16:20:57 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Roscenco http://www.scholarslab.org/digital-humanities/more-on-pandora-genres/comment-page-1/#comment-958 Roscenco Sun, 23 May 2010 06:22:41 +0000 http://scholarslab.lib.virginia.edu/?p=76#comment-958 In a spirit somewhat contrary to the position you take in this article, what I'd really love to at least play with on Pandora, if not use outright, is the ability to specify those traits, and weigh them, from the back end, so to speak, and see what kind of music is "created", rather than the other way around. I have yet to see any clue to doing this - it seems like a fun exercise, and would seem easily adaptable to the current engine. In a spirit somewhat contrary to the position you take in this article, what I’d really love to at least play with on Pandora, if not use outright, is the ability to specify those traits, and weigh them, from the back end, so to speak, and see what kind of music is “created”, rather than the other way around. I have yet to see any clue to doing this – it seems like a fun exercise, and would seem easily adaptable to the current engine.

]]> By: Phoebe http://www.scholarslab.org/digital-humanities/more-on-pandora-genres/comment-page-1/#comment-529 Phoebe Tue, 16 Feb 2010 09:39:40 +0000 http://scholarslab.lib.virginia.edu/?p=76#comment-529 Although Pandora overcomes some limitations of collaborative filtering (like tending to recommend already-popular music), I think you're right to point to the challenges for any recommendation system in creating diversity, serendipity - and simply in accurately predicting people's complex tastes. You might be interested to know that the type of approach Pandora has taken to music is also being applied to other areas. For example, Jinni's http://www.jinni.com search-and-recommendation engine for movies and TV shows is powered by our "Movie Genome," though our approach is based on semantic technology for analyzing reviews and synopses rather than human tagging. Although Pandora overcomes some limitations of collaborative filtering (like tending to recommend already-popular music), I think you’re right to point to the challenges for any recommendation system in creating diversity, serendipity – and simply in accurately predicting people’s complex tastes.

You might be interested to know that the type of approach Pandora has taken to music is also being applied to other areas. For example, Jinni’s http://www.jinni.com search-and-recommendation engine for movies and TV shows is powered by our “Movie Genome,” though our approach is based on semantic technology for analyzing reviews and synopses rather than human tagging.

]]>