Follow Slashdot stories on Twitter


Forgot your password?

Music Recommendation Engines Compared 126

An anonymous reader writes "The music recommendation/music discovery space seems to be heating up this year. Two big recent features on music recommendation engines: ExtremeTech has a round-up and reviews of eight leading services. Of the eight, emerges as the winner: " is by far the best out there, possessing a huge library of music, a great community, and a recommendation feature that will blow you away." Meanwhile, just ran an in-depth feature about the hows and whys of music recommendation software, that tells the story going back to the '90s, and interviews people at, Pandora, MusicIP, and the startup Echo Nest: '"Our hope is to answer every possible question about music that ever existed. If we can pull that off, then I think we're doing very well," says [Brian] Whitman.'"
This discussion has been archived. No new comments can be posted.

Music Recommendation Engines Compared

Comments Filter:
  • I agree - Pandora's library doesn't seem to be as big as's, but the ease of use is second to none. I've found myself several times being captivated by music I never knew existed.
  • by stupidfoo ( 836212 ) on Thursday June 01, 2006 @03:17PM (#15447723)
    What I've found with Pandora is that it seems to work best if you just tell Pandora what music you like and not what you dislike. I just seem to get better results that way.
  • by stuffduff ( 681819 ) on Thursday June 01, 2006 @03:17PM (#15447731) Journal
    Pandora can get stuck in a rut when it doesn't understand why you listen to what you're listening to. I'm a big fan of Progressive Rock. Say you start with PFM's (Premiata Forneria Marconi) 'Celebration.' After a few songs it tends to pick some kind of Indy Rock. Progressive Rock may be grandiose, but that's what I like about it, that it continues to change during a piece of music. In essence it, by it's very nature seems to escape the definitions given to it. So I'm not sure that any of these engines will ever get that kind of music right. If I "Don't like' enough songs in a row it will go back for some classic prog, but if left alone it just keeps wandering away from it instead of leading me to new prog like Wobbler or Kenso.
  • by 'aspies' are retards ( 958036 ) on Thursday June 01, 2006 @03:25PM (#15447796)
    Ok, so I've had it with the musicians who have sold their souls to the corporations. With the advert of the Internet, they don't need anyone else to publish and distribute their music to the world. So now I want to get my music from independent artists. The problem is: I know what kind of music I like, and I know which mainstream bands make this kind of music, but I don't have time to go listening to every indie artist to find out what they make.

    What I'm looking for is a site where I can enter or select names of bands or songs that I like, and get independent music recommended to me. You like Alanis Morisette? Try Jen Pitch. That sort of thing. Does anybody know of such sites?

    By the way: the example above is just an association I know from the top of my head; I'm not very much into the kind of music at all.

  • Clinko (Score:3, Interesting)

    by clinko ( 232501 ) on Thursday June 01, 2006 @03:40PM (#15447930) Journal
    I've been using Clinko for years. []

    In fact, I wrote it :)
  • Biz Model..? (Score:3, Interesting)

    by suv4x4 ( 956391 ) on Thursday June 01, 2006 @03:53PM (#15448074)
    Can someone explain to me wha is the business model of those services?

    From what I see I download a player where I can play commercial music of the sort I like for free, with CD quality and no ads...

    There are Google Ads on the site, but I can just not go on the site and play free music forever... The player doesn't seem to contain ad/spy ware.

    Where's the "catch" :)?
  • by gunpowda ( 825571 ) on Thursday June 01, 2006 @03:59PM (#15448133)
    However, what I've found to be excellent about Pandora, apart from its uncanny ability to play relevant music based on the few clues I've given it, is that the service tells you *why* it's chosen a particular track if you click that button. It's a great insight into some features of the music you may not have consciously picked up on before.

    It is true that if your taste is for a niche genre then it won't be too useful, but if you're in that position then you probably know better than any software what you ought to listen to next!

  • vs Pandora (Score:4, Interesting)

    by Supersonic1425 ( 903823 ) on Thursday June 01, 2006 @04:12PM (#15448247)

    I'll start by saying that I'm a huge fan of, and have been for years. I'm addicted to the place, and my music collection would be nothing without it. While does have a feature where artists are automatically recommended to you, I rarely use it. It's the social aspect of that sets it apart. The best way of getting recommendations is just simply asking for them [].

    I've used Pandora a few times before, and was always disappointed with what it recomended. Results are mixed to say the least—it clearly works better for some types of music than others—and some of the recommendations can be, quite frankly, laughable.

  • Re:iTunes playlists (Score:3, Interesting)

    by adpowers ( 153922 ) on Thursday June 01, 2006 @04:13PM (#15448259)
    Musicmobs [] does something similar. I'm not sure about now, but before you were able to export your whole iTunes library (or just one playlist) to XML and upload it to Musicmobs to get recommendations.

    I think they got rid of the direct upload feature because it was extremely slow (the iTunes XML file is huge, but can be highly compressed) and they now have a client - Mobster [] - which will upload your stats. I'm not sure if Mobster allows you to upload just one playlist in place of your library. You can upload a playlist for all to see, but it doesn't (yet) give you recommendations on the playlist. It might be worth checking out, though.

  • by llyenn ( 873664 ) on Thursday June 01, 2006 @04:38PM (#15448448)
    Well...a kickass podcast that a lot of people don't know about is 75 Minutes []. Great Indie music that runs the gamut from Jazz to Experimental... May not be exactly what you are looking for, but its a great start...
  • by klausboop ( 322537 ) on Thursday June 01, 2006 @05:02PM (#15448682)
    I am a huge fan of the MusicIP MusicMagicMixer application discussed in the pitchforkmedia article. MusicIP does a fantastic job of helping me navigate my own collection since I ripped it into a couple hundred gigs of files. Coupled with SlimServer, I feel like I have the best of everything: Offline, I use MusicIP to create mixes from my own collection and transfer them to my portable player or make a CD. Online, I can stream my own collection with SlimServer playing MusicIP mixes, and when I want to discover new stuff I drop over to Pandora or Last.FM. I was excited to read about something that would have the intelligence to group Heartbreaker with Living Loving Maid, though, which MusicMagicMixer cannot do. That's righteous. But I'd want to be able to turn it off...sometimes it's fun to have Heartbreaker cut into something else. Imagine it being TOO intelligent. It's one thing to have Overture always lead into Temples of's another to have it always play both discs from The Wall any time it picks Another Brick In The Wall part 2. I agree with the extremetech article about Launchcast, too. Before Pandora and Last.FM, it is where I went online to discover new music. While I still like their ratings system the best of all and feel it is the most intuitive (rate the song and/or the artist and/or the album to shape your station), I think the other services implement moods much better and generally have a more positive user experience.
  • (Score:2, Interesting)

    by gnurb ( 632580 ) on Thursday June 01, 2006 @08:19PM (#15449979) Homepage
    Yeah, I wanted this too, so me and a buddy created [] (free download for mac and pc)
  • is on the right track, but incomplete. Registered users submit the tracks they play, and the algorithm considers how many plays each artist got. But it doesn't look at a per-song level, just artists. There's no way to tell it which songs you dislike.

    This is bad because many of us have bought an album and realized we only liked a few tracks. Yet the big fans of that artist like all the songs, or different ones. Jamiroquai - Virtual Insanity got lots of airplay, but the rest of the album is much slower and disco-y. Consequently, is highly unlikely to recommend the artist and of course that song to listeners who missed it six years ago. thinks I should like lots of Radiohead, Coldplay, and The White Stripes because other users who listen to the same artists I do have also listened to those bands a lot. Well I only like a few songs from the first two and really dislike the last band. Too much whining in the vocals. If only let me tell it the songs I like and the ones I don't. Then it could find users who also dislike the same music as I. Consequently, it would recommend just songs I'm probably going to like; certain Ska songs by Reel Big Fish and others, certain Rock/Swing by The Cherry Poppin' Daddies and The Brian Setzer Orchestra.

    Then I don't have to skip through albums getting annoyed with how much of them I don't like because I'm not a huge Ska or Swing fan.

    When I listen to Best of albums by Garth Brooks and Clint Black, along with select Shania Twain, and the Black Dog soundtrack, I should get song recommendations for Travis Tritt that only include the few tracks I'll probably like.

    If could increase their computing power per user by about 30x, I think it could be recommending all kinds of obscure hits and tracks that users would never think of otherwise and human community members couldn't think of either. After all, I like a bunch of hip-hop and techno too. In fact I have extremely varied musical interests, but probably most people do and they're too stuck in a few genres because there's too much chaff among the wheat to branch out and find the select songs they'll enjoy.

No problem is so large it can't be fit in somewhere.