Out of curiosity, I seeded it with Tom Lehrer. It didn't do nearly as well, repeating Tom Lehrer within 5 songs and hitting Randy Newman twice within 10 songs. The others on the playlist (e.g. Laura Nyro, Bruce Springsteen, and Yoko Ono) didn't seem quite apt. As with Alexandria book recommendations, this seems as likely to be driven by what a lot of participants have commented on as by any real similarities in style.
That said, I can see how, when it works, it could work exceptionally well.
I actually think it uses more analysis than participant feedback. If you look at some of the display screens, it will tell you why it picked a song, based on various musical traits.
I actually was basing my conclusion on their description of the Music Genome Project. They actually have participants in that dissecting all those various musical traits for a whole host of different aspects of the work. But it's still a matter of which songs received more attention from those participants, and which received less. Songs which receive less attention in the MGP are going to have a less well-defined set of descriptors, and so I'd expect them to produce a poorer playlist.
I put in "Cocteau Twins". Persephone was the sample song (which is one of my favorites), and it was followed by a song from Elastica, then one from Nena. I thought this was really interesting, esp. since I actually own both other albums and really like them. I never would have associated them with Cocteau Twins, though. It's nice that it picks out the musical qualities, because the music it's playing for me (and we're 3-for-3 on songs I'm enjoying so far) would typically come from all sorts of different genres.
I keep having mixed results. I was hoping that seeding with two different bands (e.g. Leonard Cohen and "Break My Stride" by Matthew Wilder) would get me music that was a mix between the two; instead, it alternates between quiet folk and rather loud 80s, which is not at all what I really wanted.
I think I just need to keep poking at it. It did turn up Nickel Creek when seeded with Girlyman, which is absolutely correct, so I've still got hopes.
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
no subject
Date: 2005-12-19 11:36 pm (UTC)Wow. That - that is just amazing.
It is SO COOL to be living in the damn future! Woohoo!
no subject
Date: 2005-12-20 02:37 am (UTC)no subject
Date: 2005-12-20 03:54 am (UTC)no subject
Date: 2005-12-20 12:05 am (UTC)That said, I can see how, when it works, it could work exceptionally well.
no subject
Date: 2005-12-20 12:56 am (UTC)no subject
Date: 2005-12-20 02:45 am (UTC)no subject
Date: 2005-12-20 12:10 pm (UTC)no subject
Date: 2005-12-20 02:52 am (UTC)So cool.
-Zanne
no subject
Date: 2005-12-20 12:11 pm (UTC)no subject
Date: 2005-12-20 06:56 pm (UTC)I think I just need to keep poking at it. It did turn up Nickel Creek when seeded with Girlyman, which is absolutely correct, so I've still got hopes.