![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
juushikaEvery now and then Pandora goes through these cycles where it's like: "Hmm. It seems we've played most of the wonderful thumbed up and seeded music. I think it is time to—branch out! to try new things! to play some total crap!"
Oh, and it does. Like, for example, the last six songs. So that's when I refresh the player and it pauses, reloads, and goes: "Hey, look! Good music! I wanna play that." And it does.
Speaking (obliquely) of Pandora, you know what would be cool? A book genome project. Pandora uses the Music Genome Project, which functions something like this:
So now consider writers like Neil Gaiman. His style is also unique: he writes fantasy novels, yes, but he writes often sets his fantasy at the borders of reality, so that the two overlap; he interweaves fantasy, humor, and horror; his writing is heavily atmospheric. If you try to search for Gaiman-style books from his genre alone, you'll run into everything from mammoth high-fantasy series (Robert Jordan, J.R.R. Tolkein) to modern mainstream fantasy (Charles de Lint) to YA fantasy (Annette Curtis Klause, J.K. Rowling). Not all of the books and authors—indeed, very few of them—will actually resemble Gaiman in terms of style or content. It would be much easier to find more Gaiman-style books if you were searching by style and by content, but it's hard to Google "overlapping fantasy/reality" or "atmospheric fantasy". A book genome project, however, would be looking out for exactly those aspects: how the book is written, what the style is like, what the setting is like, what the content is like, as well as what the genre is. You would be able to enter an author, or a book, and it would spit back a list of similar books based on these various factors. And if one of them didn't work for you, you could give it a thumbs down, and the service would respond in turn, looking for less of those aspects/author and more of everything else. Ideal, really.
Amazon is currently playing with some things that are similar to the book genome project, but they tend to be clunky and unreliable, and there's no way to personalize the system. You can use item tags (the tags are user submitted); Gaiman's Neverwhere turns up results like urban fantasy, parallel worlds, and alternate London. Parallel worlds turns up everything from theoretical physics to Pier's Anthony mass fantasy to Terry Pratchett. You can further narrow the results, say by selecting fantasy ... and that turns up three results: Neverwhere, a Pratchett novel, and an out of print novel, Roadmarks by Roger Zelazny. Searching by broad tags turns up massive of results; narrowing those results to produce a manageable or more specific selection narrows the results to a bare handful. The ability to begin and narrow searches is limited by what tags have been user submitted, and the user isn't particularly trust worthy; without a formal tagging system, tags are arbitrary, repetitive, misspelled, and misapplied. This isn't an ideal system.
Devon—the boy—mentioned a book genome project when I was complaining about just this process the other day—and I really do think it would be brilliant. It would be much more intuitive and intelligent, and it would produce much more useful results. There are a lot of authors and books I've read lately that I'd love to find more books in the same vein of. So all we need is someone to make a huge list of lists of all possible categories, tags, and descriptions, and then thousands of readers to read, identify, and categorize millions of books. That's not too much to ask, is it?
But it would be lovely if it could be done.
Oh, and it does. Like, for example, the last six songs. So that's when I refresh the player and it pauses, reloads, and goes: "Hey, look! Good music! I wanna play that." And it does.
Speaking (obliquely) of Pandora, you know what would be cool? A book genome project. Pandora uses the Music Genome Project, which functions something like this:
Together we set out to capture the essence of music at the most fundamental level. We ended up assembling literally hundreds of musical attributes or "genes" into a very large Music Genome. Taken together these genes capture the unique and magical musical identity of a song - everything from melody, harmony and rhythm, to instrumentation, orchestration, arrangement, lyrics, and of course the rich world of singing and vocal harmony. It's not about what a band looks like, or what genre they supposedly belong to, or about who buys their records - it's about what each individual song sounds like.And it is surprisingly effective and in some cases ideal. One of my favorite bands is The Dresden Dolls, who describe their music as "Brechitan Punk Cabaret." Perhaps not surprisingly, that's not a genre like "R&B" or "alternative rock" that encompasses a number of bands with the same style and sounds. The sound of The Dresden Dolls is surprisingly unique, and it's difficult to find music that works well beside them. But because of the way that the music genome project functions, it's actually very easy to find more artists like The Dresden Dolls through Pandora. (Like Rasputina, like A Particularly Vicious Rumor, like Charlotte Martin).
So now consider writers like Neil Gaiman. His style is also unique: he writes fantasy novels, yes, but he writes often sets his fantasy at the borders of reality, so that the two overlap; he interweaves fantasy, humor, and horror; his writing is heavily atmospheric. If you try to search for Gaiman-style books from his genre alone, you'll run into everything from mammoth high-fantasy series (Robert Jordan, J.R.R. Tolkein) to modern mainstream fantasy (Charles de Lint) to YA fantasy (Annette Curtis Klause, J.K. Rowling). Not all of the books and authors—indeed, very few of them—will actually resemble Gaiman in terms of style or content. It would be much easier to find more Gaiman-style books if you were searching by style and by content, but it's hard to Google "overlapping fantasy/reality" or "atmospheric fantasy". A book genome project, however, would be looking out for exactly those aspects: how the book is written, what the style is like, what the setting is like, what the content is like, as well as what the genre is. You would be able to enter an author, or a book, and it would spit back a list of similar books based on these various factors. And if one of them didn't work for you, you could give it a thumbs down, and the service would respond in turn, looking for less of those aspects/author and more of everything else. Ideal, really.
Amazon is currently playing with some things that are similar to the book genome project, but they tend to be clunky and unreliable, and there's no way to personalize the system. You can use item tags (the tags are user submitted); Gaiman's Neverwhere turns up results like urban fantasy, parallel worlds, and alternate London. Parallel worlds turns up everything from theoretical physics to Pier's Anthony mass fantasy to Terry Pratchett. You can further narrow the results, say by selecting fantasy ... and that turns up three results: Neverwhere, a Pratchett novel, and an out of print novel, Roadmarks by Roger Zelazny. Searching by broad tags turns up massive of results; narrowing those results to produce a manageable or more specific selection narrows the results to a bare handful. The ability to begin and narrow searches is limited by what tags have been user submitted, and the user isn't particularly trust worthy; without a formal tagging system, tags are arbitrary, repetitive, misspelled, and misapplied. This isn't an ideal system.
Devon—the boy—mentioned a book genome project when I was complaining about just this process the other day—and I really do think it would be brilliant. It would be much more intuitive and intelligent, and it would produce much more useful results. There are a lot of authors and books I've read lately that I'd love to find more books in the same vein of. So all we need is someone to make a huge list of lists of all possible categories, tags, and descriptions, and then thousands of readers to read, identify, and categorize millions of books. That's not too much to ask, is it?
But it would be lovely if it could be done.
