Music, metadata, and monkeys

[ckd]

Now that every shipping iPod (save the shuffle) has a color screen and the ability to display lyrics saved within the song file, the metadata problem is even more acute.

Put simply, user-entered metadata sucks rocks. Cory Doctorow calls this metacrap, and notes that even when people have a financial incentive (eBay listings, for instance) they screw up (anyone want a "Plam Zire 71"?).

This seems even more true of CDDB. Sometimes I think they've got crack-addled monkeys entering data. Reversed artist and song title? Yeah, we've got that. Both of them in the song title field? Yup. Incorrect spelling and capitalization? Uh-huh.

Probably most annoying because of its semi-invisibility, though, is the misuse of the "compilation" flag. That's not there just because the iTunes developers wanted to give you a "ticky-box" option! I've now become inured to the habit of checking the "get info" box for the CD before importing, simply as a defensive measure against this being mis-set.

An album is not a compilation, unless it has songs on it by multiple artists. This means that, in general, "Greatest Hits" albums do not qualify. (Some of them may have different but related acts together, like Lionel Richie and the Commodores, or the various Crosby, Stills, Nash, and/or Young groupings.) Random everyday "one artist, bunch of songs" albums? Definitely not, you crack-addled monkey!

Cover art, at least, has Amazon and Wal-Mart (yeah, they're good for something) doing a reasonable job of offering it up. Still, there are bad scans, microscopic JPEGs, and scans with the clearly-readable "PROMOTIONAL USE ONLY" legend right across the middle to contend with.

Lyrics...don't get me started on lyrics. My iPod doesn't display them anyway. This is probably a good thing.

Comments

[alecaustin.livejournal.com]

Yeah, the misuse of the "compilation" check-box drives me absolutely batshit. While usually better than nothing, CDDB data is definitely a mixed blessing.

[whl.livejournal.com]

As I recall, CDDB started as a volunteer, open project, and got taken commercial later, and I don't think they've ever gone back and reviewed the data.

I have most of the CDs of Peter Schickele's P.D.Q. Bach works; all my CDs are ripped to my hard drive.

I got very tired of how many different ways there were to tag those works. On my hard drive, I dragged them all to the same folder. But the ID3 tags are still all over the map.

[kodi]

I've been mostly happy with the Musicbrainz data, although the tagger is Windows only, unfortunately. But it's very rare that I find something that dismays me the way some CDDB entries do. Musicbrainz does import data from freedb, but it takes a separate user-driven step, so kinks have a second chance to get ironed out.

[timprov.livejournal.com]

I've been pretty impressed with the low error rate of freedb, but then again I'm not using it on something like an iPod, where it has to be difficult to edit things yourself. Every once in a while there's a wholly bad disc, but I can almost always just recheck the db and get a close-to-perfect second opinion. I don't recall CDDB having that functionality.

[kip-w.livejournal.com]

I usually end up fixing the data in whatever CDDB titles I get. Yeah, they think the composer is the artist, if they pay any attention to composers at all. Most annoying was the number of times each cut consisted of a long long album title, a long artist/composer name, and the first two letters of the name of the cut. How utterly useless.

Consequently, everything gets renamed by me so that the composer comes first, with an abbreviated but informative title of the music (which includes what movement it is), and the performer at the end. Sometimes I have to compromise and squeeze and leave out stuff to get it to fit.

Then, when I back it to another medium, I get the message that the title was too long, so it either asks me to shorten it myself, or it does it for me and I get pieces with names like JSBA~1.mp3 -- everything I entered down the drain.

On the plus side, at least I have DVD backups of all my mp3s now. That means the gods will have to smite me twice to take it all away from me. They'll do it, of course, but I like that they have to work harder for it.

[beamjockey.livejournal.com]

Amy Proni had a similar reaction (http://viaproni.typepad.com/viaproni/2005/09/folksonomies.html) to Librarything. (http://www.librarything.com/about.php) She contrasted it with the rigid, carefully constructed taxonomies that librarians use in classifying books for their catalogues:

They talk about using folksonomies instead of taxonomies; and allude to putting power in the hands of people. Well, that's fine and good, and it's true that LC is slow to respond to changes in language / style, etc (read anything by Sanford Berman to get a better idea on that) but really... I don't see how this can be a truly useful database for anyone farther down the road, because it's going to be a mess without standardization.

Argh. I recognize that the people who create stuff like this honestly mean well, but I think they are working at cross purposes to what librarianship has been about for the past hundred years.

[ckd]

At one point I thought about whether an MLS (or MLIS) was a logical direction for me. As it stands, library science is yet another area that I have an interested layman's understanding of, but no real depth. I'm sufficiently library-geeky to prefer Library of Congress call numbers to Dewey.