You are here

Log in or register to post comments
struts
struts's picture
Offline
Last seen: 1 month 2 weeks ago
Joined: Feb 1 2007 - 12:02pm
Tagging!

Dearest audiophile siblings,

It is finally time to post on what I personally find one of the biggest bugbears of computer audio, managing metadata or 'tagging'. I have seen many posts along the way lamenting the difficulty of tagging audio files and the inordinate amount of time it takes, however very little discussion of or pointers to any articles with proposed solutions or recommendations. As my library grows it grows messier and I am now pretty much at the point of no return. If I can't find a better way of doing it now the amount of legacy data will exceed the time available for me to ever get it under control, at least that is until somebody invents a system or tool waaaay more powerful than the ones I am currently aware of. All-in-all it's enough to drive one back to analog!

So what am I talking about? Well, I see two main types of problem:
- Data structure problems, whereby the fields available to store metadata are inappropriate or their purpose is ambiguous and
- Data problems, where the intention is probably clear but the exact usage is not

Many of the structural problems I have found relate principally to classical, but affect most other genres to some degree or other. Here are some of my favourites:

  • Artist. Who or what is the Artist? In the rock world the answer is pretty straightforward, it is the band or singer performing the song, e.g. 'The Beatles'. However consider the case of a classical violin concerto. Is the Artist the soloist, the conductor, the orchestra or all three? It is even worse with opera where you have a number of principal soloists, a choir (or two), an orchestra and a conductor. In many cases (and in my experience both Gracenote and AMG are equally guilty) I have found the name of the composer in the Artist field, e.g. Album: Fidelio, Artist: Ludwig van Beethoven!? Doesn't anybody moderate this stuff?
  • Album. Easy enough in the rock world, 'Revolver', but what about classical? Here Album is (mostly) minimally defined by the combination of Composer(s), the Work(s), the Performer(s) and the Conductor(s). But most of these data are captured in other fields, so does one simply duplicate them in the Album field? Then consider the case of the 'Gouldberg' Variations of which I have at least ten different recordings of about seven or eight different performances. Better add Year and possibly also Venue to the list!
  • Track. Again, simple in the rock world, 'Here, There and Everywhere' but switch to classical and things get complicated. 'Aria' is fine, but on its own it doesn't tell you much, I mean there are two Arias in the Goldberg Variations alone, not to mention all the other thousands of Arias in other works. The question here is how much of the information from Album to duplicate in each Track. The problem is that the display fields for track information on most players are quite limited in size so the actual track-specific information quickly gets pushed out of view.
  • Genre. The twin problems here are that depending on the level of granularity one goes to genres are not mutually exclusive and that much music crosses several genres anyway. It is equally irritating to find Norah Jones in Folk, Vocal and Blues as it is to have to decide when Coldplay made the transition from Alternative to Rock. Is Weihnachtsoratorium Classical, Choral or Seasonal? And how on earth does one classify Myriam Alter??

Data problems on the other hand are less confined to classical works, although classical is once again hardest hit. Apart from the usual issues of typos, spelling errors (especially in names), and errors related to non-English characters here are some of the main ones:

  • Names. There are just so many ways of representing them and, seemingly no standards or conventions. Mozart? W.A. Mozart or Mozart, W.A.? Wolfgang Amadeus Mozart? Wolfgang Amadeus Mozart (1756-1791)? etc. Okay, it is no big deal to have to look for Beethoven under 'L' for Ludwig (and Ludvig of course, groan) but it is surely irritating when you finally find F
Editor
Editor's picture
Offline
Last seen: 3 years 11 months ago
Joined: Sep 1 2005 - 8:56am
Re: Tagging!


Quote:
I am interested to hear how others here have tackled these issues and particularly if anyone is aware of any standards, adopted or proposed, de facto or de jure or even simply 'best practices' in this area.

Great post, Struts. I am afraid I don't have anything concrete to offer. I tend to leave both my iPod and my SqueezeCenter on "Shuffle," which kinda makes the metadata question moot.

John Atkinson
Editor, Stereophile

Axon
Axon's picture
Offline
Last seen: Never ago
Joined: Sep 2 2005 - 1:44pm
Re: Tagging!

A lot of the problem with dealing with composers vs conductors vs performers is that it's just a really difficult problem in general to structure that kind of data. And most of the time, I don't really give a rat's ass who conducted the music. I have very few duplicates of pieces anyway. So in that sense, setting ARTIST=COMPOSER is a very legitimate and rational decision to make, even if I know it's wrong. It's also much easier to browse classical music that way on my iPod. Half the classical performers I've tagged, I couldn't tell you what exactly they performed...

Usually, the best solution for this sort of thing is a dumb text search: Make sure the information is tagged somewhere, and that the music player can do a substring search on it when you search for a particular work/composer/performer/etc. You'll run into collisions, like if you're looking for a Bernstein-composed piece but you wind up with a lot of Bernstein-conducted pieces, but that is of a relatively minor concern.

I'm not sure about other music players, but I use foobar2000 with the Facets plugin, and it is absolutely glorious with that sort of thing. Music is grouped by artist/album, unless I've set the WORK and COMPOSER and PERFORMER tags on it, in which case it groups by that instead. Hierarchial browsing can be based off of any arbitrary tag, or any programmatic combination of tags. That said, it's not exactly a turnkey solution.

Example (from some vinyl I ripped recently):

Code:Artist Name : Chopin, Fr

struts
struts's picture
Offline
Last seen: 1 month 2 weeks ago
Joined: Feb 1 2007 - 12:02pm
Re: Tagging!

My problem is I have a fairly large library (not much of it ripped yet, partly due to this issue) with quite a lot of duplication.

Your example (super disc by the way, I am a huge fan of Pollini's

jazzfan
jazzfan's picture
Offline
Last seen: 5 months 4 days ago
Joined: Sep 8 2005 - 8:55am
Re: Tagging!

Great post Struts and as usual you expressed these very trying issues in a clear and concise manner. The current metadata struture is very pop/rock centric and leaves much to be desired when working with claasical music. Jazz falls somewhere in the middle.

A related issue is the use of metadata to establish relationships within one's music library, i.e. while listening to a John Coltrane Quartet track to be able to find other tracks/albums that Elvin Jones plays on. AFAIK none of currently the established metadata/tagging systems can do even this simple task. Sure one can go online and find all the recordings that feature Elvin Jones but there is no way of searching one's own music library to find out this information.

I know that these complaints seem like very minor issues to most people but i feel that one of the main features of a computer based music system is being able to harness the power of a computer to browse one's library.

Axon
Axon's picture
Offline
Last seen: Never ago
Joined: Sep 2 2005 - 1:44pm
Re: Tagging!


Quote:
My problem is I have a fairly large library (not much of it ripped yet, partly due to this issue) with quite a lot of duplication.

Your example (super disc by the way, I am a huge fan of Pollini's

Elk
Elk's picture
Offline
Last seen: 6 months 1 week ago
Joined: Dec 26 2006 - 6:32am
Re: Tagging!

Wonderful post, Strut!

I feel your pain. Tagging is one of the main reasons I cannot get myself to fall in love with music servers.

I have also defaulted to Artist = composer. I then add a little info to the title of the album to indicate what orchestra/conductor/performer (there are limitations on what portable players and other servers have as available fields).

As I file my CDs by composer my method makes some sense to me. It still is often easier just to go to the physical CD to find what I want however.

CharlyD
CharlyD's picture
Offline
Last seen: 3 years 3 months ago
Joined: Jul 20 2006 - 4:01pm
Re: Tagging!

The feature I found most appealing about the new players from Olive was the design of their user interface. The schema for the music database includes a very wide range of categories including several targeted to classical, opera and jazz. You can define, for example, a comma-separated list of sidemen for a jazz entry or the act number for an operatic piece. Rather than displaying these fields in a flat table, these categories are organized within hierarchies defined by the genre. Selection of the genre in the UI defines which categories are displayed.

I haven't heard anything about the Opus No. 4 or the Melody No. 2 since their introduction last April. Has anybody out there had a chance to hear one of these or play with its UI? If their hardware products can't make it in the marketplace, it would be a shame to see the software die also.

struts
struts's picture
Offline
Last seen: 1 month 2 weeks ago
Joined: Feb 1 2007 - 12:02pm
Re: Tagging!


Quote:
Wonderful post, Strut!


Steady on "Moosie"!

Struts:

Strut:

  • X
    Enter your Stereophile.com username.
    Enter the password that accompanies your username.
    Loading