Page 1 of 9

The End of "Genres"

Posted: Fri Dec 18, 2009 5:18 pm
by Davydov
For some time now we've been discussing how to replace the present inefficient systems of "genres" used on IMSLP to classify works. After some useful discussions earlier in the year, I'd now like to present what I really believe is a solution that should please (nearly?) everybody. See here for an outline of the system, and please post any comments below in this thread. Thanks.

Re: The End of "Genres"

Posted: Fri Dec 18, 2009 7:11 pm
by varnis
The only problem I see with this solution is the problem of standardising tags, so that they can be parsed for sorting. Other than that, i think it covers just about anything, unless people want to start arguing over whether X's Requiem is really for a funeral etc.

Overall, heartily support this, obviously well thought out system. :mrgreen:

Re: The End of "Genres"

Posted: Fri Dec 18, 2009 7:52 pm
by steltz
I think this might just solve everything, but I do want to warn everyone that people will still type things in that will have to be corrected. Then again, no system will prevent this.

For instance, if someone types in "2 horns and piano", one of us who regularly cleans things up will have to put the "2 horns" in the instrumentation field and leave the category (or tag or genre or whatever it is called) as "horns and piano" so the search works properly.

I think we should all take the weekend to think through various scenarios and see where it goes, but this looks very positive!!!!

And, of course, the fact that the instrument lists would no longer have to be manually maintained would be a HUGE benefit . . . . :D (Can't think when they were last updated.)

Re: The End of "Genres"

Posted: Fri Dec 18, 2009 9:31 pm
by Operalala
Using the first example:
The work page Drang in die Ferne, D.770 (Op.71) (Schubert, Franz) would have the tag "Lieder for Solo Voice and Piano"
This would create an infinite number of "tags", especially if uploaders are expected to tag German Art Songs "Lieder for Solo Voice and Piano", are French Art Songs to be "Melodies for Solo Voice and Piano" (with a diacritic which I do not have on my keyboard). Are we also going to acquire special terms for Russian, Czech, Spanish etc Art Songs?
Going further - Why on earth would an English song be tagged a Chanson? Why would a work title "3 Trios" be tagged a trio, when the "Genre" field lists it as a Choral work? It's blocked, so I suppose you're basing your tag on the title (which may well refer to a 3-part chorus), and not the work itself, and how would you know the accompaniment is an orchestra.

"Work Type" is the same as "Genre", and thus not the "end of genres", not that I have any particular stake in what term to use.

Of course, it would be really nice if the instrumentation, which is already given in the "Instrumentation" field, could be cataloged in some manner, and for vocal works, the language as well.

Re: The End of "Genres"

Posted: Fri Dec 18, 2009 10:37 pm
by Davydov
Operalala wrote:This would create an infinite number of "tags", especially if uploaders are expected to tag German Art Songs "Lieder for Solo Voice and Piano", are French Art Songs to be "Melodies for Solo Voice and Piano"
Well, no, it would create precisely two tags:
  • Lieder for Solo Voice and Piano (in the categories of "Lieder" and "Compositions for Solo Voice and Piano")
  • Mélodies for Solo Voice and Piano (in the categories of "Mélodies" and "Compositions for Solo Voice and Piano")
The categories of "Lieder" and "Melodies" would both themselves fall into your preferred category of "Art Song", and all this would happen automatically just by adding one tag per work page. And to repeat: once a tag his been created for the first time, the categories are allocated automatically for any pieces that are later given the same tag.

As far as other languages go, the Russians have Romances (Romansy), which would be tagged "Romances for Solo Voice and Piano" in the same fashion. Comparable terms in other languages will be treated similarly, but I think we'd be hard pressed to come up with more a dozen at most. And while it would be nice to be able to search on instrumentation and language, that's not likely to be feasible for quite some time.

In answer to your other question. my examples were based on the instrumentation and descriptions shown on a random selection of work pages. As you know, there are trios (quartets, duets, etc.) for solo voices and for choral ensembles, which anyone will be able to search for directly for the first time under the new system.

Re: The End of "Genres"

Posted: Fri Dec 18, 2009 10:56 pm
by Operalala
Would it not be a better idea to look into getting the Instrumentation and Languages cataloged, than having users slog through perhaps hundreds? of categories for every permutation of Work Type, Instrumentation, and Language in the database?

Re: The End of "Genres"

Posted: Sat Dec 19, 2009 5:27 am
by steltz
You're missing the point here. Just because there are several search options does not mean that every user has to search by every one of them. It means that any one of a number of words will work. For the less common languages, the "Compositions for Solo Voice and Piano" will still work just fine, but if you are looking specifically for Lieder, you can find them under "Lieder". (Under the other system, people will have to read the forums to know that the word has been deleted.)

Because a tag can belong to a higher category tag, you can use the specific or the general and either will work.

Far from being slog, it's a really good idea that will make the site more searchable, especially by people who don't read these forums. The opposite, having only one word that will work as a search word will reduce the ease of searching.
Davydov wrote:As you know, there are trios (quartets, duets, etc.) for solo voices and for choral ensembles, which anyone will be able to search for directly for the first time under the new system."
As someone who runs a chamber music class, and is constantly searching for vocal chamber music, I can't tell you how nice it will be to search, for example, specifically for vocal duets or trios. I won't be able to do this under the other proposed system.

Re: The End of "Genres"

Posted: Sat Dec 19, 2009 6:56 am
by Operalala
Steltz, the point I am making is that with hundreds of category permutations, so much will be "mis-cataloged" that your custom-made search terms will not bring up what you are looking for, if you yourself did not catalog them, because no one will actually go through all those categories for each and every piece they upload, and get the exact vocabulary, format, spelling, conjugation etc to make it work. Like "Fantasia" ?? no uploader is going to arrive at that, even from the title of the work "Phantasie". Terms like "Fantasia" and "Lieder" may suit a personal logic, but what other uploaders will be privvy to that.

Re: The End of "Genres"

Posted: Sat Dec 19, 2009 8:56 am
by Davydov
Operalala wrote:Steltz, the point I am making is that with hundreds of category permutations, so much will be "mis-cataloged" that your custom-made search terms will not bring up what you are looking for, if you yourself did not catalog them, because no one will actually go through all those categories for each and every piece they upload, and get the exact vocabulary, format, spelling, conjugation etc to make it work. Like "Fantasia" ?? no uploader is going to arrive at that, even from the title of the work "Phantasie". Terms like "Fantasia" and "Lieder" may suit a personal logic, but what other uploaders will be privvy to that.
There will be two aspects to the implementation:
1. A designated project team will go through all the existing works on IMSLP (numbering 20,285 at this moment) and give each of them a tag, following a set procedure to ensure the correct construction of the tags.
2. The project team will also monitor all new work pages created from that point, and apply the correct tags.

So we won't be expecting uploaders to do the tagging, even though the construction is essentially very simple (= type of work + instrumentation).

Re: The End of "Genres"

Posted: Sat Dec 19, 2009 10:33 am
by steltz
Operalala, people don't get things exactly right as it stands now. I regularly edit pages to clarify and correct things (and I have a cleanup star to prove it!).

As Davydov says, there will be a set procedure, so most things will be correct. I don't believe I'll be doing any more cleanup than I do now.

And this solves many problems, so it is definitely something we should implement.

Re: The End of "Genres"

Posted: Sat Dec 19, 2009 5:16 pm
by Operalala
A "designated Project Team" at a volunteer Wiki. That pretty ambitious. Do you have a project team? Will it be around, and tagging hundreds of tags following your "set procedure", forever in the future? And are your "types of works" really "very simple"?

Re: The End of "Genres"

Posted: Sun Dec 20, 2009 12:16 am
by Lyle Neff
Just to add a little confusion: Besides "romansy" (romances), Russians also have "pesni" (songs) :mrgreen:

Re: The End of "Genres"

Posted: Sun Dec 20, 2009 5:32 am
by steltz
Operalala wrote:Do you have a project team?
There have been project teams for previous projects, probably the largest (huge, actually) being the Orchestral Parts project. That project has run extremely efficiently and as a result, IMSLP now offers full parts to many, many orchestral works, with more coming almost daily.

Re: The End of "Genres"

Posted: Sun Dec 20, 2009 4:19 pm
by imslp
This is actually quite surprising. I was thinking the exact same thing (tags) a few months ago, right after Davydov abandoned his prior attempt. However, I obviously never got to the level of specificity that Davydov has, and I think his ideas have fixed several questions I had earlier. Nevertheless, I still have a few questions that remain.

What I like about tags:

1. They are the "norm" on the internet nowadays, and it actually works very reasonably well, and have stood the test of time.
2. They are flexible and extensible, generally lightweight, and easy for users to understand and use.

What I don't like about tags and that needs to be corrected:

1. Non-standardization: Obviously the tagging system won't be very useful if every spelling (and capitalization) variant of "Sonata for Violin and Piano" is used in the system. This can be corrected by having category merger functions, where admins can merge large chunks of tags. This can also be prevented to an extent by providing the user with an existing list of tags that we highly recommend the user use (how exactly the software handles this, however, is a huge issue, since the tag list is likely to be hugely long). However, what I can't figure out is how the "tag" categories will be nested into the larger categories. Will the users do it? Admins? A new group of "librarians"? We certainly can't have larger (non-tag) categories randomly created, or else we'd be back at the same problem with having too many tags.
2. My ignorance: I do not actually know how the tagging system works (I've never used tags) on other sites. I will need help from the other contributors here to get a general sense of how tagging systems work on different sites. For example, do they provide a list of prior tags to users? Do they allow users to create tags at a whim? What other features do they have?

Summary:

I like the idea, but some details of implementation is still unclear. And I think it is these details that are the hardest to sort out. This is not to discourage: I do entirely support this idea (in fact, I wanted to do it before), but the details need to be worked on. Most specifically, there are two questions:

1. How do we tie "tag" categories to the larger, more static categories?
2. How do we let the users choose previously-used tags? They are going to be very numerous, and a dropdown box is probably not going to work. Here it may be very useful knowing how other sites deal with this.

Re: The End of "Genres"

Posted: Sun Dec 20, 2009 5:44 pm
by Davydov
Hi Feldmahler. Thanks for your response, and I was intending to contact you tonight anyway, but I'll do my best to answer your questions now.

On your first point, technically the tags are just Wiki categories, and will work in the same way as the nationality categories do now, so that adding "Tag=Octets for wind instruments" to the right section of a workpage will add "[[Category:Octets for wind instruments]] to that page. The automatic list of wanted categories will alert the project team to any that are new, or just mistyped :)

All the tagging will be done by the aforesaid project group (in a similar way to the copyright reviewing), and we won't expect users to try to do this themselves. The biggest task will be to go through all the 20,000+ works already on the site, but that's always something we knew would happen, whatever new system was brought in. This will be done by going through the current genre categories individually, adding the tags and at the same time deleting the old genre/meta-genre references from the page. When the old genre/meta-genre categories are empty, they can be deleted for good.

Once the project starts we also need to make sure that all newly-created pages are correctly tagged as well, which should be a much less arduous task by comparison. Ultimately -- when all the existing pages have been tagged -- the only work involved will be the monitoring of new submissions.

As for the way that the tags group together to form broader categories, I looked at the example of Verdi's Requiem in another thread. Like all tags, the basic structure is the same, i.e. type of work + instrumentation (both terms having strictly defined syntax). So the tag reads: "Requiems for solo voices, chorus and orchestra". All other requiems with the same scoring will have the same tag, and the creation of the tag immediately creates the category of the same name ("Requiems for solo voices, chorus and orchestra"), under which all the works concerned are listed.

The "bottom-up" approach is the novel feature of the system, so rather than start by thinking "how can we break down this large category into smaller and lower levels?", the question will be is "what larger categories does this smaller category belong to?". So, we see that our newly-created category of "Requiems for solo voices, chorus and orchestra" can itself be placed into two larger categories: "Requiems" and "Works for solo voices, chorus and orchestra". Furthermore, the category "Requiems" falls into two the higher-level categories: "Religious music" and "Funeral Music", while "Works for solo voices, chorus and orchestra" is a sub-category of "Choral works".

The connections between the categories are defined when a category is created (using the standard Wiki method of adding "[[Category:xxx]]" to the appropriate page. There is plenty of flexibility for creating new categories at a stroke if the demand exists, simply by combining those at a lower level, and without the need to do any retagging of work pages at all.

Anyway, back to Verdi's Requiem, which by adding the one initial tag of "Requiems for solo voices, chorus and orchestra" can now be found by anyone searching under the following headings:

* Requiems for solo voices, chorus and orchestra
* Works for solo voices, chorus and orchestra
* Choral works
* Requiems
* Religious music
* Funeral music

Some of these "access points" are based on instrumentation, some on the work type, and some on the occasion for performance, which is where all of our previous ideas have stalled. It will require only a minor change to the workpage template, just to allow for the new tags, and at some point the work submission form will need to remove all references to genres and sub-genres.

While the idea is very clear in my head, my typing doesn't always catch up with it, so let me know if you need a better explanation :? But if you're satsfied, then I'm currently working on guidelines for the project team, which should be ready in about a week, and a new calendar year seems like a good time make a start. What do you think?