Artist Management for Albums

[Gigablah]

The artist table and associated artist/album crossreference table is a rather complex addition to the VGMdb system, and as such testing will be limited to trusted editors and staff members. This thread is concerned with the artist-to-album linking component (crossreference table); the actual interface for managing artist entries in the main artist table will be discussed at a later time.

Firstly, the main goal of the crossreference table: to normalize the artist fields for albums (composer, arranger, performer -- and lyricist, to be added later) such that:

• All artist names will conform to naming conventions and accurate spelling
• Ambiguity is eliminated by ensuring that all aliases of a particular artist point to the same artist entry (and are listed in the profile)
• Individual artists can be displayed as links in the album view pages, which will take you to their profile pages
• Artist name display can be switched between different languages
• Queries for all works by an artist, in whatever role, can be performed efficiently and return accurate and comprehensive results

This has implications for the album view pages: If a particular album has artist reference entries, the regular artist strings for composer, arranger, etc. will be bypassed in favor of those entries. However, artist fields in the album database will remain. They'll still show up in the album submission and editing interfaces, and you can still update them. They serve a function for the new artist management interface, which I will explain now.

Let's take a look at an example album, Clannad Arrange Album MABINOGI. You can access the management interface via the Edit menu:





An album without any entries in the artist-to-album crossreference table will have no content under "Linked Artists" and "Unlinked Artists". If the artist strings for the album contain data, though, they will be parsed and listed under "Unassigned Artists":





Notice that the original strings are shown as a reference, in case anything horrible happens during parsing (for example, when track numbers are listed with commas after an artist name).

Before you click on "Assign" though, check whether any of the names contain suffixes. Suffixes are extra information after a name; anything that isn't the name itself. Therefore, we need to fix the first guy.





KEY is an alias for Jun Maeda (for this album, at least). Note that you can also just go ahead and replace KEY with Jun Maeda, but since we're following what's printed on the album liner notes, we should preserve the alias. Once you click on assign, all the names disappear from the unassigned section, and reappear in the unlinked section.





From this point on the album will retrieve these entries for display on the album view page. However, the artist names will still NOT be linked -- you will have to perform the next step.

First, notice that under "linking options", the system has found matches for Shinji Orito and Magome Togoshi. Every matching name in the artist database (yes, there could be two or more artists with the same name) as well as names associated with every matching alias in the crossreference table will be listed in the dropdown box. We don't have entries for the other artists (except Jun Maeda, but we'll get to that part later), so let's create them now. Scroll down a little bit to find the multiple artist name submission form:





Uncheck the box for KEY because it's an alias, and we don't want to add aliases to the artist database. Submit the rest, and now we get this:





Excellent! KEY needs to be manually linked to Jun Maeda (prior to this, I've already added him to the artist database with ID 4) because it's a new alias which hasn't been recorded in the crossreference table yet. After submitting this information, the system will suggest matching KEY to Jun Maeda from now on. Remember to check all the boxes if you want to update the entries (the mass checker on the top right is handy here).





And this is the result. We're done linking artists for this album. Notice that the aliases for everyone except KEY have disappeared, because the system detected an identical match between the provided alias and the artist name in the database and cleared it. (It's not an alias anymore!)


Question: What if I DON'T want to use this system? Yes, you can completely disregard this process if you want, it's only there to make adding multiple artists easier and safer. There's a single form for manually linking an artist to the album, if you wish to use it:





As long as the artist you add in a particular role matches the artist detected by the parser from the album's corresponding artist string, that artist will disappear from the "unassigned artists" section.

What if someone adds a new name through the album editing interface? That's fine, it'll show up automatically under "unassigned artists", and you can repeat the process for that artist.

Current Issues:

• No option to delete entries from the crossreference table yet
• Script does not check whether a linked ID refers to a valid, existing entry in the artist database
• Suggestions do not take into account transposed names (e.g. Orito Shinji instead of Shinji Orito)
• Artist links in the album view page are not shown yet, that'll happen once the artist profile template is finished

Comments and suggestions are welcome!

[Secret Squirrel]

I tested this out on Okami, and luckily, the procedure is easier than it first appears, consisting of 3 basic steps:

1. Assign artists, which takes names from the artists fields and associates them with the album.
2. Add artists, which adds any new artists to the artist database. (Note: have to be careful with alternate spellings, incorrect romanizations, etc.)
3. Link artists, since now, they all point to their most likely candidate artist page.

Still, I think we may want to autogenerate some of these (e.g. sort through the database and link up all the Nobuo Uematsu entries in the artist fields.) Going through this process manually for each album would be time consuming, and probably isn't necessary except for some of the more obscure artists.

[Gigablah]

Linking up an artist to all albums should be done through the artist profile management interface, which I'm still working on. It'll definitely be there.

[Gigablah]

An update to the batch artist submission form:




Artists can be defined as two types: Person or Group. Many artists can be linked to a group, as follows:




The interface for doing this will be contained in artist profile management.

[Gigablah]

Update: Staff and Trusted Editors are now able to delete assigned artist links, by clicking on the trashcan icon beside the entry ID to the left.

[Secret Squirrel]

Is there a way to turn an artist that's already in the database into a group?

EDIT -- Another question ... I wonder how doujins should be handled, since the original artists are credited as composers on a doujin, and these could potentially clutter up an artist's page. (Zun's page would be a good example, since he'll have personal compose/arrange credits on maybe a dozen albums, but there will be hundreds of Touhou dojins that list him as the composer.)

EDIT 2 -- We need some kind of auto-linking/assigning for the artists on reprinted albums.

[Gigablah]

You can switch artist types (Person and Group) on the edit page now.

Most likely I'll filter out Doujins from the discography by default, unless the user specifies otherwise.

[seanne]

Looking at the opening post at first I expected this to be a much more complex process than it turned out to be. Great work Blah, this is definitely an exciting addition to the site.

[Secret Squirrel]

Just a couple more minor issues.

What order should we put the artists in the list? Right now it's in artist-ID order, I think, but it might make more sense to do them alphabetically.

Also, the alias font is larger than the artist name font:

http://www.vgmdb.net/db/albums.php?id=4699

[Free.User]

Wow, this is awesome and much easier than it looks!

Just a thought - is there any way to accommodate artists who are both composer, arranger, and sometimes even performer? The system only allows one choice per artist per album, but that might be getting too detailed.

EDIT: Just realized that this can be solved by having the composer's name in the arranger/performer fields to begin with, duh.

[Gigablah]

- The checkboxes for unlinked artists with suggestion matches are now ticked by default, to save you the extra steps.
- Batch artist submission form in the unlinked artist section has been moved to the left, to prevent confusion with the form directly above.

[CHz]

A form which allows an artist to be added to the database without actually linking them to an album would be nifty.

The reason this came up is that the album I just added, Lost Childlen, has a bunch of arrangements by groups, most of whose members don't have any solo work currently in the artist database. It'd have been nice if I could have created the artists directly and then added them to the groups instead of creating them as an artist on the album to create the database entry and then deleting them from the album. :|

[Gigablah]

Actually, you can create artists using http://www.vgmdb.net/db/artists-submit.php , I just haven't linked it yet.

[CHz]

I just noticed that, if you view an album page with linked artists while not logged in, none of the artist names are linked to their profile pages, and the name suffixes don't show up either (aliases seem to work, though).

Is this intentional, perhaps another incentive to have an account like full access to scans? If so, it doesn't strike me as a terribly effective one while the artist browsing page is still easily accessible to anonymous users.

[Secret Squirrel]

Originally Posted by CHz

I just noticed that, if you view an album page with linked artists while not logged in, none of the artist names are linked to their profile pages, and the name suffixes don't show up either (aliases seem to work, though).

Is this intentional, perhaps another incentive to have an account like full access to scans? If so, it doesn't strike me as a terribly effective one while the artist browsing page is still easily accessible to anonymous users.

It's only temporary. Right now, the new links aren't viewable to guests, but they will be soon. Actually, things are probably far enough along that we could start showing them for all users.

[CHz]

So, suppose that you link the same artist twice to the same album, but in each link, you give that artist a different alias. This works perfectly fine in the album display page, showing each of the aliases separately. However, when you view that artist's profile, only one of the aliases actually shows up.

Before you think I'm just doing something colossally stupid just to break the system, Koichi Namiki is credited with two different aliases on GALAXY FORCE II & Thunder Blade ORIGINAL SOUND TRACK. I think this might be the only album on the planet where someone is credited twice with different aliases.

[Gigablah]

Good catch, I currently have the aliases extracted from the discography loop which groups by album, thus only the first alias for a particular album is detected.

I guess I'll do a separate query for the alias list, then.

[seanne]

We currently have the same Family/Given Name setup for groups as for individual artist entries. Should this be changed so that we have single fields for English/Kanji/Hiragana, enabling us to enter the original names for groups (probably mainly orchestras) as well?

[Gigablah]

The current practice is to enter group names into the "given name" portion leaving the "family name" blank. Admittedly I have not run into any instances where a group has first and last names... but anything can happen.

[seanne]

Ah, gotcha. I'll do that then.

[Cedille]

Originally Posted by Gigablah

First, notice that under "linking options", the system has found matches for Shinji Orito and Magome Togoshi. Every matching name in the artist database (yes, there could be two or more artists with the same name) as well as names associated with every matching alias in the crossreference table will be listed in the dropdown box.

Any reason why it doesn't also try to match a variation? Or does it require more works than by adding just one more condition to the WHERE clause?

[Secret Squirrel]

Yeah, I don't know if there's an easy way to do that. Each artist has a single full name, for which there is one per artist, so you can do a whole word match. There is also an artist/album cross-reference table where each entry is an artist's association with an album. When an artist is linked under an alias, these are found here, also a whole word match.

Unfortunately, the Variations are a comma separated list containing multiple variations for a single artist. We can't match whole words there. We also don't want to partial match, becuase it'll generate a ton of false positives for simple small names. There may be some SQL tricks that could be used, but I don't know if they would totally alleviate this downside.

The variation field was not designed to contain aliases anyway. It is intended for removal of the "x (as y)" clauses from the discography. People have been sticking unused aliases in there. Whenver I catch that in the queue, I do a search and link up those albums properly, since that is probably what they were trying to do anyway.

[Cedille]

Originally Posted by Secret Squirrel

Unfortunately, the Variations are a comma separated list containing multiple variations for a single artist. We can't match whole words there. We also don't want to partial match, becuase it'll generate a ton of false positives for simple small names. There may be some SQL tricks that could be used, but I don't know if they would totally alleviate this downside.

Well, then shouldn't we do some database normalization and create a new table for variations to make it to 1NF?

Also, I knew nothing about MySQL but I just installed it and the database seems to have a function that can match each segment split by a comma, like...

SELECT artist_name
FROM your_table
WHERE @parameter = artist_name OR FIND_IN_SET (@parameter, artist_variations)

Still, I'm not really a capable engineer and just don't get a point of the whole issue. If you think it's not so easy (for technical reasons or the lack of times), then it has to be hard.

[Datschge]

Nice find, didn't know of that one yet (though it doesn't make use of indexes unfortunately).

[Secret Squirrel]

That is one I didn't know about either, and would almost work.

There is still the question of whether that's the proper use of this field. If you enter a name in this field, it removes it from being mentioned in the discography. It's not a catch all for all aliases. The list of used aliases is generated automatically from the links. Also, once a variation is used in a linked album, it is present in the cross-reference table, so it doesn't need to be searched in the variations field.

Maybe we need another field for aliases that haven't been used yet.

[Cedille]

Speaking of variations, I think we shouldn't basically erase (as Y) of the discography. If someone is credited differently in some works, it doesn't hurt to keep such information at all.

We could just merge aliases and variations, but also set a drop-box to classify them, since not every of them the sound of 'alias' fits.

Alias
Maiden Name
Misreading/Typo
Variation - different romanization, simplified kanji ('齋' -> '斉')
Character Name
Unit Name - there are some units that only has one member.
Other - some people want to add names that aren't credits in any production.

Originally came up with this a couple years ago, but thought (and still think) it wouldn't be much approved.

[Phonograph]

I agree except one thing, character name if it's what I think (names of anime/game characters?)
not sure about "Other"