Danbooru

Garment sorting thread (formerly called 'Superfluous tagging page?')

Posted under General

Ever since forum #130116, I've been thinking of drafting up some kind of wiki about needlessly overtagging everything (henceforth called superfluous tagging) and figuring out what guidelines would be written in the wiki to determine what's good tagging vs. bag tagging. Although undertagging is often discouraged, lately I've been noticing that overtagging is starting to become an issue, and I hope such a wiki could help guide users on what's not necessary to tag.

For instance, I updated post #2715867 to remove tags such as blue_tank_top, horizontal-striped_tank_top, and multicolored_tank_top. All that's really needed are tank_top, blue_shirt, and horizontal_stripes to get the same search results. It's tags like the first three that become so ridiculously specific that it's hard to imagine anyone searching for those, when it's more likely for users to search for the more general tags that the latter three represent.

Maybe such a wiki isn't necessary, but I would like to see something that could help us reach an agreeable middle ground when it comes to tagging.

Updated

One of danbooru's key features is an extensive tagging system. There can never be enough tagging.

As long as the basic tags are also there then I don't see a problem. It's not like there's any requirement or punishment for their use. Those tags being there hurt no one and only gave the picture a higher clerical organization.

It seems sacrilegious to reduce tagging accuracy on danbooru.

There have been multiple times where I've been looking for pictures of characters in specifically colored clothes and these tags have been very helpful to me.
If we want to talk about possibly superfluous tagging, being specific about a very prominent part of the image is not the thing to single out.

I believe you yourself @Benit149 helped tag the "neckerchief," "choker," and "sailor collar," among probably some other color subsets that I began or at least helped initially seed. What are some other examples, "blouse" perhaps is also woefully undertagged in its color and other variants (collared, etc.). So what's the problem, you have to start somewhere with a tag; just because others were too cautious for many years in making the first step doesn't mean it can't be taken now.

I don't think your changes on my post were productive. They were in fact the opposite, counter-productive in that they delete useful information and squash a young tag just because it isn't very populated yet. Why is it that tank tops can't have subcategories and other conventions like the other garment tags? Maybe we should go back to not having color categories for chokers and neckerchiefs too?

Eh, I stand on the side along @evazion and other users that there is. There is such a thing as overtagging. See my comment in forum #129622.

I know some of you believe that overtagging a post does no harm. I stand on the side that it does. Again, for one it increases the workload for everyone who tags indirectly, since there's more tags to keep track of that no one ever bothered to give consensus on whether to keep or not keep. Some may find it useful, that's great, but the admins won't go through the work of applying implications for tags that only ever see use by 1 or 2 users unless they're so profoundly present and never been noted to deserve a tag before.

Tags are for recognizeable features, not to pad a number on the sidebar. If a tag is too specific, it is of extremely little use to anyone because of the gardening it requires.

My thoughts on this are this: If all a user does is focus on colors or something else that can be considered marginally important (blue tank top, horizontally striped <article of clothing>), rather than the recognizeable features that make up an article of clothing (collared shirt, wing collar, dress shirt, blazer, cabbie hat, cat hood), then it diminishes what can be considered sufficiently tagged.

It's the same reason why we don't tag everything "red kneehighs", "red thighhighs" because they can be generally assumed to be easy to search. If I wanted to search a "horizontally-striped tank top" I'd have a pretty good guess that all I'd search is "striped" (because we assume striped is by default horizontally striped unless vertical stripes is present) in tandem with "tank top".

Take it how it is, but the difference between something like multicolored_tank_top and sailor_collar is that one is pointlessly specific, and the other is frequently present and can stand in for something like "sailor_leotard" because you'd search sailor_collar leotard and get a generally good amount of hits.

Updated

I think the only superfluous tags are very general and unhelpful ones like "tag", "empty", "full" and similar which aren't very descriptive. Otherwise, it may only have 3 posts attached to it (like some cosplay tags) , but it's useful if I'm trying to find cosplay of that character (like Humboldt Penguin) and not the hundred or so Humboldt Penguin pictures.

The *_(cosplay) tags aren't superfluous at all though -- they serve a special purpose for finding cosplays of a certain character. Their function is described in cosplay (which is why you see the character being cosplayed also tagged).

chinatsu said:

I believe you yourself @Benit149 helped tag the "neckerchief," "choker," and "sailor collar," among probably some other color subsets that I began or at least helped initially seed. What are some other examples, "blouse" perhaps is also woefully undertagged in its color and other variants (collared, etc.). So what's the problem, you have to start somewhere with a tag; just because others were too cautious for many years in making the first step doesn't mean it can't be taken now.

I don't think your changes on my post were productive. They were in fact the opposite, counter-productive in that they delete useful information and squash a young tag just because it isn't very populated yet. Why is it that tank tops can't have subcategories and other conventions like the other garment tags? Maybe we should go back to not having color categories for chokers and neckerchiefs too?

Yes, I helped out with populating those because I knew the base articles of clothing were well-known and recognized enough on Danbooru that they could be separated into colors and patterns to generate stronger search results. I'm fine with populating basic clothes like shirts, shorts, skirts, kimonos, jackets, hats, etcetera. I start to feel a little iffy when we start taking really specific types of those basic clothes like tank tops, baseball caps, cropped jackets, bike shorts, or whatever else is out there and tack a color/pattern to that as well for the sake of 'good tagging'.

Hell, when I was working on the colored yukatas, I had to actually alias the colored yukatas to the colored kimonos because, like I said above, yukatas are a specific type of kimono. I think that's the first time when I started to feel uneasy about making colored tags for specific garments. If left unchecked, all of these tags could become a pain in the ass to keep track of for implications and/or aliases.

I stand on the side that there can potentially be such a thing as overtagging, since the notion for different colors of ice cream was vetoed alongside the yukata issue I had. That got me to asking the question of, "What else would people see as pointless if I tried to make a BUR?" That's what the proposed wiki would be for. It's not to quash tagging in and of itself, but rather to act as a guide for those who wants to make implications or aliases.

Either way, @chinatsu I add the tags striped tank top and blue tank top back onto your post. The only one I disagree with is the multicolored_tank_top tag because the stripes are simply different hues of blue. It would have to be blue and a completely different color other than black, white or grey for it to be multicolored.

Updated

I wouldn't mind if there were a more generalized tag for them. What I object to is calling that garment a "shirt" because it's clearly not. when you tag it as "blue shirt" instead of "blue tanktop" + tanktop you add the shirt tag to the post because of the implication from blue shirt to shirt, thus mistagging the post.

Why don’t we just extend what works perfectly fine for hair and legwear?

We have tags like long_hair and short_hair, black_hair and blonde_hair. Those can be combined as fit. We don’t have overly specific tags like short_black_hair.

We also have tags like socks and kneehighs, along with black_legwear and blue_legwear, allowing for all possible combinations. Tags like blue_socks were nuked a while ago. socks and kneehighs, for example, have excellent wikis describing this system.

Now why don’t we do the same for other articles of clothing? I have to admit that something like black_topwear and blue_bottomwear sounds a bit contrived, but it matches topless and bottomless, which we already have. This would also make it easier to find colored articles of clothing regardless of type. For example, a user might not care if it’s a blue_shirt, blue_blouse or blue_vest.

This can actually be extended to all other clothing items, which is would be more consistent and easier to tag, IMO: <color>_swimsuit, which gets rid of all the <color>_bikini tags because bikini implicates swimsuit, <color>_underwear to get rid of all those <color>_panties and <color>_bra, <color>_headwear, <color>_footwear, etc.

Bonus: This also cuts down on the amount of implications required because we don’t need any <color>_<item> implications to <item>.

I'm in full support for something like that. It's a good idea, makes everything a lot easier to tag. We do lose super-specific article combinations as a result, but the tradeoff is probably worth IMO.

Also disagree with combining panties and bra ones. They're completely different items of clothing worn on completely different parts of the body.

My biggest problem with the other things is that quite often you'll have multiple things that would then fall into the same category being worn by the same or different characters. So a character who wears a black jacket over a white shirt gets black topwear and white topwear along with jacket and vest. Exactly the same as someone wearing a white jacket over a black shirt. post #2555293 would become black topwear brown topwear cardigan jacket shirt white topwear with no indication as to what was what. I really don't want that. Yes, I have used tags like white shirt and black jacket in searching before now.

With images with a lot of characters in different combinations this quickly descends into total meaninglessness. For instance, the Kagerou class have green vests, grey vests and black vests, white shirts, white blouses, a beige shirt (Arashi, though usually not drawn as such by artists), a white serafuku top, white sailor dresses*, a brown sailor dress*, and Akigumo is wearing a red dress as well if you want to go as far as including that*. So a picture of the class could have green, grey, white, black, beige, brown and red topwear tags, with no indication of what they are referring to. And that's before you get into the possibility of neckerchiefs/neckties/neck ribbons being included*.

This is more an issue with some things than others. I'm not really fussed about the idea of shoes, sandals, slippers and boots all getting put into a footwear category (although with a name like that people would probably stick socks in all the time as well).

*And you'd have to decide what actually counts as topwear as well - do these?

So of the things above:
-1 to underwear, topwear and bottomwear (although the latter two could exist if people wanted them, with implications to them from more specific tags)
-1 to headwear because hat covers most of the things that would make sense already (I wouldn't want headbands, hair ornaments etc. included within this, and headgear often aren't even worn on the head directly)
Don't really care about footwear, but if pushed I'd be leaning against it purely because of the amount of tag gardening it would likely create to clear out things that should be legwear instead. A preemptive -1 to combining footwear and legwear together as well.
Unsure about swimsuits.

Updated

That sounds like a better option. For example, when I wrote the *_shirt wikis, I wrote them as, "Any type of shirt that is colored (COLOR)." I think that part of the sentence was obviously lost on a lot of people, so something like *_topwear and *_bottomwear would help immensely with implications and aliases.

I just don't agree with bras and panties being merged into an umbrella *_underwear tag since they're specific undergarments that are popularly depicted in images. I'm also on the fence with *_swimsuit and *_bikini being merged. Yes, a bikini is a type of swimsuit, but it's such a popular type that it's basically treated as its own category.

So from what I gather, the recommended list of umbrella (color)_(clothing) tags would be:

*_topwear: shirt, tank tops, crop top, blouse, sweater, jacket, coat, etc.
*_bottomwear: pants, shorts, skirt, dress, buruma
*_headwear: hat, headband, helmet, bandana, hairband, etc.
EDIT: Following what @kuuderes_shadow said, this portion of the list is highly debatable, while the bottom portion is more agreed upon or already in use

*_panties
*_bra
*_swimsuit
*_bikini
*_footwear: any length of boots (ankle, knee, thigh), any style of shoes (mary janes, high heels, sneakers, sandals, etc.)

I had started the alternate bottom wear tag at one point to cover some changes from canon that were not adequately covered by other tags (no pants characters wearing pants, shorts-wearing characters wearing skirts, etc.). Though if "bottomwear" is preferable as one word, I'm fine with an alias there.

kuuderes_shadow said:

My biggest problem with the other things is that quite often you'll have multiple things that would then fall into the same category being worn by the same or different characters.

That’s a general problem with many other tag combinations as well, like long_hair black_hair matches post #1621150. The only alternative is combinatorial explosion, aka “superfluous tagging”, with an unwieldily large amount of tags.

EB said:

Though if "bottomwear" is preferable as one word, I'm fine with an alias there.

I suggested the contracted version because it matches most of our dozens of other *wear tags, like *_legwear. alternate_bottom_wear is actually one of only five non-contracted *_wear tags.

I’m also not set on the underwear thing. I just listed off what I could think of. It seems like most of you want to keep the colored bras and panties.

kittey said:

That’s a general problem with many other tag combinations as well, like long_hair black_hair matches post #1621150. The only alternative is combinatorial explosion, aka “superfluous tagging”, with an unwieldily large amount of tags.

True, but this only affects situations with one character with long hair that isn't black and the other with black hair that isn't long. A search of long_hair black_hair solo should still return only images containing characters with long black hair (bar possibly a few odd situations). And the vast majority (looks like around 90% from a quick scan through) of the images in long_hair black_hair are of characters with long black hair.

This would affect any clothing item that is black and any clothing item that is, say, a jacket. These could be worn by the same person, or worn by different people. Or not worn at all. Or being worn on a different body part entirely. white_jacket (62 pages) has far fewer results than even white_shirt jacket (199 pages) and would be completely drowned out in a hypothetical white_topwear jacket search.

The two things aren't really comparable to one another.

This proposal would remove a significant level of utility from the search in a quest for a false simplicity.

What about color combinations for some of those items, e.g. a white shirt underneath a grey jacket/blazer...? Or black shorts underneath a green skirt...?

Is is grey_topwear/white_topwear for the first and black_bottomwear/green_bottomwear for the second?

Also, what about multicolored...? Will it count as it currently does, i.e. a single article of clothing is multicolored, or does having a blue jacket with a pink shirt count as multicolored...?

Also, I saw dresses being counted as bottomwear... but I'd be against that. Have a different category, like *_onepiece. That would also cover leotards, jumpsuits, robes, etc.

^And long coats or jackets, whereas shorter ones would be topwear... Arguable oversized clothes would sometimes be one and sometimes the other as well.

The more I think about it, the more counterproductive the whole thing seems. And it seemed counterproductive to begin with.

1 2