Topic: [Feature] Character-oriented tagging

Posted under Site Bug Reports & Feature Requests

Preamble / why I'm suggesting this:

So I've noticed an increase in the number of tags which try to specify multiple pieces of information at once. I'm talking about tags like, for instance, overweight_male, which take two descriptors of the same character and connect them.

I've made it clear before that I don't support this type of thing, because if the principle were applied consistently, it would dramatically increase the number of tags in use, and would dramatically increase the effort of tagging images correctly.

So I'm proposing this as the only solution I've been able to think of that allows both the specificity that is clearly desired as evidenced by those sorts of tags, without dramatically increasing the difficulty of tagging most images.

Requested feature overview description.

Effectively, the concept of character-oriented tagging is this:

Users, when tagging an image, would be able to define a character (as in, something which would count towards the solo-duo-group tags), and then would be able to tag that character with any number of tags which apply. When searching, users would have the option to either search for terms normally, which would function mostly identically to how it functions now, or to search with a certain syntax which searches through characters defined on posts and returns posts which have a character which has all such tags on it.

Tags which can't be tied to a character would remain in a general pool.

An example:

Let me pick an example image to show how this would work. This one should work:

post #1997491

There are two characters in the image. So we'd define two characters, and there would be a general pool for tags which aren't specifically linked to either.

General:
slimefur female/female feral_on_feral duo forest size_difference
Character 1:
female bayleef scarf_only mostly_nude pussy anus feral on_top looking_back 1_toe smaller
Character 2:
dergonite_(character) female dragonite dragon lying pussy anus feral 3_toes larger

I have a few things to point out here. First, you'll notice that tags like female_on_top are no longer necessary, because that is simply a character that would get both the female and on_top tags. Similarly for tags like larger_female or smaller_female - generic larger and smaller tags would suffice.

Note also that there are some repeated tags between the two characters.

Why would it be useful?

An analysis of of pros and cons:
Cons:

Most importantly, this is almost certainly technically infeasible. This would require a substantial change to the code itself, which is not true for simply adding new tags.

This would require substantial work on the part of taggers to bring it to a level that would be effective - however, mitigating that is that the current trend towards combo-tags would require probably much more work in the long-term.

It makes sense for some implications to go within the character the original tag was placed under, but it could be argued for others that it might make more sense in the general pool (an example of the former: rodent -> mammal; an example of the latter: smaller -> size_difference, though it could also be argued that something like this would be fine to remain in the character object).

Some tags would necessarily need to be repeated, and for images with many many characters visible, this could be a lot of work.

Since many existing combo-tags would effectively be replaced, users in general would need, on average, to use more search terms per search.

Searching through multiple characters on images would probably be harder on the system than simply performing one search per image.

Pros:

Many existing tags could be eliminated, as their function could be replicated with simpler tags. This is of particular note with combination-tags.

There would be a massive increase in the amount of search-specificity, allowing users to search for combinations between tags that would be entirely infeasible to tag in the traditional sense.

The amount of additional work caused by defining characters would, for the typical image, be fairly minor, and offset by the reduction in the number of unique tags necessary.

The organization of character-oriented tagging might make it easier for users to keep things clear and tag more effectively - effectively, it makes certain key tags more obvious, especially when they are missing, compared to them being in a huge jumble in a big "general" pool of tags

What part(s) of the site page(s) are affected?

Tagging system.

---

To be clear, like I said earlier, I don't think this is actually technically feasible, but there would be some pretty dramatic advantages to this system, so I thought it might not be a terrible idea to post it anyway.

Updated by Fifteen

This has been suggested before, and the answer was "it's a great idea, but impossible to make it work with current systems"

Updated by anonymous

hiekkapillu said:
This has been suggested before, and the answer was "it's a great idea, but impossible to make it work with current systems"

I suggested something similar, but unpolished and with some errors, as part of another thread, way back in the day. If i remember right it was pretty much ignored at the time, so I thought to get rid of the obvious flaws and suggest it properly.

I don't really remember anyone else suggesting anything similar to this.

But yeah, like I said twice here, I recognize that it's almost certainly not technically feasible. It would certainly require a pretty substantial change the underlying system, and I doubt that'll be done.

Still, as far as I'm concerned, it'd be better to put the work into this than creating several hundred thousand combo-tags and expecting anyone to keep up with tagging them.

Updated by anonymous

Note: I feel compelled to use the term 'group' here, as I'm mostly referring to 'any logical grouping of tags on a post, including both character-grouping and the "general" group'

Some tags would necessarily need to be repeated, and for images with many many characters visible, this could be a lot of work.

A DnD interface with one textbox per group could reduce that (and might be necessary anyway -- it would be a big pain moving 'general' tags into character groupings otherwise). Ideally each textbox would have a different border color (assigned sequentially from a predefined palette, probably) for quick identification.

More speculatively, it would be nice if groups could each have a rectangle on canvas (like notes) with the matching color; that would make it much easier to make sure you are adding tags to the right group, particularly if things can be laid out so the image and tags are simultaneously present.

I'm not sure what could be done to adapt the 'Tag' browsing mode to grouped tagging, given the need for extreme compactness. Maybe have a combobox which selects between existing groups, and have a special syntax that could be used to 'create a new group and put x,y,z tags in it'

Erroneous groups would certainly be created at times, so the ability to merge groups would help to reduce the need for admin intervention.

Searching through multiple characters on images would probably be harder on the system than simply performing one search per image.

No doubt searching with grouping would be harder on the system; it would probably still be done in one search though.

(considering that it's possible to compute the mandelbrot set or solve a sudoku in one SQL query )

Updated by anonymous

I think there's several ways it could be done. Would be easy enough to have some part of the tag form that you can select to "add a new character" which opens up a new entry box - that's sort of the solution they use for like, job application forms and such for entering in work experience.

You mentioned a few possibilities in your post.

There could also be just the one entry box but with some special syntax that denotes a group, but this would be more obscure and difficult for users to use, so I could see it only as an additional option in addition to one of the above or something similar.

As for actually searching / browsing images, we'd probably just have some special syntax for the search to interpret. It could be something like how "or" searching is done (with a ~ before the search term). This would limit users to specifying linked details of only one character at a time, but I think that would be mostly fine and might cut down on the amount of work that the system needs to do. Alternatively some different delimiter could be used - instead of separating tags by a space, tags the user wants to be on the same character could be comma separated.

Updated by anonymous

Clawdragons said:

There could also be just the one entry box but with some special syntax that denotes a group, but this would be more obscure and difficult for users to use, so I could see it only as an additional option in addition to one of the above or something similar.

I might have failed to describe that properly. What I meant is analogous to artist:xyz -- something you can dump in any tags field that will take a special action.

It seemed necessary to me because, well, suppose you are preparing the tags on your upload in advance -- you are uploading a series of comic pages, say. It would be a pain to do 4 separate copy/pastes, one for 'general' and 3 for the 3 characters present.

The form+javascript-based 'add another' system you describe seems to me more suited to sporadic, low volume uploading.

Alternatively some different delimiter could be used - instead of separating tags by a space, tags the user wants to be on the same character could be comma separated.

That's certainly more compact than what I was thinking of, and there don't seem to be any tags containing , that would present a problem for this.

To me this is also a grammatically better choice because the group of tags is actually presented as clearly being a group of tags, instead of terms that share a character prefix.

Updated by anonymous

savageorange said:
I might have failed to describe that properly. What I meant is analogous to artist:xyz -- something you can dump in any tags field that will take a special action.

It seemed necessary to me because, well, suppose you are preparing the tags on your upload in advance -- you are uploading a series of comic pages, say. It would be a pain to do 4 separate copy/pastes, one for 'general' and 3 for the 3 characters present.

The form+javascript-based 'add another' system you describe seems to me more suited to sporadic, low volume uploading.

Yeah, that's kind of my point. There should be an easy-to-use option for basic users and some group syntax that works anywhere for advanced users. I didn't realize you were also asking for something like that.

I was thinking something along the lines << and >>. This would let users copypaste easily, and do other things like merging groups fairly easily, just by adjusting those.

So to go to my example image from earlier, in the single-box format it would appear like so:

slimefur female/female feral_on_feral duo forest size_difference << female bayleef scarf_only mostly_nude pussy anus feral on_top looking_back 1_toe smaller >> << dergonite_(character) female dragonite dragon lying pussy anus feral 3_toes larger >>

Which should be easy enough.

That's certainly more compact than what I was thinking of, and there don't seem to be any tags containing , that would present a problem for this.

To me this is also a grammatically better choice because the group of tags is actually presented as clearly being a group of tags, instead of terms that share a character prefix.

I can see advantages to either system - I tend to agree with what you've said here, but I can see some advantages to the other way. In short, I think both could be effective solutions, it would depend on personal preferences.

Updated by anonymous

A while ago, I was toying with the idea of having a selection-based tagging system, which I was jokingly refering to as "Show What You Tag" (SWYT). This was more of a proof of concept to see if this couldn't work as a tagging workflow, but after mentionning it to a few people, the reactions I had was that (in an ideal world) this could be a generalization of this kind of character-oriented tagging into some kind of feature-oriented tagging.

Regarding ease of implementation, i'm not entirely sure. We've now passed 2 million images on e6 with an average of 30 tags each, and keeping searches snappy with that kind of volume means sacrificing some functionalities. Adding a "tagged zone" column or a "tagged character" column in most databases wouldn't be so hard, but optimization means a lot of that extensibility goes straight out of the window.

Updated by anonymous

  • 1