Another way to count characters, looking for feedback

I think we need a totally different mechanism for counting characters, so that we don't have a constant tension between "x should logically implicate y" and "but then the char tags will be misleading". Both are usually valid points, so the solution would need to be in functionality.

Brute force solution: add an additional field independent of the chartags: metatag where we can manually select a number of characters. We would use human judgment to count rather than trying to shoehorn conflicting desires into our tagging policy.

It could work in addition to the chartags: metatag, rather than in place of it. "numchars:" perhaps. It could be populated during tagging by typing in numchars:3 or with another text box when editing.

Obviously the challenge to adding anything like this so late in the game is: IMPOSSIBLY MASSIVE BACKLOG. But that shouldn't necessarily stop us from something that would be useful going forward. At the very least it's an addition of functionality, not a removal or replacement or alteration of existing functionality, so it doesn't harm anything if it doesn't take off or misses legacy content.

Or we could just say this isn't necessary and live with chartags: being imperfect, which works just fine for me as well. But a lot of people don't seem to be find with that, so here's an idea. Not an elegant idea, but likely an easily-implemented and simplistic to understand one.

I'd support this brute force method you propose. It'd bring back a lot more leeway with the character tags.

(ignore the tag discussion, missunderstood the issue)
i'll say this, is this a tag that we desperately need? If we don't have it, will it change anything?

Aside from that, id be ok with a tag like this, but we would need a cutoff point in terms of how many number tags we would need. we don't want 1-100 cause thats just gonna be too many tags to manage at a time

I feel we should only have bumber tags for 1-9, cause realistically, who will search for anything beyond that?

I do feel that any picture that has 10 or more charcters should reciece a +10 tag numchars:+10, cause again, i feel that would help simplify the tag a bit.

It's not a tag, it'd be a field like putting in the parent post ID number in the edit options.

NWF_Renim said:
It's not a tag, it'd be a field like putting in the parent post ID number in the edit options.

ah, ok, sorry for misunderstanding. ok, that sounds a bit more pratical then a tag.

Why do we need this? What sort of searches would this help with?

It wouldn't affect me much, but some prominent members (Rantuyetmai strikes me as an example) do prefer that the character tag count be as accurate as possible and would prioritize that over tagging other aspects (e.g. characters split into multiple taggable components (forum #54143) or cosplay (forum #58272)).

I sort of like the proposal even if it is manual, and if I had free reign and more self-confidence in my programming, I'd do all sorts of things to Danbooru's tagging system to improve semantics (which would undoubtably kill it's usability & scalability unfortunately).

I wouldn't say the impossible backlog would be a necessary deterrent here. Especially since we could default the values to the ones we currently calculate (propagating through all the old posts), and simply let people change that value where it fails to count properly. I'd suggest we use that default for new posts as well since people are going to inevitably ignore them otherwise, and the current method does work for most posts.

Of course this is all up to Albert. I like the idea of improving our semantics in general, though to me this isn't the most important statistic.

How about just making it possible to manually modify the chartags: count? That way it can be fixed when necessary and simply left alone in the default case.

evazion said:
How about just making it possible to manually modify the chartags: count? That way it can be fixed when necessary and simply left alone in the default case.

I would say have the chartags count seed the numchars count by default instead, but yes, this seems like a sound idea.

albert said:
Why do we need this? What sort of searches would this help with?

I think jxh2154 made the case pretty clearly, but the examples that springs to mind are ex-keine->kamishirasawa_keine and a whole host of magical girl transformed alias->real life name implications.

I'm actually happy with our system right now, if my previous discussions seem like I'm being anal about it, I'm sorry.

A separate system to count characters will be nice to have, mostly because there's no way to work around with the majority of original posts (we don't have 1boy_and_1girl_who_are_not_couple kind to tag after all). But this is what I consider a luxury we can live without, depend on programming time and effort it requires.

evazion said:
How about just making it possible to manually modify the chartags: count? That way it can be fixed when necessary and simply left alone in the default case.

I like this, but at the same time finding explicitely posts where numchars != chartags might be useful as well so I dunno.

+1 to the overall suggestion at any rate since people seem okay with it.

evazion said: How about just making it possible to manually modify the chartags: count? That way it can be fixed when necessary and simply left alone in the default case.

I figured people might like both options to remain.

But if it were accepted I would propose that we initially "seed" the numchars field with the chartags value. It won't be right in all cases, but it will be in the vast majority and we could tweak the remainder.

Oh, and I suppose there's also the possibility of using this to count unnamed entities in an original image, if so desired. chartags can't do that for obvious reasons.

Edit: I see rantuyetmai mentioned the original aspect already. And glasnost mentioned the initial seed. So this is a useless post!

This could replace the solo tag.

The 2girls, 3boys etc tags could be used to seed a numchars field on original images. Sometimes people use them on mixed-gender images, but I think it would be just as accurate as a chartags seed.

It would be a good idea to have an easy way to tell if an image just has the numchars seed value or if it's been validated by an actual person.

Just out of curiosity, would this new way of counting characters help/hinder searches where a post contains more than one character and there may be multiple copies of one or more characters (such as clones or mass-produced mecha) or would nothing really change?

I think I could actually cheaply allow custom tag counts by letting them be negative. If it's negative, it was explicitly set by a user and defaults should be ignored. I probably won't implement this until Danbooru2 though.

The downside is this is just another avenue for vandalism. I would probably restrict it to janitors only.

Janitors only? Come on, at least make it Privileged and up.

Edit: After thinking about it some more, I don't see why this new character count feature can't be extended to regular Members as well. I'm assuming the count changes will be recorded, so vandals will be caught quickly anyway.

On the other hand even if changes are logged, people aren't going to know to check a post to see if it is appropriate or not. A potential vandal could just go back a few months and edit things with relative impunity.

The problem with allowing Privileged this ability is that it's still a level people can achieve without by being vetted by someone higher up (by donating the $20). I can see maybe opening this up to contrib+, but it should be a rare enough occurrence that these counts need manually set, that I don't see that it'd be a problem to restrict it to janitor+.

In the worst case, we could set up a forum thread for those unable to change things to point out places it needs to be done.

Seeding it with a mix of the current character count tags and using solo to fill it with 1s where character count is greater than 1 and solo is tagged would make it completely ok to me to have it restricted to janitor+.

Hell, I'd probably fix cosplay and original myself.

Shinjidude said: The problem with allowing Privileged this ability is that it's still a level people can achieve without by being vetted by someone higher up (by donating the $20).

If someone is willing to contribute $20 to the site in order to do petty vandalism that we could reverse in mere minutes, rendering their efforts a total waste of their own time more than anything else, I say let 'em. They'd just be banned anyway.

Anyway, I don't think a fear of potential vandalism should too heavily influence whether we implement a feature. Not if it's essentially just another tag change that can be easily reversed.

I completely agree with these points made by

Shinjidude:

ome prominent members (...) do prefer that the character tag count be as accurate as possible and would prioritize that over tagging other aspects (e.g. characters split into multiple taggable components (forum #54143) or cosplay (forum #58272)).I sort of like the proposal even if it is manual[.]I wouldn't say the impossible backlog would be a necessary deterrent here. [W]e could default the values to the ones we currently calculate (...), and simply let people change that value where it fails to count properly. I'd suggest we use that default for new posts as well[.][T]his is all up to Albert. I like the idea of improving our semantics (...), though to me this isn't the most important statistic.

(Yet it is a statistic I could sometimes make good use of.)