-
Posted under General
Alright, sorry about the accidental messages.
jxh2154 said:
I am actually sympathetic to Cyberia's points. I got frustrated back when I uploaded regularly too. It's sort of part of why I almost stopped, but honestly 90% of the reason is admin work and just having minimal free time in life in general.
Wait, I'm not sure if that's what you meant but frustration from denied uploads isn't the problem. This is more of a consequence.
So just to make sure:
The "lower level" problem is that grinding for Contrib is counterproductive, since you're snatching images that were already going to be posted, rather than doing anything actually useful.
And you know it, but you have no choice.
The "higher level" (and somewhat obvious) problem is that the approval system will always tend to fail for unpopular content.
Good images can have trouble getting approved when their content turns out to be unpopular among Janitors (always approved by the same few people, so more likely to be missed or liked by no one). Examples include males, /trad, unusual art styles...
And there's not much to do about this other than becoming a Contributor yourself (or a Janitor, since it turned out to be faster).
Of course, as long as you only like bishoujo scans this won't seem like a problem to you.
The frustration from getting ninja'd when camping pixiv is just a consequence at the end of the chain, as you waste your time to fill your history for the promotion that serves making you finally able to post those Danbooru worthy images that run into a lack of approvers.
I don't know if I'm making it clear enough that serving Danbooru's purpose is the entire point.
RaisingK said:
I've never bothered trying to stalk pixiv users. I currently have 526 pixiv links (artists and artist responses) queued up in this script. Pick a random link, enable EPP's source search to find posts that haven't been uploaded, pick out anything you like, use PTP with the Danbooru mods to tag uploads faster, clear artist, repeat. Add artists of posts you fav, the artists' last 3 favs, linearts they colored, etc... It doesn't end.
I gave up on taking my time reaching the postcount for Contrib partly because of what RaisingK describes. So many open/bookmarked tabs that I had yet to visit, and already too many URLs piling up under "stuff that mightn't be approved" (since I was scanning artists galleries for possible missed posts, it was natural to find more of these than guaranteed no-brainers).
The final straw was when I started coming back to this list after some months to remember what were the images, to find out around 20% of them had already turned bad_id. Somehow I had no idea it was this common.
There I paused this job and joined the crowd at the front of pixiv to get Contrib ASAP.
Incidentally, it's a pity to see the amount of uploaders who stay glued there when they could have been more helpful to Danbooru by finding more missed images before they disappear.
RaisingK, I had noticed lately that you had been clearing a lot of artists on pixiv of their remaining images.
Correct me if I'm wrong: seeing how a good part of the images you posted this way are too plain to expect they would be approved, when not simply borderline on quality, I take it that not only you're trying to find good art for Danbooru, but you also use Danbooru to save the slightly less good one from disappearing, especially as you're the most aware person of how bad the bad_id issue is.
I don't think I'm going that far this often myself, but I certainly agree on the principle so.
I hope this shows that Danbooru isn't just about the awesome pictures.
Most (if not all) Contrib+ uploaders, including Janitors so, must have run at some point into images that they knew weren't that good, but still looked worth posting, and went "screw it, better here than lost".
~
So,
and/or,
This should also apply to Builders (translations, tagging, not sure about wikis), where the 6 tags limit and tag scripts give a huge boost to your productivity. Hell the privileges might even get more users into contributing.
I apologize if my messages are difficult to understand, but this isn't exactly the easiest topic either.
jxh2154 said:
So educate bad taggers. If they stay bad, neg them. If they are just egregiously bad, they might get banned.
So what would you do about users like the one I mentionned in my first message?
Shinjidude said:
What are others' thoughts on this? Is it worth the effort?
It should be the % of deletions on the last X posts (100-200), because this will always be the starting reason to want to inquire further.
If possible it should even be the % of deletions on the last non-comic X posts, as uploaders who get their "real" numbers obscured by comics are by far the most common occurrence of false positives for potential Contribs (and incidentally that's how the biggest ones found themselves promoted by albert).
Cases of false positives from users who just started uploading, or on the opposite became inactive, are very rare, and are spotted quickly as you start looking at their first/last posts' IDs.
The remaining cases of possible false positives, also rare, can't be detected before inspecting the users' upload history:
- Not so good uploaders who recently posted large image sets (chibis, pokemon...). On the opposite you would miss a good uploader who happened to have a recent large image set partially or entirely deleted, but it's extremely unlikely and there's no way to detect it in advance anyway.
- Uploaders whose pretty much the whole history is made of image sets.
- Not so good uploaders who largely benefited from the presence of one or several overly compliant approvers.
Assuming you ever encountered, as you say, an extremely prevalent poster who saw a sudden improvement of his deletion ratio, either the reason is one of these, or he did genuinely improved.
Generally though, so far I think all the users I've seen becoming Contrib+ got their deletion ratio slowly lowering over time (and potentially just from Janitors alone without much actual improvement on the uploads themselves), or close to 0 from the beginning. So you should already somewhat be able to spot them in advance.
Updated
I do agree in that contrib status should be affected by how crappy or good you actually tag images when uploading.
You can lose contrib by being a MR_GT or get it "quicker" if you tag well while uploading quality.
I also think the usefulness of the content should matter.
Assuming you upload stuff other people have missed and it generally gets approved you should see contrib before someone who is just sniping away at all the new stuff on Pixiv.
It might be worth looking at users (for contrib) whose ratios are not as high but are uploading generally good stuff but from less popular (potentially less often approved) series.
Make contributor status a little less about camping pixiv and more about finding interesting images and tagging well.
I completely agree with you, Pyrolight, but quality of tagging is very hard to see as a mod. First, it's not something that can easily be summarized when looking at a list of users or an individual user's profile. It also isn't even obvious by browsing a user's posts in most cases, since posts usually get embellished by gardeners and other users after posting. Expecting someone to go into the fine grained detail of viewing each post's history to find a user to consider inviting is sort of unreasonable.
Also image quality still has to be the larger part of the equation. If you are excellent at tagging, but frequently post stuff that is off-topic or unacceptably poor, you still need to be affected by the mod queue.
If you can think of a good way to make this sort of evaluation based on good tagging and posting "interesting" rather than "quick and easy" images readily visible, such that those users can easily be separated from everyone else, it'd be a great way to find people worthy of upgrades. Unfortunately those metrics are so subjective and unsuited to quantification, that I'm not sure it's possible.
Shinjidude said:
I completely agree with you, Pyrolight, but quality of tagging is very hard to see as a mod. First, it's not something that can easily be summarized when looking at a list of users or an individual user's profile. It also isn't even obvious by browsing a user's posts in most cases, since posts usually get embellished by gardeners and other users after posting. Expecting someone to go into the fine grained detail of viewing each post's history to find a user to consider inviting is sort of unreasonable.
Perhaps when someone gets a nice image ratio under their belt and are in the contributor range do a sampling of some of the images to see how they are tagging?
Basically just to quick check to see if they are part of the solution or the problem.
Also image quality still has to be the larger part of the equation. If you are excellent at tagging, but frequently post stuff that is off-topic or unacceptably poor, you still need to be affected by the mod queue.
Absolutely. You would have to have an acceptable ratio first.
The idea is that someone who has a 90% ratio and 100 posts but does not tag worth a damn would be chosen after someone with a 90% ration and 100 posts who also tags well.
Make contributor a little more responsibility.
If you can think of a good way to make this sort of evaluation based on good tagging and posting "interesting" rather than "quick and easy" images readily visible, such that those users can easily be separated from everyone else, it'd be a great way to find people worthy of upgrades. Unfortunately those metrics are so subjective and unsuited to quantification, that I'm not sure it's possible.
Really the only way I can even think of is a script that compares the upload dates of images on danbooru to pixiv by user.
Also listing the copyrights of the images might help with using popular vs unique series for uploading.
That would take care of the camping and the objective side assuming the script was easy enough to do.
I can't think of anything really that would deal with the issue of interesting images outside of the mods scanning the posts of potential contributors or the fact the images are unique enough that they actually stand out in the minds of those approving them.
Pyrolight said:
Really the only way I can even think of is a script that compares the upload dates of images on danbooru to pixiv by user.
Querying pixiv for dates would be unreasonably slow. Just looking at the illust IDs might be enough, though, if you can work out a sliding threshold based on that.
Pyrolight said:
Also listing the copyrights of the images might help with using popular vs unique series for uploading.
Too much trouble, can't feasibly automate consideration of copyrights anyway.
A simple deletion ratio based on recent posts is all I was considering. Anything more than that is someone else's business. But we've rather moved on from what I was initially talking about, haven't we?
RaisingK said:
A simple deletion ratio based on recent posts is all I was considering. Anything more than that is someone else's business. But we've rather moved on from what I was initially talking about, haven't we?
Ya I think to some degree we are asking what we want new contributors posting practices to be.
We have enough pixiv campers already and if we can change that way of thinking a bit we might see more "useful" contributions.
Pyrolight said:
We have enough pixiv campers already
That's because pixiv is 'safe'
-Possibly completely off-topic-
As a (probably out of line) generalization, most people with the ability to approve images (with the exception of Not_One_Of_Us and a few others) seem to have very, very specific tastes, so boderline content from other lesser sites are not given the benefit of the doubt
Comparing the type of posts that were transferred to Sagubooru, or even say, 3/4 years ago on danbooru, to now should show an increasingly narrower category of images being approved
However, it's an understandable mindset, given the type of lesser-quality images that get posted on sites without stricter quality adherence
One does have to ask though, what happens when the categories become too narrow? Especially with the numerous booru variants (specialized and general) that have sprung up within the last few years
With respect to content, what does danbooru offer that other sites don't?
-
Once again, it's probably not my place to insinuate anything like that all, but the question's been there for a couple of years now, and I'm curious about how everyone who's been contributing regularly feels
Shinjidude said:
My thoughts are that that is part of why Albert keeps adding Janitors, in order to broaden the scope of what will be approved whilst keeping things within the quality thresholds we want to see as a community.
Is he adding janitors with unique tastes though?
titaniachkt said:
With respect to content, what does danbooru offer that other sites don't?
The big one would be translations.
Once again, it's probably not my place to insinuate anything like that all, but the question's been there for a couple of years now, and I'm curious about how everyone who's been contributing regularly feels
I agree that at times the variety of images can seem a bit narrow but unless someone has done another ratio count on copyrights it's hard to say if it is real or just perceived.
You're thinking too far Pyrolight.
Ordering Contributor promotions based on users' merit is putting the cart before the horse, when the reality is that we already have trouble getting users promoted at all.
Shinjidude said:
Expecting someone to go into the fine grained detail of viewing each post's history to find a user to consider inviting is sort of unreasonable.
When I find an uploader in the range for a Contrib or Builder promotion or close to, I check the history on some of his recent uploads.
When the user is borderline on the uploads alone (quality or count), tagging and other contributions can help making the difference.
This mostly helps users who don't upload that much but show an all around good comprehension of the site.
RaisingK said:
A simple deletion ratio based on recent posts is all I was considering. Anything more than that is someone else's business. But we've rather moved on from what I was initially talking about, haven't we?
Please look at my reply to Shinjidude in my previous post if you haven't.
Right, I honestly have gotten sick of this problem being ignored, so I'm going to do something about it myself.
This an uploader problem. This is caused by uploaders, and it effects uploaders. It causes good uploaders to leave, regardless if the whole reason of "first" is stupid (well less "first", and more "wasted effort"), having good uploaders leave because of bad ones it not something that should be ignored.
The really bad part about this is that the worst people doing this are janitors. I mean really.
We also have people that I see downvoting when they don't get "first" (and I could actually name a few).
So I've started running what could essentially be called an "upload bot". This can auto-tag, rate, and upload faster that any user. This fixes the problem by making it impossible to happen in the first place.
It's only running on a few whitelisted artists at the moment, but will be expanded in time.
Albeit the tagging isn't perfect, but considering "tagging on upload" doesn't matter as long as it gets tagged in the end, that isn't exactly a problem.
I've actually had this thing for a few months. Have been hoping you know, uploaders would get better. Apparently not.
Uploaders probably won't like this, but honestly it's a good thing.
This pushes people into actually "looking" for content. You make it impossible to watch pixiv, people then have to look. We get more "rare" content as a result.
Sorry if my attitude sounds a bit off here, but seeing this problem just be completely ignored despite the problems it actually causes confounds me.
Regardless if the problem is stupid or not, or if it's not something is not easy to "fix", it does not mean it should be ignored.
Updated
DakuTree said:
So I've started running what could essentially be called an "upload bot". This can auto-tag, rate, and upload faster that any user. This fixes the problem by making it impossible to happen in the first place.
Albeit the tagging isn't perfect, but considering "tagging on upload" doesn't matter as long as it gets tagged in the end, that isn't exactly a problem.
I've actually had this thing for a few months. Have been hoping you know, uploaders would get better. Apparently not.
...Wait, you're running a bot that makes all other uploaders obsolete? Considering the mess your artist-URL-grabber thing made 2+ months ago, I'm worried that you're being serious.
DakuTree said: You make it impossible to watch pixiv, people then have to look. We get more "rare" content as a result.
No, watching pixiv is impossible thanks to you and your bot. Just FYI.
Besides, I suspected you of a usage of a bot much earlier. Your problem for being consciously unfair to others on a far greater degree than all the active "little taggers on upload". And to add, it will achieve the absolute opposite of what you want. Everyone who will want to grab something from front pixiv, WILL upload it with minimum tags, just to skip your shit.
Updated
RaisingK said:
...Wait, you're running a bot that makes all other uploaders obsolete?
I'm not. I'm running a script which makes it impossible to actively watch new images. That is where this problem happens.
The sheer reason I have been watching pixiv as much as I do is to push users to actually look for images. We have how many people looking for images exactly, and how many people that just sit and stalk pixiv for "first"?
We do not need anyone actually actively watching pixiv.
Considering the mess your artist-URL-grabber thing made 2+ months ago, I'm worried that you're being serious.
I've already had quite a while to test this thing on a virtualboxed danbooru. Uploading isn't exactly something that can cause problems.
No, watching pixiv is impossible thanks to you and your bot. Just FYI.
So how exactly do we get users to look for useful content?
You can't just have a thread saying "look for more useful content". As long as an easy method of finding content exists, people will use that.
Watching pixiv is essentially something a bot could do. I do not understand under any sense how we need more than one bot/person watching pixiv. Pixiv is something that can essentially be automated, so why can't it be?
Wypatroszony said:
How about you search for said useful content yourself?No, revisions don't count as useful content.
Last time I checked, I am one of the few users who actually does look for this kind of content, we have what 3? other users that ever bring "useful" content from outside the site.
Yet we have a lot more users who I never see uploading anything outside anything posted in the past few days on pixiv.
This isn't something that takes long, or is hard to do. If you can push users away from easy content and into doing stuff like browsing unchecked blogs, twitter etc., we would have more actually useful content.
Or do you care about your numbers too much?
What numbers?
DakuTree said:
I've already had quite a while to test this thing on a virtualboxed danbooru. Uploading isn't exactly something that can cause problems.
Unless it's automated uploading. How do you automate deciding what's worth uploading? Rating and tags are problematic, too. I'm still confused as to what exactly this script is doing.
RaisingK said:
Unless it's automated uploading. How do you automate deciding what's worth uploading? Rating and tags are problematic, too. I'm still confused as to what exactly this script is doing.
At the moment I've completely limited it to artists I've seen upload nothing that isn't exactly "danbooru-quality" (which isn't exactly loads). I could probably run this on a seperate account to push images into the mod queue to possibly allow more artists, unsure though.
Tagging is basically pixiv > danbooru tag translations I've managed to scrape from various userscripts and such. Several related tags are also added if they have been added to the DB (copyright & anything else that nearly always exists).
Rating defaults to Q, E if "R-18" tag exists.