Notice: My personal stance on AI generated artwork. Retweet and share if you agree. Let us discuss, and not immediately scream bloody murder.

Now Viewing: Facesitting vs. Sitting_on_face (again)
Keep it civil, do not flame or bait other users. If you notice anything illegal or inappropriate being discussed, contact an administrator or moderator.

Super_Strawberry - Group: Member - Total Posts: 3
user_avatar
Facesitting vs. Sitting_on_face (again)
Posted on: 02/25/12 11:35AM

Recently the Facesitting tag was rolled into sitting_on_face, due to them redundantly rivaling one another. I agree with merging the tags, but I think they should have been merged the other way around: Sitting_on_face into Facesitting. See, Facesitting is a common term on other websites, it's actually a name for the act. People know it, people search for it, etc. Sitting_on_face is a booru word-combining function making a tag most people haven't heard of. Case in point, the Facesitting tag has been gaining images again, and faster than it's rival tag. Seems like people were confused when the Facesitting tag went away, and started tagging images with it again. Can like, someone in charge fix all this (merge the tags the RIGHT way)?



Jerl - Group: The Real Administrator - Total Posts: 6711
user_avatar
Posted on: 02/25/12 11:43AM

Mass tagging started to retag all images tagged facesitting into sitting_on_face for the correct tag alias. Please stand by...

In the meanwhile, I'll explain to you exactly why we use the tag the way we do.

You see, the vast majority of images on Gelbooru come in straight from Danbooru. Approximately 75%, actually. And, with their images, we also get their image translations and, more importantly, their tags.

All of the tags that are applied on Danbooru are also applied on Gelbooru. While Danbooru has an aliasing system that is relatively easy to keep up-to-date, Gelbooru's alias system has some problems. For example, images that come in from the danbooru bot, which directly inserts them into the database instead of using the normal post upload procedure, doesn't directly get affected by it, and the tag remains as it was. Additionally, when new aliases are added, our search listing must be rebuilt for it to work. On top of this, aliases cannot be broken in any easy manner, and only two people on the site are currently capable of adding aliases. For these reasons combined, we generally do not use our alias system unless absolutely necessary, and instead perform soft aliases with the moderator-only mass tagging tool.

Now, to start with, lots of the tags on Gelbooru started here before Danbooru had any alias set up for them. Additionally, Gelbooru has been around for long enough for many of our images not to have been affected at all by Danbooru's aliases. I don't know when Danbooru implemented aliases, but I do know that I have seen quite a few tags on images that came in from Danbooru that were not aliased on Danbooru.

However, the danbooru scraper is still constantly running, pulling posts straight from Danbooru over to Gelbooru, tags and all. For this reason, we cannot just use soft aliases to fix the new images to fit the new tags. First of all, often the new tag has many more images than the old tag did. Other than that, though, the images with the new tag would just keep coming in even if we soft-aliased them with the mass tagger, which would require us to run thousands of mass tag edits every day. Running the mass tagger once is probably equivalent to about three or more people browsing the site normally. Depending on the tag, this may last from 10 minutes for a tag with about 200 posts to eight hours for a tag with ten thousand. Performing mass tag edits on this scale would eat up all of the site's resources and bandwidth, and would cause more trouble than it's worth.

So, instead, we soft-alias them to what Danbooru uses. This means that we only need to re-perform the mass edits very infrequently. The number of edits that must be performed in these mass edits is usually very small, even if we only perform them every few weeks. For example, I frequently mass implicate animated_gif -> animated. I've let it sit for several weeks and only found about 10 images missing the tag. This is much more efficient for tagging purposes.

So, when Danbooru sets tagging policy, we are almost always locked in to following it ourselves. There are some exceptions, but generally our tagging policy mirrors theirs.

You'll have to look to Danbooru for the reasoning behind their tag aliases. A quick search of their forums found me this thread:
danbooru.me/forum/show/38771

I barely looked at the thread, though, so I don't know how relevant it is. You may also wish to look at their tag wiki.



Anti_Gendou - Group: Moderator - Total Posts: 4370
user_avatar
Posted on: 02/25/12 12:11PM

IRC complaining of text wall.

Mod respond with text wall.

At least it got the job done.



interneth86machine - Group: Member - Total Posts: 192
user_avatar
Posted on: 02/25/12 01:53PM

Jerl said:
Mass tagging started to retag all images tagged facesitting into sitting_on_face for the correct tag alias. Please stand by...

In the meanwhile, I'll explain to you exactly why we use the tag the way we do.

You see, the vast majority of images on Gelbooru come in straight from Danbooru. Approximately 75%, actually. And, with their images, we also get their image translations and, more importantly, their tags.

All of the tags that are applied on Danbooru are also applied on Gelbooru. While Danbooru has an aliasing system that is relatively easy to keep up-to-date, Gelbooru's alias system has some problems. For example, images that come in from the danbooru bot, which directly inserts them into the database instead of using the normal post upload procedure, doesn't directly get affected by it, and the tag remains as it was. Additionally, when new aliases are added, our search listing must be rebuilt for it to work. On top of this, aliases cannot be broken in any easy manner, and only two people on the site are currently capable of adding aliases. For these reasons combined, we generally do not use our alias system unless absolutely necessary, and instead perform soft aliases with the moderator-only mass tagging tool.

Now, to start with, lots of the tags on Gelbooru started here before Danbooru had any alias set up for them. Additionally, Gelbooru has been around for long enough for many of our images not to have been affected at all by Danbooru's aliases. I don't know when Danbooru implemented aliases, but I do know that I have seen quite a few tags on images that came in from Danbooru that were not aliased on Danbooru.

However, the danbooru scraper is still constantly running, pulling posts straight from Danbooru over to Gelbooru, tags and all. For this reason, we cannot just use soft aliases to fix the new images to fit the new tags. First of all, often the new tag has many more images than the old tag did. Other than that, though, the images with the new tag would just keep coming in even if we soft-aliased them with the mass tagger, which would require us to run thousands of mass tag edits every day. Running the mass tagger once is probably equivalent to about three or more people browsing the site normally. Depending on the tag, this may last from 10 minutes for a tag with about 200 posts to eight hours for a tag with ten thousand. Performing mass tag edits on this scale would eat up all of the site's resources and bandwidth, and would cause more trouble than it's worth.

So, instead, we soft-alias them to what Danbooru uses. This means that we only need to re-perform the mass edits very infrequently. The number of edits that must be performed in these mass edits is usually very small, even if we only perform them every few weeks. For example, I frequently mass implicate animated_gif -> animated. I've let it sit for several weeks and only found about 10 images missing the tag. This is much more efficient for tagging purposes.

So, when Danbooru sets tagging policy, we are almost always locked in to following it ourselves. There are some exceptions, but generally our tagging policy mirrors theirs.

You'll have to look to Danbooru for the reasoning behind their tag aliases. A quick search of their forums found me this thread:
danbooru.me/forum/show/38771

I barely looked at the thread, though, so I don't know how relevant it is. You may also wish to look at their tag wiki.


TL;DR we use danbo's tags/rules, and our alias system sucks dick.



Super_Strawberry - Group: Member - Total Posts: 3
user_avatar
Posted on: 02/28/12 02:06AM

Not the response I was expecting, but I did read the wall of text, and I'm pretty sure I understand now. The two tags don't seem aliased here while on Danbooru they are, so Danny prolly aliased them later on, and the tags here are getting moved from facesitting to sitting_on_face to reflect Danbooru's alias, amirite? In this case yeah my complaint's not really an issue, so thanks!



Daijin - Group: Retired Staff - Total Posts: 1153
user_avatar
Posted on: 02/28/12 08:32PM

interneth86machine said:


TL;DR we use danbo's tags/rules


Not always. There are times we use a tag that Danbooru doesn't. A good example of pokemon_bw and pokemon_black_and_white.

Danbooru uses pokemon_bw and aliases pokemon_black_and white into it; the amount of images on Dan with the pokemon_bw tag cover 155 pages.

We use pokemon_black_and_white over the other other and that has 152 pages to it, pokemon_bw is only on 16 images here. I'm about to wipe those 16 tags. I'm also going to wipe the pokemon_black_2_and_white_2 tag someone is starting to put on the images.

EDIT: Actually as I'm cleaning these tags, I'm finding that overall we choose to write out the full title of the pokemon game while Dan uses abbreviations. The other possibility is that it was Dan who changed to a new pokemon tagging policy after we adopted their old one.

It may be a good idea to use the tags Dan uses for this because they cover all the games much better, i.e. pokemon_rgby is for all images that cover pokemon, red, blue green and yellow, as well as the remakes of firered and leafgreen which cuts out a ton of tags used for those images.

The problem is the amount of work that would go into it.



add_replyAdd Reply


1