[X] My Assistant

We will not provide categorization for irrelevant content (like non-H western stuff) since it'll implicitly encourage people to post it. We also want the added categorization to be backwards-compatible with existing categories and tags, since there simply isn't enough manpower to recategorize everything. This is mostly about splitting the "Western" and "Misc" categories into more manageable chunks, a more fundamental revamp is off the table.

We could consider adding new categories for Novels and possibly also Artbooks, though I'm not really sure if the latter is sufficiently different from Artist CG to deserve its own category?

Novels is a pain because the tag is used by not only exclusive novels but also by works that have a significant novel section within them. So that would require some man power to confirm which are pure novels and which are 50/50 splits and which are an anthology work that features a novel within it.

As for art books there has been an ongoing discussion thread about how to categorise published (printed) art books. As it is due to some rule changes published art books should end up in the image set category due to the lack of narrative. However most people (citation needed) do not like this. A category for this kind of content could fix this issue.

This post has been edited by Cipher-kun: Nov 20 2023, 18:10

We will not provide categorization for irrelevant content (like non-H western stuff) since it'll implicitly encourage people to post it.

I still think this is a missed opportunity but I understand the reasoning.

We also want the added categorization to be backwards-compatible with existing categories and tags, since there simply isn't enough manpower to recategorize everything.

Is there a plan on what happens to the stuff that cannot be converted properly (e.g. western galleries with no comic / western cg / western imageset or western non-h tag) and would this be the proper thread to discuss that?

We could consider adding new categories for Novels and possibly also Artbooks, though I'm not really sure if the latter is sufficiently different from Artist CG to deserve its own category?

By our current definitions, which require Artist CGs to "depict a sequence of events (or be a variant set)", most pornographic artbooks would go to Imageset. That is not enforced at all though, so most of them currently go to doujin. This was discussed here but it didn't lead to actual changes yet; maybe this is a good opportunity to fix it.

Is there a plan on what happens to the stuff that cannot be converted properly (e.g. western galleries with no comic / western cg / western imageset or western non-h tag) and would this be the proper thread to discuss that?

"western non-h" is still Misc. But otherwise, they will probably land in the Western Image Set category until reclassed. And sure, the discussion is relevant for deciding classification so it's probably fine to have it here.

By our current definitions, which require Artist CGs to "depict a sequence of events (or be a variant set)", most pornographic artbooks would go to Imageset. That is not enforced at all though, so most of them currently go to doujin. This was discussed here but it didn't lead to actual changes yet; maybe this is a good opportunity to fix it.

Yeah, maybe. On the other hand, I can see overlaps between applicable categories potentially being a problem, but it might be better than the status quo.

As for subcategorization, I guess it would make sense to have all four for Artbooks. Not sure about Novels though, would it make sense with "3D Novel" for a novel with 3D illustrations, for example?

On the other hand, I can see overlaps between applicable categories potentially being a problem, but it might be better than the status quo.

As for subcategorization, I guess it would make sense to have all four for Artbooks. Not sure about Novels though, would it make sense with "3D Novel" for a novel with 3D illustrations, for example?

For artbooks I also think the overlap will be problematic. I'd rather see a general rule like "artbooks go to (whatever sub-category) Artist CG" or something like that.
I don't care much about novels but I feel like having a separate category would be overkill.

"western non-h" is still Misc. But otherwise, they will probably land in the Western Image Set category until reclassed. And sure, the discussion is relevant for deciding classification so it's probably fine to have it here.

Great, then here are my thoughts on the most practicable way to do split western / misc. I tried to keep it short, but I thought about this a lot! At the end of the day, it boils down to how much work we want to do to increase accuracy, and this is the degree to which I'd take it.

Most importantly, I'd argue for only doing a default conversion when that can be done with sufficient accuracy. For the quoted example, I think defaulting untagged western galleries to Western Image Set is suboptimal because:
- From my experience, this won't have high accuracy (probably no more than 75%).
- Depending on how exactly it's implemented, there may not be an easy way to distinguish between "Western Image Set that was actually tagged as such" and "Western Image Set that was defaulted to the category because we didn't know any better", making after the split cleanup difficult.
- Depending on how exactly it's implemented, reclassing a gallery after the split may require much more effort than just applying the correct tag before the split.

Instead, I'd have them placed in the (temporary) category "Western Uncategorized", which would not be available for new uploads. Galleries within this category don't require a 100 MP tags to be reclassed (in practice, you could have the autotagger boost reclass tags in this category to 100 MP whenever there is one). To prevent abuse, you could have this happen at 20 MP or something like that, so that a single experienced user can get a gallery out of "Western Uncategorized".

You already talked about lack of manpower and that's true to a degree. It looks like out of ~170,000 western galleries, ~35,000 are not tagged any of comic / western cg / western imageset. But consider that adding the correct tag may require very little effort. You have to:
- Determine the format (Comic / Image Set / CG). This can be done with high certainty in just a few seconds in the vast majority of cases.
- (nice to have) Confirm that the gallery is indeed pornographic. This is usually straightforward, but can be tricky in cases of large galleries wherein the majority is SFW.
- (nice to have) Confirm that the artist is indeed western. This can be anywhere from easy to impossible.
I added (nice to have) to the last to points because if they are wrong, that means that the gallery was already in the wrong category to begin with. The point is, if all we want to do is determine the format, I'm confident that this can be done quickly with very high accuracy.

Another way to look at it: The western galleries lacking "comic" / "western imageset" / "western cg" either have some ambiguous quality that prevented people from adding those tags or they just have never been looked at with those tags in mind. I'm convinced that well over 10,000 galleries fall into the latter category and that many of those could be quickly classified at a glance with much higher accuracy than any defaulting could achieve. As for actually doing it, I'm sure some vigilantees would participate (maybe more if there is some incentive beyond toplist points) and that "Western Uncategorized" would be mostly cleared within a few months.

A similar train of thought would apply to "3D Uncategorized", but the number of untagged galleries there sadly is much higher (around 90,000). Language tags might be useful to narrow down the galleries to tackle manually. Also in my counts I didn't respect expunged galleries, but I don't think they're worth any extra effort.


Now for the specific conversions. I'm going to assume that galleries have no conflicting tags (e.g. no galleries tagged both "comic" and "western cg"). I don't think there are that many with conflicting tags and it'd be manageable to clean them up manually before doing the split. Keep in mind that the tags in question won't have 100 MP on almost all older galleries, so the autotagger may need to boost them before conversion.
- Western galleries tagged "comic" or "western imageset" can be categorized without ambiguity.
- Western galleries tagged "western cg" could be either western Artist CG or western Game CG. From my experience they're western Artist CGs in the vast majority (well over >95%) of cases. So defaulting them to western Artist CG makes sense.
- Western galleries with no further specifying tag should go to "Western Uncategorized". As described in above, they either haven't been looked at or are ambiguous cases, but they cannot be defaulted to anything with good accuracy.
- Misc galleries tagged "3d imageset" could be either 3d Image Set or 3d Non-H (aka Misc). From my experience they're pornographic in the vast majority (~95%) of cases. So defaulting them to 3d Image Set makes sense.
- Misc galleries tagged "3d" but not "3d imageset" should go to "3d Uncategorized" as described above. From my experience they cannot be defaulted to anything with good accuracy (wild guess would be that ~70% are Artist CG, but many actually are untagged Image Sets).
- Misc galleries tagged "ai generated" will be AI imagesets in the majority of cases (~90%? pretty hard to make a disctinction betwee Artist CG and Image Set in some cases). Many of those that aren't could be identified by the presence of a language tag. From my experience AI Game CG / non-h is rare enough to justify defaulting the rest to AI Imageset. Because of this, "AI Uncategorized" would not be necessary.
- Misc galleries tagged neither "3d" nor "ai generated" stay in misc.

For artbooks I also think the overlap will be problematic. I'd rather see a general rule like "artbooks go to (whatever sub-category) Artist CG" or something like that.

Not sure why this isn't the case already, but this works for me.

I don't care much about novels but I feel like having a separate category would be overkill.

I don't really care about novels either, so I'm fine with keeping them in whichever other category best fits the particular work. I'm sure there would be obnoxious ambiguity with how much text you would need to make something a "novel" anyway.

Instead, I'd have them placed in the (temporary) category "Western Uncategorized", which would not be available for new uploads. Galleries within this category don't require a 100 MP tags to be reclassed (in practice, you could have the autotagger boost reclass tags in this category to 100 MP whenever there is one). To prevent abuse, you could have this happen at 20 MP or something like that, so that a single experienced user can get a gallery out of "Western Uncategorized".
...
Now for the specific conversions. I'm going to assume that galleries have no conflicting tags (e.g. no galleries tagged both "comic" and "western cg"). I don't think there are that many with conflicting tags and it'd be manageable to clean them up manually before doing the split. Keep in mind that the tags in question won't have 100 MP on almost all older galleries, so the autotagger may need to boost them before conversion.
- Western galleries tagged "comic" or "western imageset" can be categorized without ambiguity.
- Western galleries tagged "western cg" could be either western Artist CG or western Game CG. From my experience they're western Artist CGs in the vast majority (well over >95%) of cases. So defaulting them to western Artist CG makes sense.
- Western galleries with no further specifying tag should go to "Western Uncategorized". As described in above, they either haven't been looked at or are ambiguous cases, but they cannot be defaulted to anything with good accuracy.
- Misc galleries tagged "3d imageset" could be either 3d Image Set or 3d Non-H (aka Misc). From my experience they're pornographic in the vast majority (~95%) of cases. So defaulting them to 3d Image Set makes sense.
- Misc galleries tagged "3d" but not "3d imageset" should go to "3d Uncategorized" as described above. From my experience they cannot be defaulted to anything with good accuracy (wild guess would be that ~70% are Artist CG, but many actually are untagged Image Sets).
- Misc galleries tagged "ai generated" will be AI imagesets in the majority of cases (~90%? pretty hard to make a disctinction betwee Artist CG and Image Set in some cases). Many of those that aren't could be identified by the presence of a language tag. From my experience AI Game CG / non-h is rare enough to justify defaulting the rest to AI Imageset. Because of this, "AI Uncategorized" would not be necessary.
- Misc galleries tagged neither "3d" nor "ai generated" stay in misc.

Assuming that we're limiting automatic reclassing to content in today's "Misc" and "Western" categories, which I think will be the case anyway, I suppose adding temporary Uncategorized/Western and Uncategorized/3D would be a good idea for when no relevant tags are found.

When it sets the new category, it would do so at the "set by uploader" level rather than the "set by a tag" level. The old pseudo-category tags would likely be redundant after the switchover, and placing new reclass tags on it would just make it unnecessarily difficult to fix any automatic misclassifications. As such, t shouldn't be harder to reclass it afterwards compared to what it is today.

If a gallery is ambiguously tagged, it would generally go by the stronger tag, unless they are close in power, in which case we can just throw it into Uncategorized.

Your conversion rules look good to me, so we'll probably go with those unless someone spots any issues with them. I'm sure there will still be edge cases needing manual cleanup, but we won't avoid that in any case.

We will not provide categorization for irrelevant content (like non-H western stuff) since it'll implicitly encourage people to post it.

So just for the record, AI slob is officially more encouraged than non-nude fetish content (definitely more nuance to whether something is intended as porn than sex/nudity alone, when you got fetishes like BDSM, Bondage, Corruption, Diapers, Guro, Mind Control, Scat and Transformation. an objective category split cannot handle this nuance but that's not as big of an issue if it barely makes a difference for searchability) now?

If there's a different reason it would be good to have that on the record as well. I know there has been a lot of demand for AI to be banished to its own hole but from what I've understood that's often more of a "misc isn't bad enough for it, I don't want it drowning out my 3d and western non-h but I have a (shitty) reason for not just filtering the AI tag" and less of an "I really want this content to be up there with the non-AI stuff".

If you shift the narrative from "non-h=non-nude" to "non-h=garbage/shouldn't have been posted" you'll probably just see an increase of people going out of their way to classify their uploads as "not garbage" regardless of more objective criteria.

Should also be noted that the Asian/Western split isn't as clear as people assume it to be, there have been plenty of instances where I've run into artists where either a big chunk of the content was misclassified when I looked into it (often non-Japanese Asians with more Western artstyles), or it came down to guesswork because the external information for artist nationality was not available to me.

To me the most reasonable argument against categorizing non-H content any further (not just the Western stuff people prefer to shit on when complaining that the non-vanilla content isn't nude enough for vanilla tastes, I'm sure people are lining up to put those words in my mouth but according to OP the Asian non-H stuff also isn't shown to receive any improvements compared to what it has now) is just the complexity of it. You can categorize content based on 3 variables:
Artstyle (in this case using Artist nationality as the closest approximation even though it has its imperfections): Asian/Western/3d/AI/photos/mixed
Format: Comic/CG/Imageset/Novel(/Game CG)
whether it is Pornographic (once again using an approximation to get around nuance/subjectivity, in this case sticking to nudity or acts we all agree are sexual): yes/no (though the Nudity Only and No Penetration tags do effectively serve as extensions to this split)
And Doujinshi/Manga and Cosplay/AsianPorn are splits made for more specific criteria outside those 3 general ones.
Right now Asian Pornographic content has categories and the rest is, at most, categorized on the first variable, but we do have a bunch of technical tags that basically fill the same role while keeping the category selectors compact.
The new system aims mostly sticks to those first 2 variables but adds proper subcategories for a lot of combinations currently only handled by technical tags that already exist and technical tags that would surely have been suggested in the future without this revision.
Fully separating things out based on that 3rd variable would be an exponential increase in categories where demand is admittedly modest, but having subcategories for at least one of those other variables within non-H would be neat, and I'd still prefer to have technical tags that can make the more specific stuff a bit easier.

Not sure why this isn't the case already

Because some people have been very stubborn about making "physically sold=doujinshi/manga" (which isn't even supported by the rules (anymore), just their opinion of how it should work) overrule the format/layout-focused categorisation we'd otherwise work with when differentiating between doujinshi/manga/comic and CG, or narrative-focused split that was needed to Imageset apart from either when "distributed together" became less reliable of a factor.

I don't really care about novels either, so I'm fine with keeping them in whichever other category best fits the particular work. I'm sure there would be obnoxious ambiguity with how much text you would need to make something a "novel" anyway.

While I fully support keeping Novel content where the text-only pages are only a chunk of the content in whatever category they otherwise would have had, but it leaves a bit of a weird gap:
http://ehentaihip.com/g/2546420/55d5c62877/
You sorta have to start counting covers as CG pages (which normally wouldn't count for anything with gallery-wide implications but in this case only really makes a difference on galleries that have a perfect 50/50 split between CG and panelled or no images for either) to get the gallery to a single page with an illustration or just roll with the "if it was sold physically it's automatically doujinshi/manga" mindset to find a home for those if they don't get either their own category or a modification to the existing categorisation flowchart that does give them a home somewhere.



On another note, could you elaborate on how this new system is actually going to look
Are we going to have less category selectors with subcategories effectively working the same way they already do within Western?
Would the 9+ categories mentioned in the OP be listed the same way categories currently are with some new system (dropdowns sound intuitive but sound like torture for mobile users) for further specifying subcategory?
Would it look like the current layout but with every subcategory being its own box (essentially making it completely pointless to even seperate Doujinshi, Manga)?
Some other layout that I didn't think of here?

This post has been edited by Glovelove.: Nov 27 2023, 04:51

So just for the record, AI slob is officially more encouraged than non-nude fetish content (definitely more nuance to whether something is intended as porn than sex/nudity alone, when you got fetishes like BDSM, Bondage, Corruption, Diapers, Guro, Mind Control, Scat and Transformation. an objective category split cannot handle this nuance but that's not as big of an issue if it barely makes a difference for searchability) now?

It's still porn. This is a porn site. People still post all kinds of shit here that isn't porn (or even borderline fetish nor relevant for the doujin scene), and we don't want to encourage posting not porn.

Fully separating things out based on that 3rd variable would be an exponential increase in categories where demand is admittedly modest, but having subcategories for at least one of those other variables within non-H would be neat, and I'd still prefer to have technical tags that can make the more specific stuff a bit easier.

Like I mentioned earlier, there are problems with representing additional dimensions of categories visually that makes it a no-go.

While I fully support keeping Novel content where the text-only pages are only a chunk of the content in whatever category they otherwise would have had, but it leaves a bit of a weird gap:
http://ehentaihip.com/g/2546420/55d5c62877/

We can't really design categories to handle every single edge case, there will always be examples of stuff that doesn't really fit anywhere.

On another note, could you elaborate on how this new system is actually going to look
Are we going to have less category selectors with subcategories effectively working the same way they already do within Western?
Would the 9+ categories mentioned in the OP be listed the same way categories currently are with some new system (dropdowns sound intuitive but sound like torture for mobile users) for further specifying subcategory?
Would it look like the current layout but with every subcategory being its own box (essentially making it completely pointless to even seperate Doujinshi, Manga)?
Some other layout that I didn't think of here?

Back-of-the-napkin level stuff here, but basically the category selector will be a 2D matrix looking something like:

CODE

               Native   Western   3D   AI

Doujinshi         X
Manga             X
Comics                     X       X    X
Non-H             X
Cosplay           X
Game CG           X        X       X    X
Artist CG         X        X       X    X
Image Set         X        X       X    X
Misc              X
Uncategorized              X       X

with some finagling so you can easily select/unselect/onlyselect an entire row/column of categories.

The buttons themselves will probably just be the same background color for the main type with a different border per subcategory.

If a gallery is ambiguously tagged, it would generally go by the stronger tag, unless they are close in power, in which case we can just throw it into Uncategorized.

Your conversion rules look good to me, so we'll probably go with those unless someone spots any issues with them.

Glad you like the conversion rules. Maybe we should create some of the missing conversion tags (most notable would probably be 3d cg and 3d comic) before doing the split? I don't want to rush it in case changes are made, but on the other hand the earlier we have them, the smaller the temporary Uncategorized dumps are going to be.

Maybe we should create some of the missing conversion tags (most notable would probably be 3d cg and 3d comic) before doing the split?

"3d" is a tag and "comic" is a tag, so I'm not sure why you would need to make a new one instead of just redefining "comic" to be applicable?

Thank you for giving a proper response instead of just escalating frustrations.

People still post all kinds of shit here that isn't porn (or even borderline fetish nor relevant for the doujin scene), and we don't want to encourage posting not porn.

Not going to defend the stuff that isn't even borderline fetish content (and unlike some others here I don't hold a double standard based on where it comes from, I don't change my opinion on content that has absolutely no merit on a porn site the moment someone points out the artist is actually Korean and merely misclassed based on artstyle), just very frustrated that the non-nude fetish content always gets caught in the crossfire when measures are taken against that stuff

Like I mentioned earlier, there are problems with representing additional dimensions of categories visually that makes it a no-go.

Back-of-the-napkin level stuff here, but basically the category selector will be a 2D matrix looking something like:

CODE

               Native   Western   3D   AI

Doujinshi         X
Manga             X
Comics                     X       X    X
Non-H             X
Cosplay           X
Game CG           X        X       X    X
Artist CG         X        X       X    X
Image Set         X        X       X    X
Misc              X
Uncategorized              X       X

with some finagling so you can easily select/unselect/onlyselect an entire row/column of categories.

The buttons themselves will probably just be the same background color for the main type with a different border per subcategory.

Yeah with a layout like that adding in a 3rd dimension wouldn't work, but I think some proper technical tag coverage should be plenty anyway. It feels off to me to keep non-h split this way instead of giving it the same subcategories and finding better means to discourage non-fetish, non-nude content, but it would go a long way if we could just get the Non-h Imageset tag to apply to non-H imagesets of all (sub)categories; If people only want to see Asian non-H imagesets they could still do so without any additional search terms by just using the category toggles.

This post has been edited by Glovelove.: Nov 27 2023, 21:16

My only comment related to this proposal now is that western and non-western should be based on art style and not based on whether the artist is Asian or not.

There was a vote on that some years ago, to go by "style" rather than nationality, and it was overwhelmingly rejected.

Ugh... Because it's been years, I'd try again now that this rework is in progress. If still rejected... So be it.


Edit:

It was nuked 10+ years ago. It'll probably still be accessible, there just won't be a visible UI for it.

Even if the UI doesn't help as much anymore, I think there could be a way to reach them. Maybe just a keyword of sorts or sth even if the button is gone?

This post has been edited by Jay Low: Dec 5 2023, 11:17

"3d" is a tag and "comic" is a tag, so I'm not sure why you would need to make a new one instead of just redefining "comic" to be applicable?

Good point, that's way easier.

My only comment related to this proposal now is that western and non-western should be based on art style and not based on whether the artist is Asian or not.
Ugh... Because it's been years, I'd try again now that this rework is in progress. If still rejected... So be it.

This may at first glance seem like an easy thing to do but it 100% is not and could invalidate a decade worth of cleanup.

My only comment related to this proposal now is that western and non-western should be based on art style and not based on whether the artist is Asian or not.
Ugh... Because it's been years, I'd try again now that this rework is in progress. If still rejected... So be it.

This may at first glance seem like an easy thing to do but it 100% is not and could invalidate a decade worth of cleanup.

It's even worse than just that. It also moves away from something that is objective, location, to something that is subjective, identification of style. So yeah what peterson said. It's bad.

This post has been edited by Cipher-kun: Dec 5 2023, 13:26

Assuming we managed to find the dozens of people we'd need to reliably make judgement calls when it'd come up all the time just to handle the workload, it wouldn't be a better solution to be worth it.

Lots of people don't like western styled stuff in doujins, but if you had a gallery that looked like manga, made by a:harry jones, in English with no translated tag, there'd be just as many people pissed at that being in doujinshi too.

My only comment related to this proposal now is that western and non-western should be based on art style and not based on whether the artist is Asian or not.

horrible idea, "art style" is super subjective. The current method might place a couple of artists with western styles into doujinshi/manga, and the other way around too, but that's a very small amount of works. The way it is right now works really well, and most importantly isn't adding a big workload on vigs, like art style would.

OK. I get it. Thanks

AI should be its own top-level category.
Whatever is made with the AI, how that is categorized further is not something I have much of an opinion on, but I strongly feel that all AI content should be neatly wrapped up in one big package before any further distinctions are made.

For what it's worth, I don't see any need to categorize by country of origin within AI. For one, the author should be considered the computer chips, so they are all Asian anyway.
Secondly, these is absolutely no way to distinguish an AI-generated image generated in Japan from an AI-generated image generated in Germany by looking at the end result. Computer chips don't have national cultures or traditional local art styles or any of the other root causes why Western art looks different from Japanese art, so the country of origin is a completely trivial bit of information that serves no function.
Furthermore, if anything were to be used for such a disctinction, it should be the country of origin for the machine learning algorithm or the country of origin for the data set used to train the algorithm, as those are more indicative of the country of origin for the images the AI spat out, rather than which country the person who happened to be the one to press the button to generate any given set of images was located in. Keeping track of that information is obviously an impossible endeavour.

AI should be its own top-level category.

Would be harder to filter it by format if you do that. For example, someone might not want ai imageset, but might be specifically looking for an ai game cg, or not mind ai comics. What benefit would there be to it being a main category instead of a sub category?

-snip-

I don't think there's any plans to distinguish it by nationality anyway, unless I missed something, not really been keeping up with the thread. If it's a subcategory, it could just be set alongside the main category for it's format (comic, imageset, artist cg, game cg) without really worrying about where it came from.

Right now the plans are western for Comics, Artist CG, Game CG and Imagesets. AI being sorted by western/asian was rightfully never discussed, i hope.
See: /index.php?act=f...amp;pid=6398210

This post has been edited by -terry-: Dec 19 2023, 02:39

My little AI is called Jimmy. He's a cute little Caucasian-themed bot who loves to draw mixed style comics.