I've been working on a tool that can suggest tags on posts based on a set of rules (unsurprisingly most posts have at least 1 problem) and I'm wondering how many "rules" of tagging I'm missing - here's what I've got.

Always

• Posts should contain male, female, intersex, ambiguous_gender or zero_pictured.
• Posts should contain solo, duo, group or zero_pictured.
• Posts should contain either digital_media_(artwork), traditional_media_(artwork) or photography_(artwork).
• Posts should not have any tags in the invalid category, or ending with (disambiguation).
• If a safe/questionable post contains explicit tags (pussy, penis, sex, masturbation, cum, etc) the rating should either be changed to explicit or the tags be removed. (need a full list of what's regarded as explicit)
• Artist tag count should always be at least 1 (including unknown_artist or anonymous_artist)

Usually

• Posts tagged with penis should also contain either: flaccid, half-erect, erection.
• Posts with talking_to_viewer should also contain either text or sound?
• Posts should be tagged with either clothed, mostly_nude or nude. (unless the image is feral solo)
• Posts should contain inside, outside or simple_background. (sometimes the background seems too complex for simple, but the location is too ambiguous)
• Posts with hair should have a <color>_hair tag. Same goes for fur and other stuff I guess. (except when monochrome?)
• herm -pussy is probably actually a gynomorph.

If you've got anything else, please don't hesitate to share :)

faucet said:

• Posts should be tagged with either clothed, mostly_nude or nude. (do unclothed ferals get tagged as nude?)

They shouldn't be. Ferals are considered to be nude by default, and clothed_feral is used to denote one where clothing.

faucet said:

• A herm with rating:s or rating:q is probably actually a gynomorph.

More succinctly, I think, a herm without a visible pussy is probably a gynomorph. And a pussy is automatically Explicit (same with penis, cloaca, balls, anus, etc).

This interests me too.
I have my own tag suggester tool in re621, although it's pretty basic. The default rules can be viewed here.

I have a few recommendations.

faucet said: Posts should contain solo, duo, group or zero_pictured.

If a post contains multiple of these, it's either a mistake, or the post may be a comic (multiple_images / multiple_scenes).
Also true for *_focus tags. Those should be paired with standard character count tags, but usually not the same one (aka not both solo and solo_focus).

faucet said: Posts should not have any tags in the invalid category, or ending with (disambiguation).

Pretty much, yes.
If your tool also analyzes the tags that the user adds (and not just the ones returned by the site), you should resolve aliases.
Otherwise, the tool could miss a tag that is aliased to a *_(disambiguation) tag.

faucet said:
If a safe/questionable post contains explicit tags (pussy, penis, sex, masturbation, cum, etc) the rating should either be changed to explicit or the tags be removed. (need a full list of what's regarded as explicit)

Stuff stated here would be a good place to start.

faucet said: Artist tag count should always be at least 1 (including unknown_artist or anonymous_artist)

Don't forget to exclude the warning tags in that count: avoid_posting, conditional_dnp, sound_warning, epilepsy_warning, check_tag_type.

faucet said: Posts with talking_to_viewer should also contain either text or sound.

There was actually a bit of debate about that recently: https://e621.net/forum_topics/33133
The talking_to_viewer tag implies dialogue, which indicates that it should only be used when there's text in the image, not for voice acting. Plus, there are sometimes cases where the character is clearly talking, but there is neither text nor sound.

... also, sound should probably not be suggested for static images.

faucet said: Posts should be tagged with either clothed, mostly_nude or nude. (do unclothed ferals get tagged as nude?)

IIRC, the consensus was that naked ferals don't have to be tagged nude. It's only for characters that are expected to wear clothing normally.

faucet said: Posts should contain inside, outside or simple_background.

I think it's more complicated than these 3. Check tag_group:background for more.

faucet said: Posts with hair should have a <color>_hair tag. Same goes for fur and other stuff I guess. (except when monochrome?)

Or sketch. There may be more exceptions.

bitwolfy said:
Or sketch. There may be more exceptions.

Line_art should probably also count. And honestly, I feel a bit wrong adding any color tag to monochrome images, beyond general ones like blue_theme or black_and_red, etc, or one for the background color, but not specific parts of characters or anatomy. Partially_colored also doesn't seem like the non-colored parts should be tagged with colors.

faucet said:

• Posts should contain inside, outside or simple_background. (sometimes the background seems too complex for simple, but the location is too ambiguous)

There are some gray areas with inside/outside. Is a cave inside or outside? What about a car? Does it matter if the car has a proper "roof" vs being a convertible with the top off? Is outer space considered outside or is it a separate thing?

explicit tags

genitals anus sex masturbation cum pussy_juice

You can add precum and sex_toy

bitwolfy said:

There was actually a bit of debate about that recently: https://e621.net/forum_topics/33133
The talking_to_viewer tag implies dialogue, which indicates that it should only be used when there's text in the image, not for voice acting. Plus, there are sometimes cases where the character is clearly talking, but there is neither text nor sound.

I didn't realize dialogue was limited to text. Dialogue voice_acted -text gives a lot of results. I'd argue it makes sense to tag voice_acted posts with dialogue to distinguish moaning, etc. (not dialogue) from real words (dialogue). EDIT: if anyone wants to continue this discussion, I brought it up in the thread previously mentioned by bitwolfy: https://e621.net/forum_topics/33133

watsit said:
They shouldn't be. Ferals are considered to be nude by default, and clothed_feral is used to denote one where clothing.

bitwolfy said:
IIRC, the consensus was that naked ferals don't have to be tagged nude. It's only for characters that are expected to wear clothing normally.

Unfortunately that is not the case for *_only tags when I raised the issue on topic #29983, which now implied nude for both anthro and feral characters.

To elaborate on the inside/outside thing a bit more, I do think missing inside/outside tags is a very common problem, partially because it's common for users to only tag more specific location tags instead (sauna, beach, etc.). One solution is to just keep "posts should contain inside/outside/simple_background" as a usually rule, but it does have gray areas and exceptions as mentioned. A more complicated alternative is to have "usually" rules for locations being tagged with inside or outside.

Usually inside:
bathroom
bedroom
sauna
restaurant
locker_room
...

Usually outside (these are more prone to exceptions though):
beach
forest
hill
city
cloudscape
...

faucet said:

• Posts should contain either digital_media_(artwork), traditional_media_(artwork) or photography_(artwork).

Whar determines digital vs. traditional? Objectively 3D models/animations and photoshops are digital but can't other digital art be made to look traditional with the right style? For example...

post #157091
Maybe I just have untrained eyes for details in art vs. you long-time professionals. I see a nice painting of a fox. Would never have guessed it was digital.

glyme said:
Whar determines digital vs. traditional?

If it was done on a computer (digital), or if it was made IRL on a canvas with paints or pencils and such and then scanned into a computer (traditional). It can be difficult if not impossible to tell which a given piece is sometimes though, especially since art programs have modes to emulate traditional media. There's also mixed_media, which can include art that was made traditional and then scanned into the computer for extra digital work.

watsit said:
It can be difficult if not impossible to tell which a given piece is sometimes though, especially since art programs have modes to emulate traditional media.

This was what I was actually questioning. If you can't tell, do you put either?

glyme said:
This was what I was actually questioning. If you can't tell, do you put either?

If I can't tell what media it is, I don't tag it. Generally I look for signs of digital effects, things indicating that at least part of the image was done or modified digitally (so digital_media_(artwork) would apply, even if it's mixed_media). Otherwise, if the artist lists traditional media tools they used to create the image (gauche, watercolors, acrylic, etc), and it looks like something that could've been made non-digitally, I'll tag traditional_media_(artwork) then, too.

I wish more people would tag traditional_media_(artwork) digital media is the overwhelming amount of art uploaded to this site, and while it's never a bad thing to tag it, it's even more helpful to tag non-digital work.