Topic: How the tag alias system works?

What conditions must be met for an alias to be adopted?

It depends a lot on the alias itself. Some general good points to think about when suggesting an alias:

• It should be useful (like aliasing common misspellings and alternate phrasings to the correct tag)
• In general it should not create more clean up work (Alias should not be made when it's accurate only sometimes)
• In general should be something that is too much work to clean up manually (You do not need to alias a tag that is used only once)
• Most of the time the alias should be from less common tag to more common tag

It's perfectly fine to suggest an alias even when you are not entirely sure if it's going to work, because that's what the discussion threads are for. If it's something that should not be approved, people will pop up in the suggestion thread, and explain the issues it has.

Also keep in mind that the queue is very long, so sometimes it may take a while to get approved even when it's a good one.

ruppari said:
It depends a lot on the alias itself. Some general good points to think about when suggesting an alias:

• It should be useful (like aliasing common misspellings and alternate phrasings to the correct tag)
• In general it should not create more clean up work (Alias should not be made when it's accurate only sometimes)
• In general should be something that is too much work to clean up manually (You do not need to alias a tag that is used only once)
• Most of the time the alias should be from less common tag to more common tag

It's perfectly fine to suggest an alias even when you are not entirely sure if it's going to work, because that's what the discussion threads are for. If it's something that should not be approved, people will pop up in the suggestion thread, and explain the issues it has.

Also keep in mind that the queue is very long, so sometimes it may take a while to get approved even when it's a good one.

Shouldn't label cleaning be left to some automated bot?

This level of automation should be easy, right?Why not try to automate various tasks without reducing the quality of service?

kubihazusaikou said:
Shouldn't label cleaning be left to some automated bot?

This level of automation should be easy, right?Why not try to automate various tasks without reducing the quality of service?

Aliases cannot/should not be automated in any way
If by "automating" you mean creating aliases for common typos or synonyms, we don't create aliases to create aliases, they're created when they're needed

kubihazusaikou said:
What conditions must be met for an alias to be adopted?

See e621:Tag Aliases for a quick explanation.

There are many nuances depending on the type of alias request you are making.

• 1) The primary function for alias requests is to merge tags that have the same meaning together into one tag.
• For example, female is the widely-used and established tag for characters with female genitalia or feminine appearances. There are many alternative terms that people could use, such as girl and woman, which are synonymous with female and thus does not need to exist separately as a unique tag.
• It may also be used to merge plural terms to singular terms, such as chocolates and chocolate. We would prioritise using the singular terms, unless their plural counterpart seems to be more grammatically appropriate (e.g., paws, feet).
• It may also be sometimes used to fix common typos or misspellings of a tag that many users seem to make.
• 2) The secondary function is to merge artist names together into one tag, such as in the case where an artist uses more than one username or account to post their artwork.
• For example, tojo_the_thief has various art galleries but their usernames slightly differ from one another, where they are sometimes called tojothethief or tojoafterdark.
• We would want to prioritise using the name they are most commonly associated with or the one with the most uses and alias all of the alternative username variations to the first tag.
• On occasion, some artists may post NSFW artworks solely on a different name and do not wish to have their artist tags be merged with the name where they post SFW artworks. This should be taken into account when suggesting artist aliases.
• 3) The tertiary function is to deal with ambiguous, redundant, or invalid tag uses.
• For ambiguous tags, this mainly concerns tag names that could have multiple meanings or uses, such as commonly-used terms or character names. In this case, ambiguous tags are aliased away into a "disambiguation" tag, which is a form of invalid tag that appears red in colour on the post's tag list to inform users that they need to fix it and be specific on the tag they are using.
• For redundant tags, this mainly concerns tag combinations that are deemed to be unneeded or excessive, such as when users make a combination of two separate but distinct tags in one merged tag. This is highly dependent on community discussion on whether or not it would be useful.
• For invalid tags, this mainly concerns tags that have questionable use or validity (such as those that do not meet our Tag What You See (TWYS) policy) or tags that are prohibited or abusive. For example, the works_of_japanese_artists tag that you suggested on topic #57460 which did not meet our TWYS policy. Tags such as these are commonly aliased away to invalid_tag to prevent users from using it again.

There is also the consideration of how often a tag is used, or more specifically misused, that becomes a factor on whether your alias request would be approved.
In general, if a tag is barely used or is only used by one person, you should try to manually fix/remove it first instead of submitting an alias request for it.
This is to stop people from submitting trivial requests that could be fixed within a minute without admin intervention.

kubihazusaikou said:
Shouldn't label cleaning be left to some automated bot?

It kind of is. Once an alias has been added to the system, all instances of the old tag will be automatically converted to the new one. That's the purpose of aliases, saving people the work of hunting down posts and retagging them.

kubihazusaikou said:
Shouldn't label cleaning be left to some automated bot?

This level of automation should be easy, right?Why not try to automate various tasks without reducing the quality of service?

• A poorly implemented automation is a great way to blow everything up.
• Automating the creation of a bunch of aliases would add bloat to the aliases, which means more to sift through for every search, every autocomplete entry, & every post edit.
• Creating an automation system that won't ruin everything takes time, a finite resource which us developers must allocate intelligently. This is not an intelligent usage of that resource.

We're (i.e. I'm) working on overhauling perms to allow us to appoint dedicated users to approve/reject AIBURs. This will take time, but is a far better usage of that time than a sloppy automation would be.

aacafah said:
We're (i.e. I'm) working on overhauling perms to allow us to appoint dedicated users to approve/reject AIBURs. This will take time, but is a far better usage of that time than a sloppy automation would be.

That would be a very risky thing to appoint users the ability to approve AIBURs, especially when votes can be very contentious even between staff.
I'd still much rather have that ability to approve or "veto" requests be locked behind Moderator+ roles.

Mods can't do that; only Admins. Hence the colossal backlog.

This would be just like every other staff position; you'd have to apply & be vetted for it, & there'd be a training period where you'd have to have your decisions approved by a senior staff member.

Besides, it's no more risky than letting users ban other users. Hell, I think I could ban a janitor rn if I wanted to (and was suddenly overtaken by delusions), even though that should absolutely be Admin only. Perms are currently not a s thorough as they should be.

Also, we can make whatever restrictions we want; we could only allow them to reject bad ones, we could lock them from acting until a certain vote threshold, we have a lot of control.

And finally, we do have the functionality to roll back AIBURs. It'd be a pain, but we can do it.

aacafah said:
Besides, it's no more risky than letting users ban other users.

Ehh, a 1,000,000+ strong sidekiq queue that drags the site to a crawl does a lot more immediate damage than banning someone would

aacafah said:
We're (i.e. I'm) working on overhauling perms to allow us to appoint dedicated users to approve/reject AIBURs. This will take time, but is a far better usage of that time than a sloppy automation would be.

if it's a new staff role and you guys don't call the role "Bureacrat" I'll be upset.

aacafah said:

Also, we can make whatever restrictions we want; we could only allow them to reject bad ones, we could lock them from acting until a certain vote threshold, we have a lot of control.

Letting them reject only bad ones would already lighten the load on the existing admins by quite a bit. Though the question is how "obviously bad" a BUR has to be before it gets rejected by the BUReacrat role. Also the role might kinda attract the ire of others since their only role is to reject.

Locking action until a vote threshold is met is just about the worst restriction to place. The easier decisions don't have many people voting on them, and the ones with many votes are either impactful or controversial, where it's better if it's handled by admins (plus there is probably ongoing discussion between staff about it already)

aacafah said:

• A poorly implemented automation is a great way to blow everything up.
• Automating the creation of a bunch of aliases would add bloat to the aliases, which means more to sift through for every search, every autocomplete entry, & every post edit.
• Creating an automation system that won't ruin everything takes time, a finite resource which us developers must allocate intelligently. This is not an intelligent usage of that resource.

We're (i.e. I'm) working on overhauling perms to allow us to appoint dedicated users to approve/reject AIBURs. This will take time, but is a far better usage of that time than a sloppy automation would be.

I'm just saying that the cleanup after the alias is passed should be done by the robot.

dba_afish said:
if it's a new staff role and you guys don't call the role "Bureacrat" I'll be upset.

...It will be now.

snpthecat said:
Letting them reject only bad ones would already lighten the load on the existing admins by quite a bit. Though the question is how "obviously bad" a BUR has to be before it gets rejected by the BUReacrat role. Also the role might kinda attract the ire of others since their only role is to reject.

I'm more thinking we have figurative deputies that can only reject, and when they have enough experience and trust upgrade them to a sheriff that can approve.

kubihazusaikou said:
I'm just saying that the cleanup after the alias is passed should be done by the robot.

... again, it is. That's the point of AIBURs. If you're talking about the pussy BUR, that's an extreme edge case, and one better reserved for manual cleanup.

donovan_dmc said:
Ehh, a 1,000,000+ strong sidekiq queue that drags the site to a crawl does a lot more immediate damage than banning someone would

Fair, though again, I'm gonna put restrictions on the size of the change. No one's aliasing anthro or something.

aacafah said:

Fair, though again, I'm gonna put restrictions on the size of the change. No one's aliasing anthro or something.

c'mon just let me alias all the labia tags to "cooter" it'll be fine I promise

...well... I mean, if you promise...