Getting rid of implications

Posted under General

This topic has been locked.

I want to get rid of tag implications. I haven't been approving them lately because it's just so easy to go overboard with them (a pinup tag? what's next, magazine? published_work?) Aliases are easier because there's a simple test: equivalence. Implications on the other hand can create tags with thousands of posts out of thin air, for a tag that very few people might use. They're also much harder to revert, and the way they can create cycles with aliases is hard to detect. I think if a tag's important and relevant enough to be added, it should be added manually. I'd like to hear convincing arguments for keeping them.

Updated by albert

If you want to limit them, then limit them to copyright>copyright implications (Gundam, Nanoha, etc). Those, at least the ones we have, should stay. But getting rid of the concept entirely... If you're getting too many suggestions, just restrict suggestions to mods, and have a thread where regular users can suggest but where you're under no obligation to add it, or even answer.

IMO implications should still exist, but yeah, it makes more sense for them to only be there for things that people actually use. For one example, the various panty_[x] tags all imply panties, which is certainly a good thing. But yeah, I wouldn't really mind not having them, I'd just have to remember to type the implied tags manually.

I can see some of the implications bein relevent. As stated the whole huge_breasts -> breasts, side-tie_bikini -> bikini -> bathing_suit makes sence and does help when tagging.

I'm not sure what other implications there are but I see some of them being useful.

I've been checking the implications system and I can tell that some implications might be relevant. However there is a problem with them: they only work one-way. the bikini -> swimsuit is a good example because a bikini is a swimsuit, but not the other way round. Now suppose you tag an image with swimsuit that also is a bikini, but you forget to add the bikini tag... someone will have to add it manually.
Now let's admit something: there is a lot of crappy tags out there, and -like albert said- there is going to be a lot of posts with a tag that just a few use.
My opinion is: get rid of implications. And lets start all to tag... manually, and make better use of aliases.

I don't get the anti-implication sentiment. It seems like the real problem is that too many are being suggested for stupid things. But some of them make perfect sense. Yes if it's important it can be added manually, but will people? I really like being able to get all Nanoha-franchise images by searching that tag, but will people always remember to manually add nanoha when tagging a nanoha_strikers image? That's what I like best about implications - the ability to create logical umbrella categories so we have to do less multi-tag searching, which we're limited on anyway.

Also, what would happen to existing posts with implicated tags? Would they be changed at all?

Close implication suggestions, weed out 80% of them and leave only the most common, whatever you want, but don't remove them entirely.

I think jxh2154 is right that the copyright->copyright implications are useful and that they're small enough that just the mods could keep them up to date.

The other sort of implications (the bikini->swimsuit type) are useful but prehaps don't need to be automatic.

  • They could be folded into the related tags feature as a sort of a manual override e.g. if you asked for the related tags for bikini,the implication would ensure that swimsuit would always appear in the list.
  • Or ther could be a separate 'suggested tags' link after the related ones that just uses the suggestions.

Either way since a user has to explicitly select the tags the interface could be entirely user run without the need for mod/admin approval of the suggestions.

I think implications aren't needed. As for copyrights, they can be added manually. Whenever I upload a pic of Rei, I usually add the neon_genesis_evangelion tag intuitively anyway.

I know some of you itt are coming up with good reasons for keeping them, I'm just saying that if Albert does decide to do away with them, it won't be a big deal.

This does not sound like a good idea... Implications are too useful to just do away with. There may be needless ones, but there are also very important ones, and they all have to be approved, right?

abcadeff said: I think implications aren't needed. As for copyrights, they can be added manually. Whenever I upload a pic of Rei, I usually add the neon_genesis_evangelion tag intuitively anyway.

That's not what I refer to. I'm referring to copy>copy implications, for example, 'gundam_seed' implying 'gundam' or 'mahou_shoujo_lyrical_nanoha_a's' implying 'mahou_shoujo_lyrical_nanoha'. That way you have one umbrella tag that covers the all the installments, which is useful for series universes where you want to be able both to get everything and to drill down to specific pieces of the franchise.

I don't mean 'lacus_clyne' implying 'gundam_seed'. Those are char>copy implications and a very different thing, and something I actually objected to previously.

Even those copyright implications are something I'm wary of. Let's take Gundam for example. There's UC Gundam and there's CE Gundam. Aside from general themes, these two universes have nothing in common. Is the Gundam tag all that useful in this case, aside from general curiosity? Does it really help anyone find the image they're looking for?

We don't tag Touhou characters based on what game they're from and I don't see anyone complaining there.

albert said: We don't tag Touhou characters based on what game they're from and I don't see anyone complaining there.

This argument seems to go on a different approach from what you speak of above. Not sure of what you're going for here. (It's almost like you're saying people would be ok if we only used 'gundam' and not 'gundam_seed' or 'gundam_00'). Touhou is a very different situation anyway.

I just don't see a good reason for abandoning the implication concept wholesale, especially when it's strictly controlled from the top as is. Purge most of them, sure, but don't take away the very option to have them at all.

albert said:
Even those copyright implications are something I'm wary of. Let's take Gundam for example. There's UC Gundam and there's CE Gundam. Aside from general themes, these two universes have nothing in common. Is the Gundam tag all that useful in this case, aside from general curiosity? Does it really help anyone find the image they're looking for?

In theory, yes (also Gundam tags are a mess, I should work on those...). I agree with jxh that Implications should stay, but with higher restrictions than they currently have.

I find implications way too useful to get rid of them. As I am doing a lot of tagging removing implications would mean a lot more work.
If you must, handle implications more strict, but don't remove them.

The current implication list should definitely be wiped and started from scratch. It preferable to do this instead of deleting them one by one and having second thoughts. And yes, it should me much much stricter from now on, no animal_ears or bdsm or one I really hate which is tanline -> tan. Though I would still like some that make tagging a lot faster for me like striped_panties -> panties and very_long_hair -> long_hair and the final_fantasy_# -> final_fantasy ones. I say have implications only be made by mods and janitors but only if there's a discussion in the forum for each implication.

memegui said:
...

Why would you wipe the whole thing out and start from square one? It would make much more sense to go through it, note the ones to definitely keep, delete the ones that need it, and be more careful in the future. Starting over sounds like a lot of needless work, and in the meantime people will post stuff that doesn't end up tagged properly.

That said, closing suggestions to Mods and above wouldn't really be a bad idea, and it would greatly diminish the work of maintaining the list. Allowing them to approve might also help. If you were worried about consensus, you could make it something like 3 mod votes or 1 admin vote approve it.

1 2