What the hell are we doing with hierarchical tags?
safety.air, food.processed, neurons.artificial, storage.data - there are many many more examples to be found of unnecessarily hierarchical tags. These are not tags that benefit from such a scheme. All we have now are syntactically reversed phrases, reducing readability. What person wants to look up "safety" and peruse rail vs. air vs. public? Or energy vs. data vs. thermal for "storage"?
Where we have the line drawn now is far too arbitrary. Why is republican party not party.republican? decomposition.runaway?? lights.head for "headlights"??? I feel like I'm taking crazy pills.
I propose that heirarchical tags should only be used for things that have a clear natural hierarchy that is generally used outside of Tildes for some purpose. Geographic hierarchy makes perfect sense to me. Being able to search by country, then by state/province, then by city is very sensible. For describing a section of the tree of life - yes please. Maybe we should even go so far as to use : instead of . for tags like source.youtube. When would I ever look up just source?
IMO removing all or most of the hierarchical tags would lead to an ever worse mess with tags getting divorced of their relationships with each other, which would make browsing by tags basically useless. E.g. As it stands, ?tag=safety shows all the other hierarchical
safety.*tags, and should someone be looking for a specific one (like ?tag=safety.air) they can still browse those too. But were they to be broken up intosafety,air safety,rail safety,road safety,public safety, etc. then viewing all safety related topics would be impossible.Is the hierarchy a bit arbitrary? Sure, it is in places, but it's an ad hoc system that has slowly developed and evolved over the last 7 years so that shouldn't surprise anyone. But it's not entirely arbitrary, and a lot of thought has gone into most of these decisions. E.g. We use
republican partyinstead ofparty.republicanbecause "Republican Party" is the party's official name, andpolitical party.republican partywould be unnecessarily long and redundant.And FYI, we use
source.youtubeinstead of justyoutubenot because of anyone browsingsource.*but to help people with filtering. If we only ever usedyoutubeas a tag then people filtering that would also be filtering out every topic written about the subject of youtube, not just the content hosted on youtube.Which brings up another important point. Ultimately, a lot of the decisions we've made about tags comes down to how limited the filtering system currently is. If people were actually able to filter by link sources or
videos + youtubethensource.youtubewouldn't be necessary. Likewise,politics.usawouldn't be necessary if we could filter outpolitics + usa.*or something to that effect. Etc. Etc. Etc.And yet another issue leading to a lot of these decisions is our extremely limited abilities to edit tags, especially in bulk, which has made things far more difficult to manage. So we're generally just trying to do the best we can with the features we currently have available to us, and the limited time we can devote to it. Sure, we can message Deimos to mass edit a few tags here and there, but it feels weird to bug him over every little minor tag needing to be changed. And sure, we could just manually edit all those slightly inconsistent/arbitrary tags on every old topic that includes them ourselves (which I've actually done plenty of times over the years when certain tags became more formalized and others rendered obsolete), but that's a massive PITA.
And that's why the tagging system is the way it is, love it or hate it, and I sure as hell ain't gonna be the one to change it all now. :P
I just want to say I’m grateful for all the tagging that you and other members do. I think I speak for many when I say that I do not care that much about the details as long as they are internally consistent, and you all have been doing a great job with that.
I think it's worth pointing out that no matter what nomenclature is used, it'll fail when it comes into contact with reality. There's no cloud of tags one can point to on the entire internet that is correct or even standardized. Large data systems are always messy, that's an inevitability. It's also nothing to worry about - in fact it's better to embrace the chaos. There's no point in going into wikipedianesque flamewars over minor nomenclature issues that only a tiny fraction of people even care about. At some point one has to leave OCD at the door.
The tags are for pedantic geek-centered categorization and all they really are is search++. They do a great job of zipping around the cloud of submissions for similar topics, and if a topic is important and busy enough, it gets a group which is where the tags yield to the community. I don't see any problem with that or even any missing features for the tags themselves. It's surprising how well even this simple system handles the organization.
I think it works great as it is. It'll work even better if an LLM or smart script is importing and tagging everything that's submitted. Then the humans only need to tweak the bits it gets wrong and set up some aliases from time to time. Should scale just fine to any size data set, and since the tags aren't arbitrary they remain versatile for any nodes to use any way they like.
I can't really think of a way to improve on it. I suppose some sort of confirm / reject / revert / pin mechanics might be useful to prevent vandalism on large nodes, but since there are no large Tildes nodes it's silly to bother about. Solving problems early usually leaves one a slave to old / bad assumptions. Best to wait until the problems materialize, at least then you've got better data on the causes and some real world use cases.
Yeah, real shame. That's easy to do on the front end with an extension, but I'm not sure why the site itself can't do it by default.
That one's understandable. Doing combinatorics on anything user-facing gets messy quick. Both from a UX and performance standpoint. And I do like how lightweight this site is. I'm sure this community is slightly more technical than average, but probably not technical enough to want to understand such a combo filter.
That's also fair. I'd be interested in taking a whack at it, but These last few years have kept me busy enough without adding volunteer work into the mix.
@mycketforvirrad quietly fixing the tags on the post and not weighing in is hilarious
I weighed in. I gave @cfabbro their exemplary tag. 🙂
mycketforvirChad
I don't feel as strongly about it as you do. I mostly use tags to filter out one or two things (mostly political related). But, I sometimes do feel that tags are not a really a community feature as they are strictly controlled. Clearly there is a system, but it has been slightly annoying at times that tags I actually did put care and attention to were changed.
At the same time, I do realize that for tags to work they need to be consistently applied. In order to do that the current system does work actually pretty well even though it does lead to situations like this at times.
I tend to mostly agree. Tags should only be hierarchical if one is a subset of the other in a way that's meaningful for searching or sorting.
That said, "What person wants to look up "safety" and peruse rail vs. air vs. public?" is a use case I could imagine, and "Why is republican party not party.republican", well, I could see "political parties.republican" being useful as well.
Anyway, I don't think it's a huge deal, since, 1: I'd wager not a whole lot of people use the tag system in the first place, and 2: If someone tags your post in a way you don't want, you're free to change it back and/or DM the person who changed it about it.
What I think might be a bigger misutilization of the tagging system here is the sheer number of orphan tags. I'd wager a vast, vast majority of the tags on tildes have only one post. A long time ago I recommended showing use counts in suggestions, like "
bach (12),js bach (1),composers.johan sebastian bach (3)" or at least being able to browse a list of tags by popularity.I would like to see a list of all tags too, and see if I can word cloud them and re-grouo orphans
I thought "language models.large" was a little silly when I first saw it, but I just go with whatever is already in use. Someone with the ability to do a bulk rename might do a cleanup.
:D :D
Solution: let's just keep tagging the way we are, for the next couple of years, and once enough bulk has been accumulated, feed all existing tags to a language model.large.advanced - or whatever they will be called to distinguish from the trash models of today - and let it create order from the chaos.
I think it makes sense, from the aspect of trying to avoid tag clutter. Like
food.processedcould easily end up being tagged bothfoodandprocessed food. Same goes forneurons.artificial, someone might search for "neuron" so wording it that way means one less tag overall on a post about artificial neurons.Others make slightly less sense, but seem to be more for organizational purposes in the database? There's multiple types of storage, so being able to just add on the specific type to the end makes it easier to compile how many tags involve storage. Doesn't really do much for end users though.
As for source tags, I think the main benefit of the source tags is for filtering purposes. I think Youtube would particularly be blocked by some people who don't care for videos.
By the way, are there lists anywhere of all the hierarchal tags and/or "common" tags on each group? Because that's my biggest challenge with tagging posts. I often end up opening Tildes in other tabs to see if tags exist for other groups that don't pop up on the one I'm posting to.
I don't have strong opinions on this, but I can speak to (what I presume is) the idea behind tags like "source.youtube". It's not that someone might want to search for "source", but that someone searching for posts about YouTube wouldn't want to get every post from YouTube.
In this case, it's more like a semantic namespace to distinguish a bare "youtube" tag from more specific things like "source.youtube", "video hosts.youtube", "google services.youtube" etc. Not that we necessarily have a lot of those other tags to distinguish from yet, but it's nice to set up a good taxonomy early.
Can this get tagged "tags.questions" and "doing.hell.whatarewe."
(◠‿◕)
My fun tags get removed so I clearly don't know doing.hell.weare
Tags.hierarchal.opinions
laniakea.virgo.milky.way.oort.earth.net.tildes.tildes.opinions.tags.hierarchical.opinions
I just follow whatever Mycket does. And if I have to tag my own stuff, and it's apparently not good (it's never good), Mycket will fix it. I will admit it, I don't entire understand the tagging system here, either. But I feel like as long as we have someone who does seem to understand it AND is willing to tag all the things, then whatever. I don't really care. Should we have more than one or two people doing it? Probably. But all things considered, I don't think it's that important.
Maybe if Mycket gets tired of our awful tagging, or lack of it, one day and leaves, then maybe that'll be the day for us to have this discussion. *Shrug*
The "tag aliases" feature has been brought up several times before as a solution to this. So the current hierarchical tags might be chosen as the "canonical" tag and then aliases would exist, so we could have both instead of forced to choose one over the other.
This is one of the things I've been wanting to add when I have time/energy to work on Tildes properly again. As an aside: I don't need to explain to the techie crowd here, but the software industry is in many instances moving to an extremely demanding—and on an individual basis likely unsustainable—pace. So the maintainers here are also swamped with work, and of course non-work life things as well.
Re: Aside - I don't quite understand exactly why the industry is squeezing workers more other than "Be careful, don't want to get replaced by AI!". But if that's it I don't really understand as I can't see our phase of AI replacing developers.
Because they can.
There are many more people looking for work in the sector, that puts the pressure on for anyone still having a job to raise their output or lower their salary/pay.
This is why we need unions and workers rights.
Even well paid workers are disposable when the owners see a chance to raise their profit.
The tag system is a bit difficult to apply, but I just try my best and trust that someone will add/remove tags later.
I thought the tags on the recent gooning post were pretty funny though. I wonder if the
gooningtag will ever be used again.Maybe we can bring back the writing competition with the already-used obscure tags as the prompts. A wizard cursing someone for gooning over monkey bottoms—
Actually wait, what post talked about
monkeys bottom? Because I'm not finding any and I want context now!Huh, apparently it didn't show up when I tried entering the tag since I wasn't subscribed to ~humanities.history. It showed up when I searched, but the tag is beneath the cutoff on the search page and I figured it was more literal.
I am now mildly disappointed the wizard's curse wouldn't involve a literal monkey's bottom, but at least I get to read a neat article and I'm now subbed to the subgroup!
Those tags are a work of art
Lol a writing competition that uses the
gooningtag may be a little too “Letters to Penthouse” for TildesDepends on how much interpretation you allow. I could see "gooning" be slang for working as henchmen for supervillains.
That would probably warrant the creation of the
gooning.baddiestag.Would that therefore make a judgment call that the current gooning is therefore good?
I think more professions.villains.gooning
But what if you're gooning to baddies?
Gooning.motivation.baddies
(I have self appointed as the tag whisperer. )
At the database level it wouldn't be so hard. Filter by tags with a
., then go from highest frequency tag to lowest and make a judgement call to update them all at once. Stop going once you're in the long tail and run out of steam.I would gladly do the work such that all he needs to do is give me a report on what tags are in the system and apply the updates from a CSV.
Edit: I can also put together a small website that verifies you have a Tildes account and then crowd sources proposed tag changes and votes on those changes if we don't just want one or two people to decide how the changes are made. But honestly I think one person (myself or a volunteer) making consistent changes is more important than democratic procedures.
Isn’t this pretty close to the current system but you just disagree with the choices?
Yes :P
I don't do a lot of tagging - I try to tag what I post but that's it - but I think that in general most of the hierarchical tags makes sense. Most of the time, it seems to come down to if something is more filterable or not. To grab some of the examples:
foodandfood.processedmake sense to be hierarchical. If I want to filter out all food topics, that includes processed food.partyandparty.republicandon't make sense to be hierarchical. If I want to get rid of things tagged party, I don't necessarily want to get rid of Republican party.I feel similarly about
safety.airandstorage.data. I'm not sure about the utility of neurons as a tag, so I'm not sure ifneurons.artificialis useful, but it at least passes the sniff test to me; if I don't want to see things about neurons, I don't want to see things about artificial neurons.We don't necessarily have to go with my personal choices for tag updates, but I would like to see some kind of review done.
I'm intrigued how people actually use these tags. For me, they're mostly just noise that I blend out, and if I needed to do some sort of filtering, I'd probably just filter by group rather than think about tags. But presumably there are people using them - what sort of tagging structure works best for you guys? Or is this just a write-only feature?
In your tildes settings you can filter out specific tags. Since they are applied pretty consistently, it allows me to filter out some topics I really don't care for. For example, I realized a long time ago that I really don't care for political discussions on Tildes since that is not what I visit this website for. Now I don't need to stumble upon them since I filtered out some tags related to politics (mainly the politics tag though ;) )
I come here for things to quickly read and possibly discuss. I filter out videos because I don't want to invest more than 3-5 minutes most of the time. It frustrates me when I click an interesting link and a video appears that might take an hour to watch.
when i first joined tildes (almost right away), i thought that instead of set groups it would all be tag based. art.popart.keithharing would give us a listing of all art at the gp, pop art discussion only at the parent and specifically keith’s work at the child; then we sub to tags and that builds our feed.
it would make sense for certain user levels to not be able to ‘create’ tags at the base level — e.g. popart couldn’t be created as a base but it could be a subtag
lights.headlolI have nothing more helpful to add but am amusing myself so....
I think you have to have
lights.vehicle.head and lights.wearable.head to distinguish headlights from headlamps