8
votes
Thoughts on restricting tags?
Often when I go to post, I have to check first what the most commonly used tag is for the thing I want to post. There is quite a bit of tag duplication/proliferation. Granted, this could also be dealt with in part by suggesting tags as the user types, but:
What do you think about having a specific set of tags to choose from, and when there are a handful of posts about a specific topic, new tags are added by administrators? This is essentially the way Lobsters works, and I think it works pretty well.
On one hand, this could reduce tag proliferation, but on the other hand, it could also corral people into only posting about certain topics that are sufficiently similar to existing tags.
I don't see much of a point of going for something so extreme until we try out autocomplete, popular tags, and the like.
This seems like a last resort if no other implementation works out. This site is open to change and experimentation to figure out how exactly to solve these problems, so there's no need to jump to something like that.
There is an issue in the to-do list for the site to automatically suggest existing tags to people as they're typing tags on their topics. So, for example, as someone starts typing "poli...", they'll see a suggestion of "politics" as an existing tag - which should deflect them from creating "political" or "politicians" as new tags instead.
This should prevent synonymical tags, but still allow people to create genuinely new tags.
One of the core ideas of this website is having a significant portion of the userbase be some level of moderator so we can have human answers to problems like this. Doesn't mean that alone will fix the problem, but it's worth taking into account.
Sorry, it does? Source?
This is disappointing.
What about something like popular tags? So you have an idea of common ones used?
Once we have 'suggest title' capability here, the submission form should pull up all of the existing tags from whatever's at the other end of the link. The user only has to proofread the title and tags that come up when they do that. We want the laziest submission process we can get, especially for people on mobile where typing is more of a pain. There's a wealth of metadata out there on every site now, and not many places tap into it.
That could lead to problems.
Take ~humanities, for instance. I believe all topics there should have a tag showing which of the humanities the topic is about: history, linguistics, theology, or philosophy (the arts is covered in ~creative). I doubt that those particular tags will come from the source and, if people get used to the submission page automatically providing tags for each topic, they won't add any extra tags that Tildes requires.
While the groups are this generalized, that's true. Once we have a lot of subgroups, though, those subgroups will apply their own tag signature, so if you submit something in ~humanities.philosophy aka ~philosophy, it'll also inherit that group's own core tags. My biggest worry with that system is helping people find the right groups for their submission, especially new users. If Tildes tracks which tags are most common in each group, it can take a guess where something belongs based on that. Over time the accuracy of that system should get pretty good as the site builds up a big tag library.
It's never going to be perfect, and that's why we want to have a decent percentage of the userbase able to add tags to a submission in the communities they frequent. I think we can get a good chunk of relevant tags from hoovering them up off of the linked sites, and the users can handle the rest even if the OP doesn't get it right.
We're also going to need tag aliases. Similar tags for the same topic should just be linked together in the system. We can probably let the mods set a lot of this up in their own groups somehow, since it's not something an automated system can easily guess at. That should cut way down on the duplication problem.
From where will a topic acquire tags like "utilitarianism" or "consequentialism" or "nihilism" or "ask" or "debate"?
From the source in most cases.
Here's the thing - the end user rarely sees the totality of the content tags that are out there now. It's there, but not part of the user interface on most platforms. Let me give you an example.
Take this video on youtube: https://www.youtube.com/watch?v=uvmz5E75ZIA
It's on the topic of utilitarianism, but if you only look at the youtube classification, it's simply "Education." Looks like there's nothing useful to classify the video, right? You need to view the page source to find the quiet, juicy bits that aren't being shown to the users. If we do that, we find this...
Specific enough for most topics, certainly. The amount of metadata out there now is fantastic. It's a consequence of search engines becoming so ubiquitous. Read any guide about putting content online and they'll tell you if you want anything to be seen, tag the hell out of it like this. There will be some sites that don't tag or don't tag well, but the vast majority of sites, especially any that are even remotely popular, will have this information. Tildes can import it to make tag suggestions and take the lion's share of the work out of this process for everyone.
If you say so.
Think of it like this. If you can find it in a search engine, it's already tagged to hell and back, almost certainly by the author(s) which means the tags are highly relevant and quite correct. Anyone who uses professionals to manage their web content, or reads even the most basic guides to doing it themselves, will be aware of this and will use it. If they don't, nobody's going to find or see their content and share it here in the first place.
We've checked into this. It works for news sites, all popular video and audio streaming services, education sites, image hosting, blogging sites, all social media sites like twitter or tumblr or facebook... it's hard to find a site that isn't doing it now. Reddit itself would be a good example, closest they've got is the subreddit name.
Going in through the API of the sites usually provides even more data and in an easily queryable format. That's probably going to be the way Tildes interacts with most sites when pulling up this data. I'll wager well over 90% of all submissions can auto-tag with high accuracy, because almost all submissions are going to point back to these services. Over time, this is only going to improve and get better. A few years from now we'll probably reach a point where most sites won't even accept content if it isn't tagged by the author, you'll have to do it if you want to post/share/submit anything.
This is a fairly recent development. Tags weren't this ubiquitous even three years ago. Reddit could have never taken advantage of this, because when reddit was created, this invisible ocean of metadata simply did not exist yet.