9
votes
Is there such a thing as too many tags?
My intuition says that’s too much. I’ve noticing that most people use a lot more tags than I do, but I don’t really know how they work, hence the question: is there such a thing as too many tags? What’s the best practice?
They don't hurt. My take on tags isn't that they're supposed to help you digest the info in the post, but rather to link it. See also the discussion about zettelkasten. Tags are an essential part of crossreferencing that info. So those tags are supposed to help you a) discover tags through which to find related posts and b) help discover the post itself.
It's like garlic: The more, the merrier.
It's not very well-implemented yet either, but eventually they should also be a strong component of search. For example, if someone searches for "hydroxychloroquine", it's good that it includes this recent post about some of the studies in the results, even though that term isn't in the title/url/etc.
Thanks.
The purpose of tags in general is not lost on me, I was actually wondering if their implementation on Tildes had any specificity that could make a certain amount of tags more or less ideal.
Just imagine every link submitted has all the tags auto-imported from the other end by the submission tool. Users just give them a quick proofread before clicking on submit. The site has a synonym database for more uniformity. This clears up most of the problems with tags. We're not there yet, not even close, but someday.
A lot of the Tildes subsystems are placeholders. Designing them in from the start even if they aren't much more than stubs saves a lot of refactoring later.
That would be great indeed!
The ideal purpose behind a tag is to group things about the subject (or person if they're noteworthy) named in the tag. My post goes:
Politics: Reparations on slavery are a political topic. (A tad vague, I personally define political as 'something the government has a stake in and is divided on, although that's pretty problematic).
Ta Nehisi Coates: Author's name; mainly because @dubteedub says they're noteworthy.
Racism: It's about skin color defining your life.
History: It talks about the history of the above.
USA: It's specific to the US.
Long read: It's 14k words long.
Segregation: Segregation is an aspect of history the article touches on.
Slavery: see the above
Race: See racism. (I wonder if racism should be a subtag of race, given it's a product of race.)
Reparations: That's what the article is arguing for.
Opinion: It's an opinion.
Duplicate post: Repost (From ~-a year ago).
I do have the concern that using too many tags may lead to irrelevant results. If they cease to bring the threads I’m looking for I’ll be less likely to use them. But I may very well be wrong. Hence the question.
I'd be more worried about vague tags causing irrelevant results, rather than too many tags causing problems. There's some bizarre tags that don't really seem to resolve to a cohesive topic—for example, this post has
testing
.It's impossible to infer the intent of that tag without appealing to the topic of the post—which for me is a red flag against the use of that phrase as a tag. Does
testing
mean software testing? Pen testing? Medical testing? (Obviously the latter).Same goes for
usa
, which is a dangerous one because it gets crowded so quickly. IMO should only be used for federal-level or nationwide discussions, and can often be substituted for another tag, or a nested tag. Anything even remotely more geographically appropriate should have the state abbreviation appended.My hope is eventually we'll have "lightly nested" tags (not that scientific taxonomy tree that was attempted a while back), with each tag having its own page, a short description, use cases and guidelines around use of the tag, and perhaps some synonyms to help people out. Like Stack Overflow.
Again there's a bit of a disconnect from how it's working in practice right now, but I think there should be some inherent context in tags. When the tag is "testing" and the topic is in ~health.coronavirus, it should be implied that it's related to coronavirus testing. A topic about software-testing shouldn't be in that group at all.
There will be lots of these other terminology conflicts, but the group context should usually make them clear. For example something tagged "physics" in a hypothetical ~comp.gamedev isn't the same subject as "physics" in ~science, but the context should make the difference obvious.
Okay, cool. I wasn't aware the intention was to use group information to provide tag context, but that seems sensible. Thanks for explaining.
This is a search ranking problem. Does having more tags help or hurt for getting the right results to the top?
But it's only after a lot of tinkering with the search engine's ranking algorithm to make it smarter that we'll know, and I don't think that work is very far along yet. It's possible to get really sophisticated about it, almost arbitrarily so if you use machine learning.
A sufficiently smart search engine probably doesn't need extra tags at all, but it's easy to optimize by fiddling with the weights on them, depending on how much they're helping.