What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 133 Stack Exchange communities.

StackOverflow currently has well in excess of 30000 tags. But even on page 100 out of 852 there appear to be extremely few questions.

Is there a general mechanism to review and revise existing tags? I'm noticing that there are many tags for specific data structures within specific languages; wouldn't it make sense to coalesce all those into a few, general tags for that language? Ultimately, if you tag your question as c++_unique_ptr, you're almost surely going to tag it as as well.

By not offering too many specific tags we encourage people to add all the appropriate general tags without getting bogged down. And reversely it is already the case that lots of lesser-quality questions with general tags are lacking the specific tags that would apply, but end up getting answered anyway.

Removing overly specific tags is one way to clean up. Another way that comes to mind is to de-separate version-specific tags (if that makes sense), but I have no opinion on whether version-specific tags are currently useful. Finally, there are some bogus tags like which isn't helping anyone because it subsumes Perl HTML templates and C++ templates. If the tag can't be used meaningfully, do we need it?

Enough with the ranting; the main question is whether there's any mechanized, general process to prune the SO tag soup, and if anyone feels that pruning would be desirable.

share|improve this question
1  
So how, exactly, would we go about creating an automated process to do this? Cleaning up tags is a potentially great idea, but it's one that needs to be done by humans that can read questions, evaluate context, and make decisions. I wouldn't leave this to some kind of script. Just because a tag isn't used by very many questions (~130 on the page you linked to) is not an indication that there's something wrong with the tag. You're likely correct about specific examples like template, but how is a machine going to know when it belongs and what to replace it with--Perl, C++, etc.? –  Cody Gray Jul 20 '11 at 9:57
    
Well, an assisting machine would be useful: Some magic website where you can drag tags around and see what would happen if you were to change one tag for another... something that lets you explore the consequences of and potential for tag rearrangement. The whole thing could be powered by a hint system where you say c++_unique_ptr falls under c++, but a human could be processing the hints and check whether any particular merge makes sense. –  Kerrek SB Jul 20 '11 at 10:01
    
Yeah...we don't have that... :-) –  Cody Gray Jul 20 '11 at 10:10
    
Have you heard about tags synonyms ? meta.stackexchange.com/questions/57479/… –  Cyril Gandon Jul 20 '11 at 10:12
    
@Scorpi0: That's already quite useful. How do I view the collapsed tag collection of SO which suppresses all synonyms and aggregates their statistics into the master tag? [Edit:] Oh, are the synonyms already removed from the main tag collection? –  Kerrek SB Jul 20 '11 at 10:42
    
today i find some useless tag such as "facbook"(Missing 'e' after 'c') and or "asp.netmvc"... –  Jeson Park May 1 '13 at 14:05

2 Answers 2

up vote 2 down vote accepted

the main question is whether there's any mechanized, general process to prune the SO tag soup, and if anyone feels that pruning would be desirable.

See Jeff's answer on the question Should We Zap The Low-Occurrence Tags?

I am convinced this is a useful cleanup task, to remove these orphaned tags from the system automatically. I've implemented a routine that removes tags created more than 6 months ago which have been used only a single time. It will run monthly.

That's the only automated tag pruning that goes on. Any other tags that need to be removed either need to be set to a synonym of another tag (but only if they literally mean exactly the same thing) and merged, or just removed from questions manually.

I'm not sure I agree with your point about coalescing more specific (but still used) tags into more general tags. For every question that gets answered despite not having specific tags, I can find one that didn't get answered because it was not tagged appropriately. There have also been tags that have been used a lot but were almost totally useless (subjective, not-programming-related, off-topic), so it would be very difficult to come up with an automated way to judge the usefulness of a tag. In short, I think humans need to keep doing this.

share|improve this answer
    
Alright, that looks very good. So, if we synonymize a tag, does the synonym get removed entirely from the system as far as visible tags are concerned? But I was also wondering about "generalizing" tags that are too specific - those wouldn't be synonyms, but genuine substitutions. –  Kerrek SB Jul 20 '11 at 12:46
    
@Kerrek SB: The merge still has to be done by a moderator, if I remember correctly. You can flag a question with the tag that needs merged and let us know that it's a synonym of another tag. –  Bill the Lizard Jul 20 '11 at 12:48
    
@Kerrek: See the last paragraph I added for my thoughts on generalizing tags. –  Bill the Lizard Jul 20 '11 at 12:59

Not sure how to best thread this conversation. Should I have opened a new topic?

Check out the tag : It is way over-specific, and it should be subsumed under "Windows API". However, since it has too low a score, I cannot even propose a synonym. How would one best get rid of such a tag, and why was it created in the first place?

share|improve this answer
    
1. You should probably just edit this into your original question as an update. 2. I would say that that tag is too obscure to even propose as a synonym, and I'll re-tag those. Synonyms are for things that would commonly be used in the place of another tag but needs to be unified under one tag –  Daniel DiPaolo Jul 22 '11 at 14:32
    
Cheers, thanks. –  Kerrek SB Jul 22 '11 at 14:34
    
Also, indeed, synonyms are really not the right thing for the gripe I've been trying to bring up. This tag here is the true example of what I want to prune: Generalise overly obscure tags upwards into more useful categories that people might actually look for or respond to. –  Kerrek SB Jul 22 '11 at 14:36
    
Uhh, there are no questions with that tag. So you don't need to get rid of it, or do anything at all with it. It's not being used anywhere on the site. –  Cody Gray Jul 22 '11 at 14:42
    
Not fair, there were three questions when I asked! :-) But I do would like new users not to be able to select "getmodulefilenameex" as a tag -- is that possible? –  Kerrek SB Jul 22 '11 at 14:45
    
Hrm. I don't see anything wrong with that tag. Yes, it should obviously be combined with [winapi], but if I want to know about using a relatively common function, I should be able to search all questions with that tag. There is a setwindowshookex tag that I think is quite useful. –  Cody Gray Jul 22 '11 at 14:48
    
Hmm... 40 questions in three years? Couldn't that just go under WinApi too, with "Hook" in the question title? I mean, OK, some functions may like their own tag, but somehow it feels gratuitous. The nearest library/framework tag above should provide ample charactarisation, don't you think? –  Kerrek SB Jul 22 '11 at 14:50
    
@Cody yes there are none because I re-tagged them (as stated above) –  Daniel DiPaolo Jul 22 '11 at 14:51
    
@Daniel: Weird. Why? What was that tag hurting, and why should we bump a whole bunch of old questions just to remove a tag? If the tag is bad, a moderator should ban it, automatically removing it from the questions without leaving an edit history trail. –  Cody Gray Jul 22 '11 at 14:52
    
@KerrekSB: It could. But with that logic, why not just tag the question windows? Or programming? The idea is to categorize questions. Just because something is asked about relatively infrequently does not make that categorization mechanism any less useful. We don't need to ban or eliminate any of these tags. They all serve the purpose of tags: accurately categorizing content. I use it all the time in searching questions. I do the same thing with a number of other WinAPI-specific function tags I'm hesitant to reveal here, lest they be removed... –  Cody Gray Jul 22 '11 at 14:53
    
@Cody: Come on, no reason to be snarky, I'm just trying to help improve SO - of course we won't have a single "programming" tag. But we have tons of "C++" questions which are all well structured and don't have sub-tags "for-loops", "overloading" or "partial specialization". There's a hierarchy of specificity going from tags to question title to question body, and the separation is worth making. Categories are very useful, of course. But I'd like to think that a category should identify the field of the question in such a way that a competent practitioner of that field ... –  Kerrek SB Jul 22 '11 at 15:05
    
... should have a reasonable chance of understanding the question. That's why I'd like to stop at the most specific library or framework, if that notion applies, and not dive into individual functions. If you know the Windows API, you're probably able to help. But would there really be someone who only knows one function well? You see, I think the specificity "function" is something perfectly suited for the question title, but it's too narrow for the category. Hm, I hope that's making sense. –  Kerrek SB Jul 22 '11 at 15:07
    
@KerrekSB: I'm not being snarky. I'm genuinely upset at the idea or reality of having tags like that removed. As I mentioned, I regularly use them to find content. It's not like SO's search system is fantabulous. I have a lot more success searching by tag than I do searching by keyword. I don't understand what those tags were hurting by being there. If you aren't interested in them, don't search for them or filter them out. This is not improving SO, this is hurting my experience and ignoring the purpose of tags as categorizers for content. –  Cody Gray Jul 22 '11 at 15:09
    
There is absolutely nothing wrong with a [method-overloading] tag or a [for-loops] tag. I've added those to questions myself. They help people to find questions concerning topics that they're interested in. Isn't that the point? Yes, there are other parts of the question that can contain that information, but that's irrelevant to the point here. There are a whole bunch of C++ questions that are just tagged with [c++]. That's completely worthless. I get nearly 80,000 questions when I search that tag. I can't find anything that way. –  Cody Gray Jul 22 '11 at 15:10
1  
@Daniel: I quite disagree. As mentioned, there's no use in bumping old questions just to remove tags. Those tags weren't hurting anyone. You can't accurately gauge the usefulness of tags based on the number of people following them. I don't "follow" tags, although I do use them extensively. There are quite a few other tags that are only used by a few questions, but there's no reason to assume that they're "bad" tags any more so than they cover topics/subjects that just aren't very popular. I don't understand the crusade making it harder to find questions about unpopular material. –  Cody Gray Jul 22 '11 at 15:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .