A real solution to ambiguous tags

Question

Introduction

Even within a site's relatively focused range of topics (it's unlikely anybody on SO will think sun refers to the solar system's star), some tags are inherently ambiguous.

There are currently two unsatisfying options:

1. Declare that the tag is only to be used for one of the possible meanings, write a tag wiki explaining it, and link from there to the newly created tags for the other meanings, and keep cleaning up the tag whenever someone ignores the tag wiki popup explaining the tag.
2. Use alternative tags for all meanings (e.g. cd-command and compact-disk, instead of cd), and keep removing the undesired tag whenever it pops up again. And it will. The rep requirement is not deterring users from creating unnecessary or redundant tags.

Blacklisting tags is a complicated ritual occurring once in a blue moon, at most, and involves a dragon, a team member, and sacrifices to be burnt. It also makes those other, unambiguous tags not easier to discover, if it were applied to ambiguous tags. I don't consider this to be a viable option in this case.

Proposal

Here's the proposed feature-request:

Just like a tag can be declared a synonym of another tag, basically changing all references to the synonym to references to that other, canonical tag, it should be possible to make it a disambiguation tag (or ambiguous tag) for multiple other tags. It's a logical progression from "You entered a term with the same meaning as this other term" to "You entered a term with the same meaning as one of these other terms, please choose the correct meaning".

Use of the tag should force the user to choose one of the alternatives presented:

• "You entered 'as', did you mean as-keyword or actionscript?"
• "You entered 'vb', did you mean vb6, vb.net, or virtualbox?"
• "You entered 'vm', did you mean virtual-machine or virtual-memory?"
• "You entered 'sas', did you mean serial-attached-scsi or sas-software?"
• "You entered 'kvm', did you mean kvm-switch or kernel-based-vm?"
• "You entered 'mono', did you mean mono-framework or audio?"
• "You entered 'dsl', did you mean digital-subscriber-line or damn-small-linux?"

^{Not all of these examples might be good (I'm not sure e.g. about the first two SO examples), but that's not really the point here anyway — they should just convey how this could be used.}

Note how some of these tags would probably never be used if users had to type them, and yet everything shorter is ambiguous.

This could probably be mostly solved by extending the synonym/tag wiki excerpt dropdowns shown when a user enters the tags, proposing the alternatives for an ambiguous tag entered by the user (like a synonym suggestion, but with multiple options). If ignored, the user could still be shown an error message like with blacklisted tags, but by designing the selection dialog well, it should be possible to prevent that in most cases.

Related topics

• Suggest disambiguation to user when tagging

  This topic is related, but the user seems to have accepted the inclusion of tag wiki excerpts in the tag suggestions drop-down as sufficient.

• How should ambiguous tags be dealt with?

  This requires extensive discussion and team involvement (i.e. the status quo)

• How to handle tags with multiple meanings?

  This topic has no real solution, except manual retagging.

Answer 1

My biggest issue with ambiguous tags is that writing tag wiki for these is a torture.

* ^{For wma, am I supposed to write about Windows Media Audio format or about Java ME Wireless Messaging API?
  For cdc, am I supposed to write about USB Communications Device Class or about Change Data Capture in SQL Server or about Connected Device Configuration in Java ME?}

And even if I decide to cover all popular meanings in tag wiki, imagining how it would look like to reader makes me sick. Two, three, more totally different topics to cover at single page? give me a break!

As far as I can tell, proposed feature request helps to get rid of that headache which is why I am all for it.

Answer 2

As far as I can see, tags on these sites need a tuneup, and I don't think that this proposal goes far enough.

Commonly, bad questions are accompanied by 'a tag sentence'. Low-rep users just start typing away in the tag field, and discover that the 1000 most commonly used words in English are all existing tags, possibly including 'a' and 'is' on a bad day. As far as techo-jargon goes, I think that the tag box might be turing equivalent, containing as it does essentially all of the keywords of all commonly used programming languages.

At this point in, at least, stackoverflow's maturity, I wonder if it would make more sense to require high-rep action to bring a new tag into existence. When someone with a middling amount of rep used a tag for the first time, it would not be available to anyone else until a high-rep user or mod explicitly endorsed it.

Answer 3

As an active user of Microsoft Unity, I always try to protect the tag unity (which is firstly used for that product) from misusage.

Two key points that I think should be added to this proposal is,

Friend Tags for Each Tags For each tags, we should allow friend tags to be defined. For example,

• unity should have inverse-of-control, dependency-injection, ioc, and so on.
• ubuntu-unity should have ubuntu and so on.
• unity3d should have 3d, game, and so on.

Automatic Suggestions As each questions can have 5 tags, we can tell from the other tags whether the usage of unity is correct or not.

For example, if a user uses ubuntu and unity, he/she must mean ubuntu-unity, and we should generate a warning.

If a user uses '3d' and 'unity', he/she must mean 'unity3d', and we should warn too.

Only when the ambuguous tag is the only tag used, we display all ambiguous options equally. This is more convenient, as users can easily learn which option is most likely the proper one to use.

Answer 4

Re-edit: As was pointed out to me in the comments - There are some points that should be addressed about the proposal.

Possible Solution:

Require that Tags need to have a minimum 100 character description in their wiki to be used. All current tags that are considered ambiguous can have the Wiki entry replaced with 'ambiguous - use either A, or B'.

They cannot be applied to any posts until this happens, and an error that says soemthing like "users can't apply new tags without a description on the tag wiki. The tag 'stupid-tag' is new and/or has an insufficient length. Try using an existing tag instead, or writing a new description if you are sure this tag describes something different than those that already exist."

The creation of a Tag Wiki would be changed to be a 1500 reputation locked action. These locked tags could then not be applied to existing posts until a decision is made on how to make them non-ambiguous or they are left too short, so that they need to be abandoned.

People could, over time, make edits to these older, badly tagged posts to fix the tagging. Edits to old posts would not allow people to leave any newly non-acceptable tags on the posts.

Answer 5

Another solution might be to make it better known, or have a gui front end such that you can easily do a multiple tag query. For instance, the "table" tag is overloaded. If you just click on the button for "table", you will end up with posts that deal with different subjects, notably sql database tables and html tables. But if you go to the box at the top right and type

[table] [sql]

then you get a list of questions dealing almost exclusively with database tables.

Alternatively, if you type

[html] [table]

you will get a list of questions almost entirely related to html tables.

With multi-tag queries the problem of overloading goes mostly away so it should be promoted somehow.