Initialisms (DFT, CFD, etc.) are common in Science and Computational Science, so naturally tags that are initialisms may represent two very different subjects. It would be helpful to sort out some guidelines to include in the FAQ, even if there are few concrete examples as of yet.

What is our policy on ambiguous tags?

First an example: DFT. For computational chemists, this traditionally refers to Density Functional Theory. For applied mathematicians, this refers to Discrete Fourier Transform. So the tag , should it:

  1. not be used, and be replaced with an expanded tag.
  2. be reserved for the more popular subject.
  3. be modified to suit both subjects.
  4. invoke some special mechanism (to be developed by the SE team), for example the tagger will be asked to replace it with one of two choices.
  5. be none of the above choices.

The concern might be for novice question askers. They might want to tag their question with 'DFT', and they will use auto complete or the drop down list of suggested tags to help them find the right tag. If 'DFT' is not an option, what will be the result? Maybe this should resolve itself organically with more examples, but it might be worth having a strategy in advance.

There was a similar question on SO meta about this subject: Ambiguous tag: sts. There the suggestion was to replace 'sts' with 'sts-option1' and 'sts-option2'. The option1 and option2 additions can be long. In the above 'dft' example, this would mean 'dft-density-functional-theory' and 'dft-discrete-fourier-transform'. Of course both examples would fail because they are too long to be acceptable tags.

So what's the best approach? Should we just let things happen and deal with them when they do?

  • 1
    $\begingroup$ Actually, I proposed in the chatroom that all Fourier stuff be subsumed under fourier-analysis; that's the right mix of specificity and generality, I'd say. $\endgroup$ – J. M. Dec 4 '11 at 4:34

First of all, something as short and non-descriptive (to the non-specialist) as definitely shouldn't exist as a tag on its own, because it's easy to define a tag synonym for it. Posters will still be able to type "dft" into the tag text box in questions, but it'll get replaced with the synonym upon posting. Since the system will perform the conversion automatically, there's no reason not to standardize on the longer, more descriptive name. (Despite everyone's best efforts, a lot of people don't read the tag wiki excerpts.)

Now, for an ambiguous abbreviation which could mean two different things, it gets a little more complicated because you can't have multiple synonyms for one tag. But we should still definitely not have the tag exist on its own, because then people are going to use it and it'll make a mess of the semantic classification. What we can do is make use of the fact that when you're typing a tag into the text box, the popup that shows up will display any tags which include the string you've typed or whose synonyms include the string you've typed. So we can arrange for both and to appear when someone types "dft" by making sure each tag has one synonym which includes the string "dft".

Suppose DFT is more widely recognized as "discrete Fourier transform" (which would be my guess). We can then define as a synonym for , and something like as a synonym for . That way, when a poster types "dft" in the tag box, both and will pop up in the list, and the poster will be able to choose the correct one. But nobody will be able to actually submit a question using the tag .

As for the actual mechanics of it, tag synonyms can be proposed and approved by community members with sufficient reputation, but it takes 5 people with significant numbers of upvotes in the tag. It will probably be more efficient to build a list of ambiguous tags that need synonyms declared, and then when we get pro-tem moderators, they can just create them all directly.

  • 1
    $\begingroup$ I would be even more general and use fourier-analysis instead of discrete-fourier-transform... $\endgroup$ – J. M. Dec 6 '11 at 9:35
  • $\begingroup$ I did not know that synonyms are replaced with their parent tag; this is great and fixes a lot of potential problems. $\endgroup$ – Yann Dec 6 '11 at 14:05
  • 1
    $\begingroup$ I have no preference as to which tag "owns" the synonym dft, but I think the popularity of DFT with chemists/computational chemists is under estimated. For example, I've seen many academic publications simply use the initialism without expanding it, even if this practice may not be good form. An alternative synonym is fine. Theres a chance that it might not be what we initially intend it to be, because as more comp. chem. users join they might prefer quantum-dft or ab-initio-dft. I think for many of these ambiguous tags, we will just have to let things happen organically. $\endgroup$ – Yann Dec 6 '11 at 14:12

As the perpetrator of the DFT tag, I am of the opinion that the more descriptive name should be used, and unless we run up against character limits, then the acronyms should be relegated to the tag wiki. So, option 1.

Previously, on [theoreticalphysics.se], I had run into a character limit when using , so without testing I used . However, I just successfully retagged this question with , so the character limit isn't an issue.

(Truthfully, I believe I was not remembering the situation correctly, and I had been trying to use dynamical-mean-field-theory which is 27 characters while density-functional-theory is 25 characters.)

  • 2
    $\begingroup$ I agree, but what about short abbreviations with no as-of-yet recognized ambiguity? I would vote to leave them short until/unless an ambiguity arises. $\endgroup$ – David Ketcheson Dec 3 '11 at 17:38

We should fix overly-ambiguous tags. DFT should definitely be spelled out.

  • 1
    $\begingroup$ But isn't the point of the tooltip to give a little bit more details on the nature of the tag? $\endgroup$ – FrenchKheldar Dec 4 '11 at 3:47
  • $\begingroup$ DFT has several meanings, sure tootips can give more data but too ambiguous and users will use the wrong tag. $\endgroup$ – aterrel Dec 4 '11 at 5:12

I'm leaning towards 2, using DFT as Discrete Fourier Transform. A compromise could be "dft-fourier" and "dft-field"...


You must log in to answer this question.

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