# Usage guidance for software-specific tags

Our usage guidance for the tag is nice and clear:

Use this tag for any on-topic question that (a) involves R either as a critical part of the question or expected answer, & (b) is not just about how to use R.

This advice is not part of the text for the other software-specific tags: , , , , , etc - these just mention what the software is and who makes it. Would it be helpful to have text similar to the R guidance for these? I can edit them if this is worthwhile.

(I made an edit suggestion for one before realising it was a common issue, and that I should raise it here before editing several)

• I think yes. I'd keep a very brief definition in the tag excerpt though. E.g. for tensorflow: A Python library for deep learning, developed by Google. Use this tag... blablabla. – amoeba Oct 1 '19 at 14:49
• @amoeba Thanks - I've started to edit these since your comment seems to have a fair bit of support and no opposing views have been expressed. – mkt - Reinstate Monica Oct 2 '19 at 6:51

## 2 Answers

It is a good practice for the stackexchange communities to explain the tags by how they are used and not by what the specific term in the tag means.

See for instance the points 2 to 4 in the list in this blog post from Jeff Atwood

Tags are the de-facto map of allowed (and implicitly disallowed by omission) topics on your site.

...

1. Avoid generically defining the concept behind a tag, unless it is highly specialized. The “email” tag, for example, does not need to explain what email is. I think we can safely assume most internet users know what email is; there’s no value in a boilerplate explanation of email to anyone.

2. Concentrate on what a tag means to your community. For “email” on Server Fault, mention the server aspects of email including POP3, SMTP, IMAP, and server software. For “email” on Super User, mention desktop email clients and explicitly exclude webmail, as that would be more appropriate for webapps.stackexchange.com.

3. Provide basic guidance on when to use the tag. In other words, what kinds of questions should have this tag? Tags only exist as ways of organizing questions, so if we don’t provide proper guidance on which questions need this tag, they won’t get tagged at all, rendering the tag excerpt moot. Think of it as a sales pitch: in a room full of tags screaming “pick me!”, what would convince a question asker to select your tag?

This is a strong argument to extend the software related tags, explaining how they should be used, in the same style as has been done for .

• I agree that it can be useful to explain the use of the software tags in the excerpt (and that's why I added this answer besides my other answer). However, I have sort of split feelings about this. The phrase "(a) involves $x$ either as a critical part of the question or expected answer, & (b) is not justabout how to use $x$" doesn't look so pretty to me (which may lead to the internauts not reading it fully) and it feels that it should be more concise/shorter/simpler (although on the other hand it already seems short, and it is just the feeling that it is an ugly extension). – Sextus Empiricus Oct 3 '19 at 7:19
• I also feel that this addition is more as 'a disclaimer added in a sort of terms & conditions that nobody is gonna reed, but shows up in discussions when a user complains that his question got put on hold' rather than a 'guideline' that is gonna warn users on the spot. (are the excerpts visible when the tags are selected while composing a question? I do not remember, haven't made much questions myself. I am currently on my phone and can only check it there, and there is no excerpt displayed when adding a tag). – Sextus Empiricus Oct 3 '19 at 7:23
• Thanks for this second answer. I agree that a shorter formulation would be a good idea, if we can come up with one. – mkt - Reinstate Monica Oct 3 '19 at 8:23

Using this script I looked up some descriptive statistics for the specific tag names:

from beginning to 2019-29-09

Name         N      views   score    closed    answers   accepted  answer    time to answer
median  avg      rate %    avg       rate %    rate %    median [minutes]

r            20665  454     2.8       9.6      1.0       36.2      69.8     235
python        2614  325     2.3      12.6      0.9       30.8      63.9     234
spss          1628  474     1.8       7.2      1.0       22.2      66.6     155
matlab        1175  583     1.8       9.7      1.0       39.7      72.9     232
scikit-learn  1158  387     2.3      11.1      1.0       36.4      70.7     317
stata         1149  483     2.0       7.7      0.9       30.3      62.4     238
sas            614  397     2.0       8.5      1.0       31.1      67.1     190
tensorflow     445  256     2.2      15.7      0.9       33.0      65.2     527
excel          359  517     2.4      14.5      1.3       36.8      80.2     100
jmp             47  336     2.2       2.1      1.0       38.3      78.7     448

total site  235918  196     1.6      20.4      1.0       20.4      41.9     179
including
~ 1/3 deleted

• What we can see the easiest from these statistics is that the number of questions with the tag is much larger than the others

• However the amount of questions that get closed is smaller for the tag. And this might indicate that the others are lower quality or more often off-topic (or they are similar and just end up differently in the statistics for other reasons)

With this script you can also observe the number of deleted posts:

      Name    Total    Deleted   Deleted    Closed    % deleted
directly  after
Closing

r             38655    7387      10603      1988      46.5
python         5266    1004       1648       330      50.4
spss           3221     817        776       117      49.5
matlab         2656     702        779       114      55.8
stata          2298     434        715        88      50.0
scikit-learn   2166     511        497       128      46.5
sas            1312     292        406        52      53.2
tensorflow      976     185        346        70      54.4
excel           813     136        318        52      55.8
jmp              79      22         10         1      40.5

total        235918   52184      38450      9645      38.4


Which shows that the tags are similar in having around half the questions deleted, whereas the entire site has 38.4% of the posts deleted.

I would say that based on the above numbers one might argue that for it might be valuable to add a similar description as with the tag. The number of questions are still reasonable large and the number of closed question indicates that the questions can be problematic.

However for it doesn't seem to be a problematic tag, and tags like and others do not occur that often. With tags like , or , which have high close rates, one might also wonder whether it is due to the issue of being software specific or whether it is due to other reasons (as these tags may correlate with other issues, like being broad unclear or self-study).

• Thanks, it is helpful to see this data. I agree that it's hard to know whether such edits would materially change usage behaviour. Unlikely to have more than a minor effect, but I'd take a small improvement for such little effort on our part. – mkt - Reinstate Monica Oct 2 '19 at 15:20
• My argument is not to balance the small improvement versus the little effort of changing the tags. But to balance it with keeping the tag descriptions simple and... descriptive. The part (b) of the description is now more about how to use the site rather than how to use the tag. For the cases 'r' and 'python' this seems fine because it may help a large amount of off-topic questions and most importantly reduce the meta discussions about it by people that refer to the presence of the tag as an argument that the site invites questions about r. But do we need this in every other tag? – Sextus Empiricus Oct 2 '19 at 15:42
• I don't think you can infer anything much about quality from these data alone; still, interpretation is fun. One tag where I have some affiliations which you don't use is stata. Questions specifically on that often get closed because various of us are vigilant in advising posting on SO or more commonly Statalist. I think that is likely too with some of the other software tags. The extent to which very many high reputation users use R routinely makes them likely to look at questions with R code and/or illustrate answers with R code. There is a feedback loop there. – Nick Cox Oct 2 '19 at 16:16
• Conversely, my prejudice is that statistically clueless questions are more likely with software X, Y and Z. (Sorry, I have to run before I can spell out the details.) – Nick Cox Oct 2 '19 at 16:18
• I have added 'stata' now. I had a look at the most popular tags of the last year. Either 'stata' has lost popularity or I overlooked it. – Sextus Empiricus Oct 2 '19 at 16:23
• Thanks for adding stata. No offence implied or inferred! – Nick Cox Oct 2 '19 at 17:46
• "But do we need this in every other tag?" No, of course we don't need it. But my impression, based in part on this answer by gung was that excerpts should offer usage guidance where helpful, and I think this meets that goal. FWIW, several of the excerpts previously contained a great deal of unnecessary detail about the programs, so I don't think these changes make them any more complex. – mkt - Reinstate Monica Oct 2 '19 at 17:46