11

The same fundamental question about combining data keeps being asked and answered again and again. This is a waste of (human) resources which could be better spent on more unique problems. I am referring to: (4332) and its many duplicates.

Today I took the step of renaming the original with the hope of making it more clear but I fear that is not enough. I had imagined that a search would find at least one of those many duplicates which if not answered satisfactorily itself would have lead to this network of Q&A with its two dozen answers. Clearly this is not happening. Not only are people not finding an answer before (re)asking the question but apparently those who answer cannot recall any of these duplicates to provide a link.

I see this as a failure of site organization on my part and I seek guidance in correcting it.

How can we prevent a question like this one from being asked and answered again and again?

8

My thoughts:

You can't stop people asking the question

Some will just lazily ask without searching at all. Some will search but not find, because no matter how crystal clear you make the title of the original, they will be thinking of the problem in different terms ("matrix operation", "lump data", "interpolate duplicate abscissa values" etc)

You can't stop people answering the question

New users won't know that the question has been asked before. We certainly don't want to discourage new users from answering, so we probably just have to live with this. However it's frustrating to end up with a collection of good answers spread among multiple copies of the same essential question. Perhaps a policy of withholding upvotes subject to moving the answer across might help? "Great answer, but the question is a duplicate. If you delete this answer and repost it here I will gladly upvote it"

Of course the issue of people answering duplicates can be reduced by quickly identifying them as such and getting them closed ASAP, which brings us to the most important point:

Finding duplicates is hard and unrewarding

The issue of rewarding users for finding duplicates has been raised before and presumably dismissed by the SE team as (unless I'm mistaken) there are still no imaginary internet points on offer for finding duplicates. So the problem is how to make it easier to find duplicates. Here's my half-baked suggestion:

A list of "originals" for commonly asked questions, sorted by the answer

What I mean by that is that while there are usually many many ways to ask a question, there is often one key function that holds the answer. For the question under consideration, the majority of answers use GatherBy. So the list would have the link to 4332 under GatherBy.

The list would look something like this:

OP is looking for:

GatherBy

NumericQ

ParametricPlot

  • (for plotting y against a function of x) some link

ReplaceAll

  • (to extract values from rules returned by Solve etc) some link

Table

Transpose

Hopefully you get the idea. I believe a list of commonly duplicated questions ordered like this would make it much easier for experienced users to quickly find the originals.

  • Your point of sorting by the answer seems good. One could also think about a way to reward more finding duplicates? – chris Oct 4 '14 at 7:28
3

What do you all think about creating more canonical lists of recurring problems, similar to 18393?

There could be one for common list manipulation tasks, and perhaps others if the need is there. Each answer in such a Q&A would contain a statement of the problem that is general (unlike most real questions), links to where this problem has previously been answered and if there is an obvious standard way of solving it then the answer could also include that. Or possibly it could include the most standard way of solving it and then refer to the duplicates for more exotic options.

I suggest this because I think that by stating the problems in general terms and collecting them in one place they will be much easier to locate. I am lazy when it comes to looking for duplicates, but even I refer to 18393 when I suspect the answer is in there.

  • 4
    That question is quite easy to remember because it refers to a certain "user profile" instead of a certain "problem". Classifying knowledge by type of problem, assigning meaningful names, selecting the best representative questions & answers, etc is nearer to writing a book than to participating in a site like this one. It requires a lot of highly focused effort and perseverance. I don't think many users here can afford the time (or have the will) to do it. Another model is "alla Wikipedia", but again that requires a highly organized (eventually distributed) governance. TLDR It's a lost cause. – Dr. belisarius Sep 29 '14 at 17:29
  • @belisarius the c++ tag on SO has a string of canonical answers to combat this problem. Yes, it is time consuming, and akin to writing a book, but it may be the only thing to do. – rcollyer Sep 30 '14 at 2:35
3

I think editing titles to give more accurate description of the real problem posed in the question is likely to help. This is the reason I already put some effort into editing question titles. I think most of this effort goes for nought, at least as far as those who ask questions are concerned. The software that suggests similar questions can't really match up anything very useful to a title like "How to solve my equation" (this is the kind of title I tend to rewrite). But I do it anyway, because I think it will help our corps of frequent answerers when they search for duplicates.

Another thing that I have thought about, but have not had the courage to implement, is editing an original question and its duplicates to add a set of key words and key phrases that would increase the probability that related questions turn up when a search for duplicates is made. My lack of courage stems from not having the confidence that I can come up an adequate set of keys.

2

What if we had a tag that is added at some point after a particular question has received a variety of well-thought-through answers and provides a meaningful complement or expansion of the documentation? This approach may address the problems that community members have:

  • Experienced users wasting time looking for duplicates - a question that is asked/answered multiple times on the site would qualify for having one of those questions marked as
  • Beginner users could be encouraged to include in their search. Granted, this solution doesn't address new users who simply don't search at all, but I wonder how many of the new questions arise from an unsatisfactory search.
  • Experienced users finding the search for duplicates 'unrewarding' - while not a direct benefit, perhaps the community could agree that are not generally CW questions and directing new users to the Q&A's (that they have likely contributed to) will provide some rep benefits.

The tagging system, in my opinion, is an untapped resource for simplifying the organization and indexing of content. As has been alluded to in another answer, there are "users" and there are "problems". If I have a problem, there is an easy place for me to go for solutions, but what if I'm interested in really getting a handle on or ? We might benefit from thinking about how the tag system can be used to classify questions from a 'users' perspective rather than a 'problem-type' perspective.

0

The duplicates are reguraly put on hold and later deleted. This solves a part of a problem, since an important part of this problem is that the questions became too numerous, difficult to search through. However, one of my problems is the wording of the questions placed by others in which I often cannot recognize my problem.

What if (before deleting a duplicate) we include its title somehow into the data base, such that it would appear together with the title of the originally asked question? There might be several of such titles to the same question, no problem. Could it be enabled practically?

You must log in to answer this question.

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