What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 131 Stack Exchange communities.

I have found myself repeatedly answering 'R' questions that follow the formula "reshape your data like this, feed it into ggplot like this, look at this graph." There are some very good reasons why this question is being asked multiple times, the primary of which is the fact that while the 'ggplot' package's syntax is well-documented, its use is not.

'R' users frequently have trouble combining plots or convincing 'ggplot' to auto-grouping their data, and I don't mind answering the question multiple times. However, once the question is closed, all of this duplication waters down the ability to find a suitable answer on Stack Overflow, and we end up answering the question again.

I wonder if we can find a way of establishing a reference guide for certain topics that come up repeatedly and elevating that guide in Google/Bing/DuckDuckGo search results. I would love to help curate an ongoing reference page on the use of the 'ggplot' function using the questions and answers we have available on Stack Overflow. Once incorporated into the reference page, it would make sense for those answers to redirect to the reference page rather than the original question, so we could elevate a good set of answers for people to find that would provide a more complete and useful answer.

Update 2013-02-06

To make my criteria for a good solution more discrete, I will attempt to list them here along with the reasoning:

  1. The set of good questions and answers should be clearly visible above the "noise" of similar questions and answers when using an internet search engine. Otherwise, internet searches lead to Stack Overflow without leading to a good answer and results in a new post of a similar question. Currently, if I search 'ggplot multiple plots' on Google or DuckDuckGo, a top result is a 1-year old question (Q: multiple graphs in one canvas using ggplot2) that is a rather uninformative question with a rather uninformative answer. No other Stack Overflow results appear on the first page of the search results.

  2. The set of good questions and answers should be curated and well-explained to address the broader need of the community of users. In the case of R's 'ggplot2' package, many users ask about how to combine a set of data into a single plot when the answer is the use of the 'reshape2' package to make the data easily auto-grouped from within the 'ggplot' functions. The correct answer is not the direct answer to the question being asked.

  3. The solution should reduce the "noise" of similar questions within Stack Overflow itself. We ask our users to search for similar questions before posting, but when I search 'ggplot multiple plots', I get 121 results that are all asking something very specific (but usually end up with very similar answers). If users don't quickly see a question that is an exact match, the result is a new post of a similar question.

There have been several suggestions in the comments and in an answer, and I will summarize what has been contributed so far:

  • Make minor edits to an existing question to make it more broadly applicable. This is a good suggestion if I can find a question that can be made broadly applicable with minor edits. If the appropriate question can be found, this addresses criterion 2. The problem is that the question will still be only one among a sea of similar questions and will not address criteria 1 and 3.

  • Add an entry in the tag wiki for 'ggplot'. This also addresses criterion 2 by allowing the entry to be completely customized to suit the needs. However, I have in my time never seen a search result link to a tag wiki, so this does not appear to address criteria 1 and 3.

  • Just look at the related questions at the side. This does not appear to address any of the criteria.

  • Write the "perfect" question and give a "perfect" answer. I like this idea, and this again addresses criterion 2. I could write a good, broadly applicable question with a solid canonical answer and get some others to participate on the question. I am unsure how this addresses criteria 1 and 3, though, since the answer would not be particularly elevated above the rest of the "noise."

  • Merge questions as exact duplicates. Unfortunately, the questions I am talking about are similar and not exact, so I don't think we should merge them. I still think we should address people's immediate and specific needs by answering the questions they ask. I am proposing adding a new layer of information to help build a reference of all the useful, related information to a common set of related problems, so we can write some good comprehensive answers that apply to a variety of related problems.

The solution that I picture is more like this:

diagram of topical reference page

There are several of us who typically answer the 'ggplot' questions, and we could be wrapping these questions into an larger reference to help people with the standard 'ggplot' problems. As subtly different questions emerge, we can answer them in the usual way but then roll the answer into the reference page. I would expect the reference page to also be elevated in both external and internal search results, so people see the reference page first before asking a question.

share|improve this question
1  
Sounds like content for a tag Wiki perhaps? Although I don't know what the search engine visibility of those is. (I've personally never stumbled on a tag wiki as a search result). –  Bart Feb 5 '13 at 15:57
    
It seems that in your critique of my question you took each piece and stated how that piece doesn't address everything you want to do. The entire point of my suggestion is that you don't just do one of those things, you do all of the steps I provided, and as a result of doing all of them in combination it covers all of the issues you've brought up. Also note that if you create a canonical question and it gets a lot of votes and is used to close a lot of other questions as duplicates it will end up with lots of Google juice, as well as reducing clutter (through dup closure). –  Servy Feb 6 '13 at 15:30
    
@Servy, I'm not rejecting your suggestions without consideration. I'm simply evaluating them and explaining why I think they do or do not provide a suitable answer, so you and others can participate fully. Duplication closure doesn't fix the problem, since the questions are not duplicates but rather outcroppings of the same fundamental problem. I don't think mergings or closings are the answer, because people still need an answer to their specific question. I would welcome some others' input on whether or not my question needs more clarification or needs restating. –  Dinre Feb 6 '13 at 16:36
    
@Servy, also, the questions I see appearing in Google, DuckDuckGo, and Bing are actually questions with low votes, so I don't think high up-votes will address the criterion of elevating the results in search engines. Try searching 'ggplot multiple plots' as mentioned in the post, and let me know if you can tell why those questions are elevated in search. If we can hack the system to elevate specific questions by doing something simple, that may solve part of the problem. –  Dinre Feb 6 '13 at 16:39
    
@Dinre Clearly the questions are not duplicates of each other, but if you make a new more general question that addresses the underlying problem that all of these questions have in common you could indeed close them all as a duplicate of that question. I never suggested merging, as that's clearly not appropriate here. –  Servy Feb 6 '13 at 16:39
    
@Dinre And are there any questions in those areas that have lots of votes/views and that are the dup target of a lot of questions? The views matter a lot, and being the dup target means that you'll have lots of pages linking to that question. Such canonical questions also tend to be linked a lot in general in comments and outside sites, all of which improve the search ranking. It won't happen overnight, but it will happen if the question is actually of very high quality and is valuable. –  Servy Feb 6 '13 at 16:41
    
@Servy, from what we know about SEO, Google does currently base page-rank partly on outside links and page views. Stack Overflow as a site does have some control over how pages are indexed, though, and we could perhaps build the reference page and then remove the individual question pages from indexing. Or there may be other ways we can elevate a reference page over less useful question pages to improve page visits. I wrote this feature request in an attempt to encourage improvement in how we serve our communities and not as a complaint. Are there other suggestions? –  Dinre Feb 6 '13 at 16:55

2 Answers 2

You don't need any new features to do this; you can do it with all of the tools already available to you.

  1. Look through a number of the duplicates on this topic for one that is particularly well asked, this is clear, and somewhat more general than the rest; something less localized. If you can't find any and all that you can see are either poorly asked, or not overly general, then ask your own question. (Make sure that if you do this you spend a noticeable amount of time crafting the question so that the question is of high quality, or it will just be closed.) If you can make a few minor edits to an existing question to make it a bit less localized, or just of high enough quality to work then feel free to do so.

  2. Answer the question that you have either chosen or written. Spend a lot of time writing a high quality answer, format it well, go into depth, provide multiple examples and links to external content.

  3. If you feel that other members of your tag's community will be able to improve the content of the post then consider marking it as community wiki to encourage others to improve or add to the answer to make it "the community's answer" rather than just your answer.

  4. Favorite that question so you can find it easily.

  5. When you see similar questions use that as the target of the exact duplicate closure.

share|improve this answer
2  
...and put a link to the post in the Tag Wiki. –  Robert Harvey Feb 5 '13 at 16:11
    
@RobertHarvey I would only do that if it applied to a large percentage of the questions asked in that tag. Most likely it wouldn't apply to a tag like R, but it may very well be worth adding to the ggplot tag or some of the other more narrow that tags questions like this ought to be tagged with. –  Servy Feb 5 '13 at 16:12
    
This doesn't fully resolve what I want to do. I would like to "bury" all of those related questions, so the set of good questions and answers can be elevated for search. For something like plotting, multiple problems and answers are very helpful in learning to work with the functions. As long as we have the mound of related questions, I think we will continue to answer similar questions indefinitely. –  Dinre Feb 5 '13 at 16:16
    
I assume that, if the question is common enough to require collapsing the duplicates, it probably justifies a link in the "frequently-asked questions" portion of the tag wiki. See the tag wikis for PHP and C++ for guidance. –  Robert Harvey Feb 5 '13 at 16:17
    
@Dinre That's why there's a "Related" section on the right hand side of every question with lots of other similar questions. You can also include links to several other high quality examples in your answer if you feel it's appropriate. Personally I'd much rather have links to 3 or so answers I felt were very good than a dump of dozens of answers, many of which are mediocre at best. –  Servy Feb 5 '13 at 16:19
    
@Dinre: You're describing what we call a "merge." If there are a flock of exact duplicate questions that need to be merged into a single question, make a request here on meta. But generally, these questions should link to the canonical one anyway, and different variations of the question means that folks arrive at the same canonical question from different search results in Google, which is OK. –  Robert Harvey Feb 5 '13 at 16:20
    
This all sounds close to what I'm thinking but not quite there. The questions I am thinking of are related to a bigger picture and are not exact duplicates. People have trouble learning to use the 'ggplot2' package correctly, so we see a large number of questions that are relating to its basic use. Sometimes, people don't even know that the question is actually about 'ggplot'. It sounds like we have a number of mechanisms already built into Stack Overflow, but no one of them seems to directly solve this problem. Is SO perhaps not the right place to give the robust meta-answer? –  Dinre Feb 5 '13 at 18:19
    
@Dinre No one single feature does exactly what you're requesting, no. My point is that using a number of existing features, in combination, solves the problem that you are proposing in a similar, but slightly different manor. I'm saying we don't need this proposed feature as a way of addressing this problem, because we already have all of these other features that can be used to solve the problem. –  Servy Feb 5 '13 at 18:24
up vote 0 down vote accepted

I'm placing this answer here for the community to vote up or comment on if they think it's the appropriate answer to my question. Please feel free to edit it.

Q: Can we find some way to collapse similar questions into a canonical answer set with multiple examples?

A: Stack Overflow really isn't the place to aggregate answers into a guide. We ask and answer questions, and it's okay that we have a lot of similar questions and answers. The best approach is to take cues from Stack Overflow questions as an indicator of what content people need/want, repackage SO questions and answers along with your own content into a reference guide (giving credit where due), and publish it elsewhere on a blog, website, or in print. That way, you and others can reference it as a resource when answering similar questions in the future.

share|improve this answer
    
It's looking like this is perhaps the most appropriate answer to my question, since I haven't received any answers other than a collection of half-solutions. I'll leave this open for just a little bit longer. –  Dinre Feb 26 '13 at 0:10

You must log in to answer this question.

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