30

I have been working on a research paper to apply a technique X to solve certain problem Y. The field is in computer science. Approximately some months ago I started to check it up if there was something similar to my proposal, but I did not find anything at all.

A couple of weeks ago and when I have already finished the tests for my proposal, I decided to check it up again. In my new search, I found an article that was published about two years ago and when I read it, I saw that it was very close to the idea on which I have been working. The differences were very subtle, I was using a simplified version, data was collected in a different manner, and other tiny differences. If we talk in percentages, the differences would be like 20 % between my article and the one I found.

The reasons that I did not find this article in my first search were that it was published in a not so well known journal, and while it has been cited before, it were only self-citations and there was no strong relation between the article and the citing ones.

So what I can do in this situation? I have not based my work on this paper, but I know that I must cite it like a related work. Should I drop my paper or present it and see what the reviewers have to say about it? I would not like to be pointed as a case of plagiarism because of the similarities.

  • 8
    In this situation (which is not that rare, particularly in active areas) one has no choice but to drop the publication of the current article. This situation occurs because after having worked something out one understands the literature, and how to search it, better than one did before working the thing out. But there's no publishing something already published (unless methods are very different), and feigning lack of awareness of the already published article would be dishonest, morally like plagiarism. All one can do is use the results obtained as a building block for something more. – Dan Fox Apr 9 '15 at 8:35
  • 4
    You are way too honest for the academia. I think the best you can do now would be: write up you paper, cite the similar one (let's call it X), write a couple of sentences like "X is a very interesting appraoch", but "My version is simpler" (and thus easier to use, I hope?), "data collection is different" and so on. If the reviewers say you have enough contribution, then you get a paper. According to my experience in CS, the reviewers will hardly bother to read the "related work" section. – user12956 Apr 9 '15 at 10:36
  • 4
    @user12956 You obviously know nothing about CS, so "the reviewers will hardly bother to read the related work section" is as silly as it can get. a) We do READ related work b) We do KNOW related work if it is on our area of expertise. – Alexandros Apr 9 '15 at 12:34
  • 12
    @DanFox This situation occurs because after having worked something out one understands the literature, and how to search it, better than one did before working the thing out. A bit like the phenomenon where you start writing a question for StackOverflow and in doing the research necessary to write a good question you realise you've found out the answer, even though you thought you'd already researched... – starsplusplus Apr 9 '15 at 13:09
  • 4
    You are way too honest for the academia. — [citation needed]! – JeffE Apr 10 '15 at 0:27
34

First of all, you are absolutely right that you need to cite the paper. Citations should not be to "previous work you based your work on" but to "the state of the art" - whether or not you explicitly based your work on a previous article. The reader should be able to place your work in the context of what is already known on a topic.

Second, giving any advice will be hard for us, since we don't know the specific situation. If all you add to this previous article is a simpler design and collecting data in a different manner, then your paper likely does not offer a publication-worthy contribution to the state of the art.

Your best bet is likely to rework your paper heavily. Identify weaknesses or open questions remaining in the previous article and address these. Then you can build on the previous article and expand upon it, and you will have a genuine contribution.

Of course, this will be a lot of work, and much of what you already did may turn out to be wasted effort. That sort of thing unfortunately happens in academia. You will need to analyze your specific situation to find out how much you can salvage.

The positive side is that you have found someone who can meaningfully review your work. And if, as you write, their work has so far mainly been self-cited, they will appreciate some external attention on it, so if you get them to review your paper, you may have a sympathy bonus. Alternatively, you could even contact the authors of the earlier paper and see whether they would be interested in collaborating with you.

20

It's not plagarism to come to similar conclusions to someone else, as long as you're open and honest about everything plagarism isn't the issue. Since you know about it you'd definitely have to cite it. As long as you don't claim to be the first to have discovered it then citing the earlier paper and confirming that you found similar results before encountering it is, if anything, positive for the author of that paper. They get cited and an independent researcher, you, has confirmed their findings.

You did the work, ethically you're in the clear as long as you're honest.

I disagree with the other answer in that I don't believe that such repetition is "wasted" to other researchers. Lots of methods/technique papers gloss over weaknesses or "just happen" to use ideal datasets or don't mention the things that make the algorithm/technique unworkable in the real world and sometimes another paper from someone doing the same thing who's more open about such weaknesses can be invaluable.

Politically on the other hand it can be harder to get published. Since it's so similar you're less likely to be adding significantly novel data to the field. It's less sexy but confirming previous work is of value. You're likely going to need to be more rigorous about it than if you were showing something completely new.

  • 2
    This is potentially true, but it depends whether the new paper is adding useful value by confirming/reproducing claims in the earlier paper, or is simply duplicative. The former is more likely for an experimental area (where additional experiments that confirm and reproduce earlier results are valuable). The latter is more likely for mathematical/theoretical area (if they show an algorithm and prove a theorem about it, and you do the same, there's not much additional value to you doing the same thing a second time). The question doesn't have enough information to tell which case we're in. – D.W. Apr 9 '15 at 20:47
  • 1
    Obviously it depends exactly on the topic matter, but reproducing work is an essential and underratedpart of any scientific process, IMO. Actually also the work can probably be considered novel in its own right if the process to the result is different (e.g. in maths, a different proof for the same theorem can be considered as a contribution). – user1207217 Apr 9 '15 at 23:58
  • @user1207217 A second proof of an existing theorem is only a useful contribution if it is somehow better than the original one. Better could mean shorter, simpler, more rigorous or better by some other metric which may well be completely subjective. In experimental sciences, there is benefit in confirmation: knowing that something happens ten times in a thousand is better than knowing it happens once in a hundred. But a mathematical proof is supposed to be a fully convincing argument: there's no additional benefit in being convinced a second time. – David Richerby Apr 10 '15 at 13:32
  • 1
    @David Richerby Since the OP talks about data I'm inclined to think that it's either experimental or testing algorithms rather than pure math hence it's likely still valuable. I've read enough CS papers outlining an algorithm which quietly neglected to mention why they happened to be using data that had certain properties. – Murphy Apr 10 '15 at 14:35
  • @Murphy Agreed. I was addressing the specific comment, rather than your answer. – David Richerby Apr 10 '15 at 14:49
7

In general, all papers tell a story: "Here's this important problem. Here's what's been done about it. But what we don't know is X, Y, Z. So here in this paper, I address X, Y, and Z."

Do your best to frame your introduction to acknowledge this previous paper in the "here's what's been done about it" section, and set up your novel 20% as the "X, Y, Z". Don't forget to explain to your readers why that "X, Y, Z" is important and useful to others. Your project may be perceived as a relatively incremental advance, and therefore may not appeal to a top publication. But it may still have value to a specialist publication. If you're already done with the computational work and in the process of writing up, you might as well give it a shot.

If it gets rejected for not making enough of a novel contribution, then, based on the responses, I'd consider whether it's worth following Stephan Kolassa's advice, and doing more work to more explicitly establish your project as an extention of the earlier paper. It might be. Or you might find that your time (and excitement) are better devoted to other projects in your pipeline.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.