Sunday, September 30, 2007

"Related Work" vs "Prior Work"

(ed: As deadlines draw near, I get increasingly caught up with the minutiae of paper-writing. Be warned)

"Related work" or "Prior work" ? I've seen both section titles used when describing all the work that came before the magnificient specimen of research that I'm reading right now. Personally, I prefer 'Prior work': it seems somehow closer to the spirit of 'I stand on the shoulders of giants', even if that was really just a dig at Hooke. 'Related Work' to me has the tone of a grudging 'yeah yeah, these guys did something that might vaguely resemble what we do here, and I guess I should include it, but don't imagine that their work in any way influences my brilliant thoughts, which are sui generis'.

An issue that appears to be related to this is the placement of the Related/Prior Work section. I personally prefer to have prior work discussions just after the introduction, unless discussing prior work requires the use of notation that I introduce in a definitions section. In many papers (and some of mine as well), the prior work gets shunted off to the no-man's land before the conclusions. I wonder if there's some correlation between the choice of the title, 'Related Work' and the banishing of the section to the very end, a kind of double dismissal of the value of the cited works.

7 comments:

  1. I call it related work, because that way I can cite more stuff on other problems as long as I can find some excuse for calling it related. Prior work to me implies that it really has to be on the same problem. As for where to put it, I agree with you that in or near the introduction is best, rather than burying it at the back of the paper.

    ReplyDelete
  2. I also think that "Related Work" is a superior title mostly for the sake of accuracy and the ability to cite a wider range of papers and still be "on-topic". As for placement, I generally think that Related Work actually works better towards the end because it is otherwise quite distracting. For one, you're in a better position to understand what the related work is about (and how it is different from the current work) only after you've read the current work. Moreover, when it's at the beginning, if i'm reading in order (which I sometimes but not always do), I can get overwhelmed with various things before I even get to the point of the paper. This way it just doesn't get in the way. I think it depends a lot on whether you're reading a paper to see if it's any good (related work early is good, sets the tone) or because you know its supposed to be good and you're just trying to understand it (related work a little distracting at the beginning).

    ReplyDelete
  3. Personal pet peeve: why the need to have a separate "related work" section at all? I find it much more pleasant when relevant work is mentioned in an integrated way throughout the Introduction.

    ReplyDelete
  4. Personal pet peeve: why the need to have a separate "related work" section at all?

    So that experts in the field can easily skip over it. This is particularly important when the related work section is rather long.

    ReplyDelete
  5. I have strugged with "what to put in the intro" vs "what to put in the related work" myself, but I agree with anon 2. In the intro, you want a nice flow leading up to your main contributions, and too much related work distracts, especially when you only wanto to cite what I guess David would call the "Prior Work".

    ReplyDelete
  6. I tend to have two separate sections: one called "Background" and one called "Related Work".

    The "Background" goes after the introduction and lists and discusses extensively the prior work directly related to the methods described in the paper. The paper either builds on these methods, or proposes alternatives that fix their weak points.

    The "Related Work" goes at the end, and tends to cover related papers that have connections to the presented work but would be too distracting to discuss early on.

    Plus, when at the end we can explain what properties differentiate the "related" work from the work described in the paper. We can also point to experimental comparisons, explaining that "in Section X, we showed that our method works Y% better than the work by Z"

    ReplyDelete
  7. Are you trying to draw out the relationships, perhaps illuminating new connections, or are you trying to give a history lesson? I prefer Related Work to Prior Work, as it gives you more freedom as to what to include or not.

    I agree with Panos, although if I can't find a way to work a paper into the beginning background section, I'll either not cite it at all or perhaps mention it as part of an open problem or something. Related work at the end seems pointless (for my kinds of papers).

    ReplyDelete

Disqus for The Geomblog