Exploring “Related Work”

 Suggested Guidelines for Finding Materials to include in the
“Related Work” Sections of Conference Papers

 

These are my personal suggestions and not sanctioned by any other individual or organization.  Some of this material may be controversial.

In no particular order.

If you have a particular conference in mind:

  • Look through the proceedings for the past two years or so of that conference. If the work there is not relevant, then you are probably planning to submit to the wrong conference, find the right one.
  • Sometimes an advance program for an upcoming conference is available on the Web, and you can find the camera-ready or earlier technical report versions of the not-quite-yet-published papers on the authors’ websites.
  • Read the papers cited by the papers from recent proceedings, directed in part by what the authors say about the papers they cite – but do not assume their comments are an accurate reflection of the cited paper’s content.  Then read the papers that those papers cite, etc.
  • Use citeseer to find more recent papers that cite the proceedings’ included and cited papers. Then find the even more recent papers that cite those papers, etc.
  • Search for relevant papers written by the program committee chair(s) and the members of the technical program committee, possibly for past instances of the conference as well as for the instance to which you plan to submit. Work by the general chair, publicity chair, local arrangements chair, etc. are less likely to be germane.
  • Read the papers cited by those papers and use citeseer to find more recent papers that cite those papers.

In general:

  • It is inappropriate to cite papers you have obtained only because you or your colleague were asked to review that paper, unless the authors have already made that paper public, e.g., as a technical report. If they have not done so, you can notrefer explicitly or implicitly to the information in that paper.  However, you can try to track down other relevant published or posted work by those authors (or a subset of the authors).
  • Papers published before 1997 or so may not be posted on the Web and thus not indexed by google, but they still exist. Go to your library. If you are lucky your library will have online search capabilities for at least the authors, titles and keywords of its collection, with possibly entire documents for major portions of the collection scanned in.
  • Search the digital libraries of acm.org, (IEEE) computer.org, and any other professional societies relevant to your field, where you may find papers that are not accessible to google web crawlers.
  • When tracking down additional work by an author, you may find that the senior author’s website has not been updated since 1999.  Junior authors, particularly students, are likely to have more up to date publication lists on their websites.  Also look for lab or group websites.
  • Do not cite from a cite, the citing paper may be inaccurate, go to the source.
  • When there are workshop, conference and journal papers presenting the same body of work, a journal citation is usually preferable over a conference and a conference over a workshop.  However, keep in mind that if the workshop, conference and journal were all published on the same date, chances are the workshop paper reflects new work done in the past 6 months, the conference paper work done 6-12 months ago, and the journal paper work completed a year or more earlier.
  • Google may not always be helpful in finding the most relevant work, e.g., if you choose different terminology than the authors.  If you know a particular organization is doing work in the field, go to their website, read their papers, read the papers they cite, read the papers (from citeseer) that cite them.

When writing your related work section:

  • Citing papers more than 2 years old is very appropriate when referencing seminal work or to indicate the earliest known results on a topic.
  • Citing papers more than 2 years old is usually not appropriate when comparing your results to the state of the art.  Use citeseer to find more recent papers citing those papers.  Track down the websites of the authors and see what they are doing now – it is unlikely that a research group would publish a paper in 2001, which means it was written in 2000 or earlier, and then totally drop out of the field doing nothing since then.
  • Actually read the entire papers that you cite, do not assume that the title and abstract accurately reflects the content of the paper.  If you intend to compare to the FooBar system, make sure that the cited paper actually includes the information germane to your argument, rather than focusing on some unrelated aspect of FooBar.
  • Do not be shy about contacting authors to ask what is the “best” or “earliest” paper to cite about the xyz issue in their FooBar system. Although you are probably more likely to get a response from junior authors than from senior authors, if there is a conflict the more senior author is usually more likely to be accurate.
  • If both a technical report and a published version of a paper are available, cite the published paper unless the technical report includes details omitted from the published paper that are important to your discussion or if you are trying to indicate an early date by when the results were first achieved. Or cite both if space permits.
  • Use complete citations, with correct author names and order, paper title, forum of publication (e.g., workshop, conference, journal, book, website), date, URL if available.  Many publication forums have their own specialized citation formats, which may or may not permit URLs, however all of them include authors, title, publication forum, year.
  • Record the complete citation information for a paper when you first find it, this information may not be included in your printout or online copy of the paper.
  • When discussing related work (and your own!), use generally accepted terminology from the field, not the peculiar jargon that has arisen in your lab.  When employing the latest buzzword, explain in conventional technical terms what it means.
  • Do not over-cite your own work, your own previous work should be a very small percentage of the cited papers.  Citing your own prior papers should be particularly minimized for conferences intending “blind review”, since even if you remember to avoid saying “we” did this and instead say “Smith et al.” did this, you are not fooling the reviewers, particularly when you are referring to an earlier version of your own system.

When writing/research in general:

  • Do not assume your advisor or some other authority invented a concept just because you first heard about it from him/her.
  • Do not assume you were the first person to invent a concept just because you cannot think of anyone else with the same idea off the top of your head.
  • Do not assume the authors of a paper have actually done the work mentioned in their “Future Work” section, but on the other hand do not assume they haven’t. Find out!
  • And …drum roll… It is strongly preferable to read related work before you do your own work, certainly before you start writing your paper, rather than as the last part of the paper to be finished a few hours before the submission deadline.

More coming as I think of it… Suggestions welcome.

last updated May 21, 2004.
Copyright © 2004 Gail E. Kaiser.

The original link was broken (http://www1.cs.columbia.edu/~kaiser/relatedwork.htm). This page is a clone backup.