A good tutorial on publishing good research

Eamonn Keogh gave a tutorial at SIGKDD 2009 entitled "How to do good research, get it published in SIGKDD and get it cited". Someone forwarded the slides to me. I recommend them to you.

Particular ideas that I noted are:

  • Have your paper written well in advance of the submission (this is necessary for the following two steps).
  • Send a preview copy of your paper to authors whose work you are extending or refuting.
  • Put your paper through a mock review process.
  • Do your utmost to make your results reproducible. For instance, make datasets and high-quality code readily available.
  • Provide auxiliary resources, such as tech reports that go into subsidiary material in more detail (fuller literature review, more exhaustive experimental results).
  • For double-blind reviews, make supporting materials available anonymously online.
  • Gather real data and make it available -- this can enormously boost your citations.
  • Don't believe a result just because it has been published, and don't cite a paper just because someone else does (a number of telling and amusing examples are given in the slides). Reproduce the former, critically read the latter.
  • Techniques with lots of parameters are suspicious. An amusing quote from John von Neumann: "With four parameters I can fit an elephant, and with five I can make him wiggle his trunk".

The tutorial is quite long, but informative and entertaining.

Update 9/7/2009: Eamonn will be in Melbourne as a keynote speaker for AI '09, which runs December 1st through 4th this year (thanks to David Newman for pointing this out).

6 Responses to “A good tutorial on publishing good research”

  1. Eamonn Keogh says:

    Thanks for your kind words. As It happens, I will be giving the tutorial in Melbourne sometime in late nov 2009. Hope you can stop by.

    eamonn

  2. william says:

    Excellent! I will keep an eye out for the announcement.

  3. Shane C. says:

    Many of the same ideas (and more) have been advocated in Simon Peyton-Jone's thought provoking research howtos. Ever wonder why people mysteriously started putting the "Related Work" section at the end of research papers instead of the beginning recently? Look no further:

    http://research.microsoft.com/en-us/um/people/simonpj/papers/giving-a-talk/giving-a-talk.htm

  4. william says:

    Great link, thanks. I must admit I have not been a fan of the "related work at end of paper" structure, as it comes across as a slightly dismissive afterthought. The approach I have attempted to emulate was one in which related work is discussed directly as part of the introduction, with the problem statement being framed in terms of previous research. However, I can see Peyton-Jone's point that discussing related work after the introduction places a barrier between the reader and the presentation of your idea. I'll have to ponder this.

  5. Shane C. says:

    I don't like putting the "Related Work" section at the end at all if you have an experimental section. For theory papers, it seems like a reasonable idea. However, if you want to do a respectable experimental evaluation, you must have baselines. How can you make a persuasive argument without explaining what others have done, and showing how the new methods compare to other solutions?

  6. [...] of the side benefits of reading about how to write academic papers is that it improves your skills as a reviewer, too. In fact, the latter improvement is greater than [...]

Leave a Reply