Reviewing Rejection Top Ten

1
Reviewing Rejection Top Ten

• ( The most common reasons I reject papers that I
  am asked to review )
• James Davis
• UC Santa Cruz
• 2005

2
10 Poor figure captions

• Figure captions must be self explanatory.
  Reviewers and other readers often read figures
  first to get an overview of the paper. You simply
  cant assume that the reader knows the details of
  a paper before looking at the figures. This is
  what captions are for to explain the figure.
  The captions can be simplified, since there is
  limited space. However they should leave the
  reader feeling that they understand something,
  and not more confused than before they looked at
  the figure.

3
9 Inscrutable figures

• Spend time to construct figures carefully. It is
  not ok to just take a screenshot. They should
  convey information, or they shouldnt be present.
  Diagrams should have the minimum number of
  components that convey the idea. Graphs should be
  labeled with units and in a legible font size.
  Labels should go next the item they are labeling
  in the figure. .e.g. its not ok to tell me the
  red line means X and the blue line means Y in the
  caption, this forces the reader look away from
  the figure. The default look of your MATLAB graph
  is almost certainly wrong. Fix it.

4
8 No statement of motivation

• The very first paragraph of your paper (or at
  least the introduction) should motivate why you
  are doing this work. This needs to be done at two
  levels. First, why is this important
  commercially, or to society, or in some vague
  high level way? Secondly, what is the particular
  problem you are addressing? What is the
  challenge? Why is the solution not immediately
  obvious? A statement of the goal that the paper
  addresses needs to be made explicitly. The
  introduction section needs to convince the reader
  of two things. That the work is important and
  that there is in fact a problem to be solved. You
  cant just have one or the other. You need both.

5
7 Infatuated with math

• The paper text needs to be readable by people
  who skip equations. You can assume an appropriate
  level of mathematical sophistication, however you
  cant assume people are in the mood to read all
  your Greek letters and tiny subscripts. Its
  necessary to include equations of course, however
  they dont supercede the need for a clear
  discussion of your work. Similarly, it cant be
  assumed that the reader remembers what Xi1
  refers to three pages after it was first
  introduced. Use plain language to describe the
  concepts. Read your paper skipping all equations
  and symbols. If its not understandable, go fix it.

6
6 Bad grammar / English usage

• The paper needs to be easy to read and
  understand. Poor grammar, spelling, and strange
  English usage make it hard for people to read.
  This is nearly always caused by writers for whom
  English is their second or third language.
  Sometimes the writing is strictly correct, but a
  native speaker would just never say it that way.
  Reviewers should endeavor to be as understanding
  as possible. Writers should get a native English
  speaker to read their work and help to fix it. Is
  this fair? Nope, not fair. I apologize. However
  it has to be done. Most computer science
  journals and conferences are conducted in
  English. However, if youre submitting to the
  Tibetan conference on graph theory, get a native
  Tibetan speaker to read and correct your work.

7
5 Incomprehensible writing

• Your writing needs to be clear. Perfect grammar
  in no way insures that the reader can figure out
  what you are talking about. The text should be
  organized to follow some logical flow. The flow
  should be obvious to the reader. They should know
  when they are reading background, when they are
  reading your new idea, when its a demonstrated
  result, and when you are just speculating. The
  goal is to communicate your ideas, there are no
  extra points for confusing the reviewer to show
  how smart you are.

8
4 Unconvincing results

• Somewhere your paper should have results. These
  results should be convincing. I have reviewed
  animation papers with terribly ugly videos. I
  have reviewed papers with a single unlabeled
  MATLAB plot. I have even reviewed papers which
  directly say that their results are inferior to
  prior techniques. Now why would I accept a paper
  if the new method is worse than the old method?
  Include results. Make sure they are convincing.
  Make sure they indicate that your new work is
  superior to old work. If you work in graphics
  make sure they are aesthetically beautiful. If
  the results arent convincing, continue your
  research until they are convincing.

9
3 Not sufficiently novel

• Your work needs to contribute something to the
  scientific community. Its not worth publishing if
  it simply repeats what others have done. (With
  the exception of certain scientific studies that
  need validation from other researchers.) If part
  of your work is new and part is old, emphasize
  the new stuff. Sometimes the part that took 90
  of the time, effort, and heartache isnt even
  mentioned in the paper. Bummer. Write about the
  new stuff.

10
2 No relation to previous work

• Present your work in the context of the
  previously existing work. You have a related work
  section. This is not simply a list of papers that
  are similar. You need to compare and contrast
  this previous work against your new work. If you
  tell the reader that there are these twenty other
  similar papers and fail to make a comparison, the
  only logical conclusion is that your work isnt
  very novel. The related work section is a chance
  to defend the novelty of your work. Anticipate
  the readers belief that someone has done this
  before. Explain why each previous class of
  methods does not sufficiently address the
  challenge that you identified. Politely. Dont
  imply previous authors are stupid, rather attempt
  to praise their work, just be careful to claim
  they were addressing something different than you
  are.

11
1 No statement of contribution

• State your contribution. Explicitly. Repeatedly.
  At the end of the introduction section there
  should be a sentence that says, The contribution
  of this work is The rest of the paper is your
  attempt to back up that single sentence. Dont
  make the reader guess what the contribution is.
  The reviewer is attempting to determine if your
  contribution is important, novel, and achieved.
  If they dont know what the contribution is, this
  makes their task more difficult. Seem simple?
  This is by far the 1reason I recommend rejection
  of papers. State your contribution. Explicitly.

12
Reasons for rejection

• 01 No statement of contribution
• 02 No relation to previous work
• 03 Not sufficiently novel
• 04 Unconvincing results
• 05 Incomprehensible writing
• 06 Bad grammar / English usage
• 07 Infatuated with math
• 08 No statement of motivation
• 09 Inscrutable figures
• 10 Poor figure captions

13
Reasons related to the work itself

• 01 No statement of contribution
• 02 No relation to previous work
• 03 Not sufficiently novel
• 04 Unconvincing results
• 05 Incomprehensible writing
• 06 Bad grammar / English usage
• 07 Infatuated with math
• 08 No statement of motivation
• 09 Inscrutable figures
• 10 Poor figure captions

14
Reasons related to the writing

• 01 No statement of contribution
• 02 No relation to previous work
• 03 Not sufficiently novel
• 04 Unconvincing results
• 05 Incomprehensible writing
• 06 Bad grammar / English usage
• 07 Infatuated with math
• 08 No statement of motivation
• 09 Inscrutable figures
• 10 Poor figure captions

15
Wrong way to outline a paper

• 1 intro
• 2 related work
• 3 my method
• 3.1 module 1 lots of details
• 3.2 module 2 lots of details
• 3.3 module 3 lots of details
• 4 results
• 5 conclusion

16
Right way to outline a paper

• Motivation
• write a sentence explaining the high level
  motivation
• Challenge
• write a sentence explaining the challenge
• Contribution
• write a sentence explaining the contribution
• Related work
• write down all the related work you can think of
  and for each write a sentence about why it does
  not sufficiently address the challenge you
  specified
• Your actual work not important really, its not