Transcript Document

Pitfalls of your first paper

Shu Cai Institute of Computing Technology, Chinese Academy of Sciences [email protected]

Why this report  My first paper is rejected from ACL-IJCNLP 2009.

 Unlucky thing, but a good opportunity to learn from this experience   I believe that successes could have many, many different reasons… but failures tend to have more clues I write this report to summarize the lessons I learned, and hope other people could avoid such potential errors

Outline  Idea  Experiments  Writing  Asking for comments

Idea  There are various ways to get an idea…  What I will emphasis here is an idea of your own  You did NOT simply receive it from an advisor or a colleague…  You propose your own idea when you read papers, think about your research, or you just “blurt out” the idea during discussion

Possible flaws of the idea  Too idealistic (sounds reasonable but hard, even impossible to implement)  Low significance (Similar problem has already been tackled in better ways, or it will be of interest to few people  Wrong understanding (You may misunderstand the problem you want to solve)

And how to avoid these traps …  Analysis: see if this idea is feasible (Complexity analysis on time and space; Resource checking)   Comparison: read related work and compare this idea to others, try to find not only advantages but also disadvantages Discussion: ask for experienced people’s advice, discuss your idea with them

Polish your idea  Since it is your first paper, the idea might be coarse  Polish it to shine!

 Early optimization could save a lot of time (see software engineering books for proofs)  Think about details of your idea  Try to find extensions of your idea  But do NOT expect it to be perfect

Be serious about your idea  Don’t be afraid of rejection  Rejection is better early than late!

 The idea might be applicable in other areas  Think about your readers: what do you expect them to get from your idea?

Experiments  After you make sure that your idea is valuable (and feasible), you need to do experiments to prove it.

 Important but intricate process

Experiments: possible pitfalls  “Stuck” in the middle  Results are not as expected  Time is not enough

Possible solutions  Divide your implementation into phases  Check at the end of each phase whether you have implemented as you thought  Explore why you did not obtain the desirable results at each phase  Allocate time for thinking, redesigning, and handling accidental interrupts.

A summary of Experiments  A way to test our ideas  Carefully design your experiments  Experiments should support your ideas  Experiments could be a time-consuming process, so allocate enough time!

 Manage your time and energy!

Writing  Deadline is approaching….

 You should begin writing your first paper!

 Exciting? But Challenging also!

Pitfalls of writing your first paper  Write it like a technical report  Describe every detail of your implementation  Not enough time

Take precautions  Follow a clear skeleton  Delete unnecessary details  Allocate enough time

Revising your paper  Step by step, Section by Section  Ask for comments  Remember, your ultimate aim is present the readers with your work. Often, you need to persuade them.

Asking for comments  DO NOT only ask your advisor for comments   Ask someone in your research area, but do not know about your research to give you advice They are “pseudo-reviewers”  Ask for experience people which conference/journal you should submit to

Pitfalls after submission  Transfer to another project/ research quickly

What should be done after submission  Arrange your files (experiments, papers), make them easy to be found  You will soon go back to these files after you get the result (whether acceptance or rejection)

Thank you!

Any comments are welcome!

Please send comments to: [email protected]