How to write a paper Toby Walsh National ICT Australia and UNSW, Sydney http://www. cse.
unsw. edu. au/~tw How to write a good paper Toby Walsh National ICT Australia and UNSW, Sydney http://www. cse. unsw. edu. au/~tw How to write a good enough paper Toby Walsh National ICT Australia and UNSW, Sydney http://www.
cse. unsw. edu. au/~tw Why me? • I’m not sure • Many other researchers who consistently get their papers published Peter van Beek, Pascal van Hentenryck, .
.. Often win best paper awards • • Why me? • My papers get rejected too! • • This should be sent to a journal” “Very good work, but I’m not sure why Alan Bundy hasn’t written this? ” Anonymous review: “Clearly the author fails to understands Walsh’s previous work on this topic” • 1st Lesson • Don’t lose heart • Even if you do everything right, reviewing is imperfect Good papers will be rejected But try to learn from your knock-backs! • • Why you? • Academic career • Publish or perish • Have an impact • • Communicate your results Many have not had the impact they deserve for being bad writers • Writing is fun! Outline • How to get your paper rejected •There are many traps even experienced researchers make Myself very much included • • Hints about how to write a paper • Writing is a craft not a science! How to be rejected • Submit over-length • Blind man: send in 7 pages even though the instructions clearly say 5 Once they see quality of work, they’ll be pleased you sent in more material • How to be rejected Submit over-length • • Diplomatic immunity: put extra 2 pages in appendix Appendices clearly don’t count Similarly, bibliography doesn’t count • • How to be rejected • Submit over-length • LaTeX hacker: • • • change from 11 to 9 point font squeeze inter-line space … No one will ever notice How to be rejected • Submit late • • • Deadlines are meant for everyone else Review schedules have plenty of slack Your paper is worth the wait! How to be rejected • You don’t have room for space wasters like: • Motivation, Background, Related work • Why do review forms always have these on them anyway? How to be rejected • Annoy reader/reviewer • • • • Proof is trivial, when it isn’t Prove the trivial Fail to cite their work Only cite yourself How to be rejected • Annoy reader/reviewer • Don’t bother to spell check • What do computers know about spelling anyway? •Use all the old cliches • “There has been a lot of interest recently in global constraints. ” How to be rejected • Annoy reader/reviewer • Be pompous, boring, .
.. • This is science not literature guys, who said it should be fun? • Be overly formal • Theorems and formulas add weight How to be rejected • Annoy reader/reviewer • Make them really work • After all, these are dif? cult concepts and it took you some time … • Ignore reviews • Just keep sending paper in, eventually it will be accepted How to write a paper • Hints about how to write • • • Preparation Writing itself Ethics Preparation •Read, read, read! • To learn how to write, read a lot • I spend over 20% of my time reading • 1 day/week in library Read, Read, Read • Related literature • • • So you can cite it So you don’t re-invent wheels So you know what others think are important research questions Read, Read, Read • Other conference/journal papers where you intend to publish • • So you learn the “house” style So you can place your work within the bigger picture So you learn how to ask good questions • Read, Read, Read • Any sort of literature • Magazines, novels, biographies, ..
. • Writing is a skill, learn from others •I read approx one novel/week as a way of trying to learn how to write And I have the luxury of writing in my own language! • Review, Review, Review • Review as much as you can • • • So you see good/bad writing So you see the newest results (but see ethics) So you ask yourself good questions • What is the contribution here? What are the weaknesses? …
Write, Write, Write • The best preparation to writing is to write • • Writing gets easier the more you do it Writing is easier if you’ve drafted much of what you already need Writing is the best way to organize your thoughts Writing is a good way to record what you have done • Writing • Work out the timetable • • • Rushed papers frequently rejected Late papers are almost always rejected If you always write to deadlines, writing will seem more painful than it is Writing • Work out the message • You should be able to convey this in one sentence • “We propose a new global constraint, provide a ? ltering algorithm and show it useful on some standard benchmarks” Writing • Work out the message • You should be able to convey this in one sentence • “We identify an important class of symmetry, and show how to break it” • Write to the message! Writing •Distribute the work • • Play to your strengths If you have a native speaker, have them write intro/conclusion … • Write to length • Brutally cut papers are frequently rejected Writing • • Structure paper before you write it I write template for paper with sections and subsection headings ? rst • Intro, Background, Theoretical results, Empirical results, Related work, Conclusions Writing • • Start where you are most happy Often write from the middle outwards • Theoretical results, Experiments, …
, Conclusions, Introduction, Abstract Writing • • Rule of Three Say everything 3 times! • • •Introduce idea (introduction) Develop idea (body of paper) Summarize result (conclusions) • But don’t copy verbatim the same text! Title • Make it meaningful and brief • • Don’t make a joke Remember someone reading reference needs to be able to work out likely contents Good: the TSP phase transition Bad: Easy Problems are sometimes Hard • • Abstract • Executive summary • Try for one sentence or so on: • • • • Motivation Method Key result Conclusions Introduction • • • • What is the problem? Why is it interesting? What are your contributions? What is the outline of what you will show? Introduction Lure the reader in a with a good ? rst sentence • Bad: There has been a lot of work recently on phase transition behaviour… Good: Global constraints are central to the success of constraint programming…
• Background • Often need to set scene • • • De? ne formalism Get reader up to speed Identify research problem Body of Paper • • • • Derive theoretical results Propose new algorithm Describe system engineered … Results • Bullet proof paper • Theoretical results • Experiments only provide a limited view • Experimental results • Theory doesn’t show if results are useful in practiceRelated work • Has many purposes • • • You give proper credit to prior work You are not re-inventing wheel You can compare what you do with what has been done before Conclusions • • Remind reader of what you have done Place work in wider context • “What general lessons might be learnt from this study? ” • Flag all the exciting open research directions Acknowledgements • Thank all who have helped you • Provided code, data sets, .
.. • Thank ? nancial sponsors Writing • Keep it simple! • • • • Active, not passive Present, not past or future Long words Short sentences Writing •Avoid temptation to include every result you have • • • Paper needs to be coherent Paper needs to be understandable Many papers are rejected for having too many results! Ethics of Writing • • • • Authorship Citation Submission Publication Authorship • Who should be an author? • Anyone who has made a signi? cant contribution May not have written any text! Always err on the side of caution • • Authorship • Who should be an author? • • • Ask! You’ll be surprised how often people refuse You can be sure they’ll not work with you again if they feel they should be Authorship • Should my advisor be an author? • • •In ? rst few years of thesis, probably yes .. Once you graduate, you should (be able to) write papers on your own Again, ask! • Citation • Cite all relevant work • • • Reviews always ask about Related Work You’ll want them to cite you It’s central to the scienti? c method • We stand on the shoulders of others Citation • Do I cite myself for a blind review? • • • Yes! You must credit all previous work Either cite [Author, 2004] Or write “As Walsh has shown previously [Walsh 2004] ..
. ” Submission • Can I submit to multiple conferences? • What’s the deal with the disclaimer (”This paper is not under review .. )? • Can I submit to a journal immediately? • • No hard and fast rules My rule, once reviews are back and paper is effectively in press … Publication • Can I publish my conference paper as it is in a journal? • Probably not, even though conference is not archival Most journals ask you to extend conference paper substantially • • Proofs, more experiments, .
.. Final words • It takes time to learn how to write • Don’t be put off if at ? rst your have papers rejected All of us have papers rejected • • Spend time learning how to write • It will be worth the investment