HOW TO WRITE A GOOD PAPER

1
HOW TO WRITEA GOOD PAPER

• Jehan-François Pârisjfparis_at_uh.edu

2
Introduction

• You have interesting results
• You have learned to write correct English
• Now
• Your adviser tells you it is time to write a
  paper
• You have to wrap up your term paper
• What should you do?

3
What we will cover today

• How to present your results
• Cover the basic parts of a scientific paper
• Focus on what to say in each of them
• Review a few basic rules to follow

4
How to present your results

• Your results should be interesting
• Improve upon the state of the art
• In a significant manner
• It is your job to explain that to the reader

5
The twenty percent rule

• To be significant, your contribution should be at
  least 20 percent better than state of the art
  solutions
• Picking the right performance index is important
• Increasing a cache it ration from 96 to 98
  percent is less impressive than cutting in half
  its miss ratio!

6
Your future paper (I)

• Five classical parts
• Abstract
• Introduction
• Review of previous work
• Your own contribution
• Conclusion
• Followed by a list of references

7
Your future paper (II)

• An alternative organization
• Abstract
• Introduction
• Your own contribution
• Review of previous work
• Conclusion
• Followed by a list of references

8
Which one should you pick?

• Pick the traditional organization if your paper
  builds upon previous work
• If your own work addresses some issues that were
  not addressed by previous authors
• Pick the alternative organization otherwise
• Lets the reader go straight from the introduction
  to the section describing your own work

9
The abstract (I)

• Not an introduction
• A very brief summary of the results of the paper
• Describe the problem
• Sketch the approach
• Present you best results
• The idea is to entice the reader to read further

10
The abstract (II)

• No need to explain why the results are important
• Introduction is there for that!
• Must be self-contained
• No unexplained acronyms
• No references!
• Use very sparingly mathematical symbols
• Write it last!

11
The introduction (I)

• Should present the topic and the scope of your
  paper
• Key objectives are
• explaining what you will do (and not do)
• making it sound interesting
• captatio benevolentiae
• getting the goodwill of the reader

12
The introduction (II)

• Main difficulty
• Must describe the problem and your results in
  very general terms
• Must simplify things without being inaccurate
• Some authors end the introduction by a roadmap of
  the paper
• Section II of this paper reviews

13
The review of previous work

• Mention previous work that is relevant to your
  paper
• The key word is mention
• You should not describe in any detail any
  previous work unless it is necessary for the
  understanding of your own work
• Do not sweat it!

14
Your own contribution (I)

• Explain in a few lines its scope
• You will investigate an idea that you find
  interesting
• You have a better idea that solves one of the
  problems other proposals did not solve
• Then develop your ideas

15
Your own contribution (II)

• Unless you present the first known solution to a
  problem, you should compare the performance of
  your solution to that of previous solutions
• May require running experiments or simulations
  that evaluate these solutions
• Not always possible

16
Checklist (I)

• Do not forget to mention the limitations of your
  approach
• Markov models of disk arrays must assume that
  disk repair times follow an exponential
  distribution
• Not true!
• Necessary for having a tractable model

17
Checklist (II)

• Be sure you present your results the best you can
• Tables
• Graphs
• Do not forget to label the axes and the curves
  and to indicate the units

18
Checklist (III)

• Use an equation editor for your equations
• Avoid
• ax2 bx c
• All variables used in the text must be italicized
• The number n of disks in the array

19
The conclusion

• Should summarize the results of the paper
• Overlaps with abstract
• Often written in perfect tense
• Research papers often include directions for
  future work
• Many people read the conclusion first before
  deciding if the whole paper is worth their time
• Should summarize your main results

20
The list of references (I)

• Normally appears as an unnumbered section
• Very important part of paper
• Indication of good scholarship
• Format obeys to strict conventions
• Another indication of good scholarship
• Take it very seriously

21
The list of references (II)

• Always cite papers describing pioneering work
• First paper on broadcasting for video-on-demand
  was by Viswanathan and Imielinski
• Must be cited even though much better techniques
  have been devised
• In case of doubt, cite the papers that other
  people cite

22
We almost forgot the title

• Should be descriptive
• Might be affected by the target audience

23
And the list of authors
24
In reality

• Very poor taste not to mention as co-authors the
  names of the people supervising your work
• Mentioning people without their permission is
  almost unforgivable
• Jim Grays rule
•  "He who types the paper is first author." 

25
Theses and dissertations

• Typically follow the classical organization
• Chapter I Introduction
• Chapter II Previous Work
• Chapter III to ?? Your own contributions
• Chapter ??1 Conclusion
• Summary
• Directions for Future Works
• References

26
Writing the paper (I)

• Avoid the passive voice
• BADThree different block sizes were selected
  for our experiments
• BETTERWe selected three different block sizes
  for our experiments

27
Writing the paper (II)

• Use short sentences
• Uses figures whenever it helps
• You can borrow figures as long as you acknowledge
  your source
• Figure 9. A page table (from 9)
• where 9 is a reference to a paper you cite

28
Tables and figures (I)

• Tables have titles that are above the table

29
Tables and figures (II)

• Figures have captions below them
• Make your table titles and figure captions as
  self-explanatory as possible
• Many readers will look at the tables and figures
  before reading the text

30
Other rules to follow (I)

• Spell numbers at the beginning of a sentence
• BAD52 of the observations
• BETTERFifty-two percent of the observations

31
Other rules to follow (II)

• AWFUL3 different block sizes were selected
• BADThree different block sizes were selected
• BETTERWe selected three different block sizes

32
Other rules to follow (III)

• Spell out percent everywhere in your text
• BADWe observed a 20 increase
• BETTERWe observed a 20 percent increase

33
Other rules to follow (IV)

• Its is not a possessive
• BADIts main disadvantage
• BETTERIts main disadvantage
• Its is a contracted form of it is
• Most good writers avoid these forms in scientific
  papers and reports
• youre, cant

34
A last word

• Good writing is a craft
• You must exercise it to become better at it
• Good writing takes time and effort
• Will become more enjoyable as you progress