COMP70100: The Paper Writing Process

1
COMP70100 The Paper Writing Process

• Norman Paton

2
Outline

• Places in which to publish.
• Paper writing.
• Paper reviewing.

3
Papers and Theses

• Both thesis and paper writing involve
• Describing your research results.
• Making an argument for their novelty.
• Making an argument for their quality.
• Convincing external experts.
• Writing papers
• Usually comes first.
• Should come first.
• Is easier principally because of scale.

4
Publishing Research Results

• Why publish?
• To obtain feedback.
• To obtain validation.
• To have an excuse to go to conference!
• To publicise results.
• To obtain ownership.

5
Selecting Places for Publication

• Criteria include
• Reputation of publication.
• Nature of publication.
• Permissible paper length.
• Speed of publication.
• Size of audience.
• Nature of audience.
• Accept/reject ratio.

6
Know Your Area

• In your area identify
• Top journals.
• Middling journals.
• Less good journals.
• Top conferences.
• Middling conferences or workshops.
• Less good conferences or workshops.

• Then offline
• In each case, mark those from which you have read
  papers.
• In each case, mark those in which your supervisor
  has published.
• Discuss your list with others in your group.

7
Rating Publication Places

• Conferences
• Reputation.
• Accept/reject rato.
• Programme committee.
• Publisher.

• Journals
• Impact factor in ISI (www.mimas.ac.uk).
• Publisher/series.
• Editorial board.

Check out www.researchindex.com for (informal)
list of most cited sources in Computer Science.
Check out scholar.google.com to find out how
specific people and papers are cited.
8
Personal Profile

• Most published-in journals
• Information Software Technology (9).
• Data Knowledge Engineering (6).
• Nature Biotech (5).
• Bioinformatics (4).
• Information Systems, JVLC, Concurrency Practice
  Experience, SIGMOD Record (3).

• Most published-in conferences
• British National Conference on Databases (9).
• Interfaces to Database Systems Workshop, Rules in
  Database Systems Workshop (5).
• Data Management in Grids Workshop, Visual
  Database Systems Symposium, Scientific and
  Statistical Databases Conference(3).

9
Comments on Personal Profile

• Balance is towards less good places.
• Too many once only places.
• Next to no unpublished workshops.

10
Citations

• If no-one ever cites your papers, then you are
  doing something wrong!
• Examining citations of your own work or that of
  your supervisor or that of your competitors can
  be illuminating.

11
Active Rules Experience

• Timely first paper good for profile.
• Rapid follow-on capitalised on initial prototype
  and associated experience.
• But then
• Second phase of work much too slow to be done.
• Second phase pitched at journals, and appears
  after huge gap.
• Area well and truly dead by the time later papers
  are published.

12
DOOD Experience

• Initial difficulties getting somewhat radical
  story across.
• Principal publications in good places in
  mainstream database literature.
• However
• Never really reached certain parts of the DOOD
  community.
• Results quite late in lifecycle of area, and thus
  too late to influence/rescue area as a whole.

13
TAMBIS Experience

• Project became quite well known by word-of-mouth.
• Initial paper not very good, but widely read.
• Ontology paper peripheral to overall idea has the
  most influence.
• However
• Too radical for bioinformatics community, too
  pragmatic for computing community.
• Thus not easy to publish at top level in either.

14
Immediate Lessons

• There is no one best time or approach to
  publication.
• You need to be careful, as a result can only be
  published a few times.
• A mistaken publishing strategy can cause results
  to vanish without trace.

15
Repeated Publishing

• How many times can you publish the same result?
• You can never submit/publish the same paper to
  two different places.
• You may be able to report on a single result in
  different ways to different communities.
• You can publish an extended version of a
  conference paper in a journal.
• It is legitimate to re-publish material in edited
  works, as these do not require novelty.
• In general do not push your luck.

16
Summary

• You should publish.
• To do so successfully you need to understand the
  literature in your area
• What is your chosen community?
• How do different places rate?

17
Outline

• Places in which to publish.
• Paper writing.
• Paper reviewing.

18
Publishing As a PhD Student

• You should if you cant your work is probably
  not good enough for a PhD.
• Good practice for how your work will be examined
  peer review of written account.
• Writing-up your work is a good way of getting
  your story straight.
• Students who publish well pass.

19
Papers and PhD Students

• How many papers should a PhD student publish?
• 0 too few. No external evidence that your work
  is any good. No impact.
• 1 perhaps adequate. But are there no
  publishable sub-components to thesis?
• 2 perhaps reasonable. But probably not coverage
  for many of your main chapters.
• 3 quite good going. Coverage for quite a lot of
  thesis material.

20
A Classical Thesis Structure

• Introduction.
• Related Work.
• Technical Context.
• Your basic idea.
• Enhancement.
• Evaluation.
• Conclusions.

• Not a paper.
• Possible if good.
• Not a paper.
• Should be possible.
• Should be possible.
• May be possible.
• Not a paper.

21
Writing Papers

• Plan the publication process. Writing and
  presenting papers is time consuming.
• Pitch carefully at your selected community you
  may have a decision to make here.
• The first papers on a piece of work tend to be in
  workshops or conferences.
• Check how long journals are taking both to accept
  and to print submissions.
• Be realistic, but not over cautious.

22
Authoring

• Not all the authors of a paper need to have
  written text authors should contribute words or
  work.
• The first author of the paper need not be the
  person who wrote it.
• Establish understandings within the group on how
  authorship is handled.
• Lots of actual authors slows the process down.
  The coordinating author is the boss!

23
Telling a Clear Story

• A paper should have a clear aim that
• Makes it clear what the target is, and why.
• Makes it clear what the target is not, and why.
• It should be possible to
• Tell a lay person what the point of the paper is
  in a minute.
• Explain what makes your work different from other
  work in one minute.

24
Example Situating a Result

• Several proposals have been made that
  support join reordering during query evaluation.
  For example, describe how an optimizer can be
  invoked at query runtime to identify a new plan
  in the light of information on progress to date.
  However, these approaches are coarse grained in
  the sense that they either reuse complete join
  results or restart join evaluation from scratch,
  discarding results produced using an earlier
  plan. As a result, they are not well suited for
  use in pipelined plans in which many operators
  are likely to be being evaluated simultaneously.
  Where pipelined query processing has been
  addressed, proposals have either required
  post-processing in the form of stitch-up plans
  or have been limited to a single join
  algorithm . As an alternative to plan
  reordering, Eddies change the order in which
  tuples are routed through joins continuously
  during query evaluation. As such, they can be
  extremely fine grained, but obtain this
  flexibility by materializing (e.g. ) rather
  more intermediate data than may be required by
  other strategies. This paper adopts a
  middle-ground, by supporting adaptation during
  operator evaluation, but in the context of
  established join algorithms. The most closely
  related work supports the adaptive reordering of
  indexed nested loop joins the results
  presented in this paper are more general in that
  they support multiple join algorithms, used
  together or separately, and allow changes to the
  algorithms used and not only to join order.

Problem Statement
Weakness of Related Work - 1
Weakness of Related Work - 2
Weakness of Related Work - 3
Claim
Weakness of Related Work - 4
25
Example Stating Contribution

• The contributions are
• an algorithm that computes the work remaining to
  be done by a pipelined plan containing select,
  project and join operators
• a description of how the algorithm can be
  integrated into a query processor and the
  approach used to monitor progress at runtime and
  
• an evaluation of the approach demonstrating a
  range of circumstances in which significant
  performance improvements have been obtained.

26
Paper Planning

• Have a thesis-long publication strategy.
• How should the pie be sliced?
• Select size and shape of pieces.
• Dont just publish the first thing you can.
• Decide where you want to publish.
• Select a target community.
• Choose a target location.
• Know your target 6 weeks out.
• Read the submission instructions (e.g. length,
  format).

27
Paper Organisation

• Decide on the title.
• Plan the contents.
• Write this out to sub-section level.
• State what will be in each section.
• State the length of each section.
• Agree the contents with co-authors.
• Allocate authors to sections.
• Write your bit in order if at all possible.

28
Paper Assembly

• Agree delivery dates, such as
• Individual authors contributions 1st March.
• 1st complete draft 7th March.
• Comments on 1st draft 14th March.
• 2nd draft 21st March.
• Comments on 2nd draft 27th March.
• Submit 30th March.

29
Write the Paper

• Agree on formatting and diagram tools.
• If at all possible choose LaTeX!
• but consider publishers requirements.
• Deviate from your planned structure only with
  reluctance.
• This should have been carefully constructed in
  advance.

30
Make It As Good As Possible!

• Steal good paper structures and styles.
• If it worked for them it may for you.
• Think about what made the other paper work.
• Consult a book on research writing for the
  details.
• J. Zobel, Writing for Computer Science, Springer,
  1997.
• Dont just describe your work, make a case. Your
  supervisor is not the target audience!

31
Politics

• Dont ignore or slag-off members of the programme
  committee/editorial board.
• Select places that like your type of paper
• Are you citing earlier work from the same place?
• Has this place published similar work to yours?
• Is your paper relevant to the call for papers?
• Write the paper in a way that takes into account
  the background of the reviewers.

32
Outline

• Places in which to publish.
• Paper writing.
• Paper reviewing.

33
Peer Review

• Is the most common way of assessing research
• Research degree assessment.
• Paper publication.
• Rating departments.
• Assessing grant application.
• Assessing grant outcomes.

34
Objectivity

• Peer reviewers
• Are not necessarily objective.
• May misinterpret your work.
• May not understand your work.
• May miss the main point of your work.
• May find your work boring.
• but are often right, and are here to stay.

35
The Reviewer

• Was busy before your paper
• arrived.
• Received eight other papers as well.
• Is not especially interested in your topic.
• Doesnt know exactly your area.
• Doesnt share your perspective on the area.

36
The Reviewers Dislikes

• Papers writing in rong english.
• Papers that are difficult to understand.
• Papers out of scope of the conference.
• Undefined terms.
• The wrong amount of formality.
• Papers without a clear purpose.
• Papers that ignore the related work.
• Papers that ignore the reviewers work!

37
Handling Feedback

• It is OK to have (some) papers rejected.
• On the whole, referees are not idiots.
• Papers generally improve in the light of referees
  comments.
• Negative comments do not always correctly
  identify the real problem.

38
Trying It Out

• Ask yourself how useful your written comments
  would have been to the author.
• Ask yourself if your comments demonstrate that
  you read the paper carefully.
• Ask yourself if you would have respected a
  reviewer who wrote a review like yours.
• If you are asked to review a paper, take the task
  seriously peer review is crucial to science.

39
Summary

• Peer review will happen to you.
• Sometimes this will be unpleasant.
• It will only occasionally be unfair.
• It will sometimes be sloppy.
• You can learn a lot from it.
• Treating feedback seriously
• Can teach you about your work.
• Can help you convey your work better.
• Can help in community identification.

40
Further Reading

• Michael J Katz, From Research to Manuscript A
  Guide to Scientific Writing, Springer, 2006.