1

Writing a Scientific Article

Bogdan Dumitrescu

“Politehnica” University of Bucharest, Romania, and

Tampere University of Technology, Finland

2

General contents

� What is a scientific article ?

� When do you start writing ?

� General structure of an article

� Style issues

� Review process

� Revising a paper

� ...plus some English and a big homework

3

1. What is a scientific article ?

� A scientific article

� is a written communication presenting results of

scientific research

� may contain theoretical results and their proofs

� often presents experimental data that support the

theory

� is addressed mostly to specialists

� is published in a journal, typically after a peer-review

process

4

Types of articles

� There are two main types of scientific articles

� Research articles: dedicated to communication of original

research results. Depending on the length:

� Regular (full) papers: length e.g. 8-10 pages double column or 20-30 pages single column (draft format)

� Letters (technical notes, etc.): shorter, e.g. 4-5 pages

� Review (survey, overview) articles: synthesis of recent

results in a field, a topic, a problem. No original

contribution, but typically the authors have significantly

worked in the area and are recognized specialists

5

Full paper or letter ?

� You need original contributions for both !

� If theoretical contributions are minimal, probably a letter is better

� Letter also better if you improve on other results, without coming with an original approach

� If in doubt, write the paper. You’ll decide when the paper is almost ready

� Warning: some journals don’t accept both letters and full papers

6

Comments and replies

� A less significant type of article: the “comments”

� Comments are very short

� Referring to a previously published article, they

� point out a significant error and maybe give a cure

� affirm that the original contribution was actually

published elsewhere

� give a shorter or more elementary proof

� A “reply” is an answer of the authors of the initial

article

7

Terminology examples (1)

� Automatica

� Survey papers - Extensive reviews of established or emerging research topics or application areas

� Papers - Detailed discussion involving new research, applications or developments. [10 printed pages, i.e. 10 000 words.]

� Brief papers - Brief presentations of new technical concepts and developments. [6 printed pages, i.e. 6000 words.]

� Technical communiqués - New useful ideas and brief pertinent comments of a technical nature. [4 printed pages, i.e. 4000 words.]

� Correspondence Items

8

Terminology examples (2)

� IEEE Trans. on Signal Processing

� Regular paper (max 30 double-spaced pages, 11pt

font)

� Correspondence items (max 12)

� IEEE Trans. on Automatic Control

� Full paper (max 32 double-spaced pages, 12pt)

� Technical notes and correspondence (max 12-15

pages)

9

Where do you publish ?

� Target: best journal that might accept the paper

� Why ?

� good audience—many potential readers

� more likely that your article will be cited

� as a researcher (even only for PhD) or professor, your

publication list is primarily evaluated based on the

journals

� (a refined evaluation is done based on the articles

themselves, but you must survive the first evaluation)

10

Narrow the target from the beginning

� Before starting to write your paper, choose at

most 2-3 journals, one of which will be the final

destination of the paper

� Check the requirements of these journals: format

of the submission, length, other details

� Download Latex or Word templates

� Although they affect only marginally what you’ll

write, these details provide a useful framework

and free your mind for the main job—writing

11

How do you tell a good journal ?

� Tradition and reputation:

� you have read many good articles from it

� famous researchers have published in it

� your professors used it for teaching or research

� etc.

� Scientometrics information:

� impact factor

� other quality measures

12

Tradition vs. noname

� Journals edited by societies with tradition are

usually good (or at least not bad)

� IEEE, IFAC, SIAM, IET—good labels, generally

you can rely on the title of the journal to know its

contents

� Relatively new journals: there is a risk, try to get

as much information as you can

� Bad labels: WSEAS is a good example of low

quality (but certainly not the worst)

Romanian journals

� In the latest few years, many Romanian journals

managed to get indexed in major databases

� Before submitting, read at least the contents of a

few issues

� Even if the quality is not the best, it is important

that the contents is focused

� Counterexample: Metalurgia International

publishes papers on materials science,

management, environment, social sciences

Electronic journals

� Some journals are published only electronically

� They can be good or bad, as the others

� These journals are not necessarily free, the readers have to pay (in fact, only few are free, and not the best)

� At some journals, the authors are offered the “open acces” option: free access to all readers

� The authors have to pay a fee going from 400 to 2000 euro (?)

Databases

� Good journals are indexed in databases

� Reciprocal is not true: databases contain also lower rank journals, conference papers

� Main databases:� ISI web of science (maintained by Thomson Scientific's

Institute for Scientific Information)—the most used

� Scopus (Elsevier)—emerging and quite good, but very accurate only for data after 1995

� Google scholar—free, very extensive, but many “gray”area papers (i.e. “garbage”)

16

Do you have to pay ?

� Publishing is free in good journals !

� However, some journals impose a maximum

page number (usually big enough)

� You’ll have to pay for the extra pages, if your

article is very long (>10 pages at IEEE TSP, >12

pages at IEEE TAC, in the publishing format)

� If money is a precondition for publication, go to

other journal

Copyright issues

� At most good journals, the authors transfer all

rights to the journal

� So, the article becomes property of the journal

� If you’ll want to reuse pieces of text or figures

(e.g. in a book), you have to ask permission to

the journal

� If you want to protect the methods or the devices

described in the article, you must apply for a

patent before publishing

17

18

Conference upgraded to journal ?

� Some conferences promise to publish your article

twice

� in the proceedings

� in a journal (sometimes only selected papers)

� This is not exactly good practice…

� An article can be published only once !

� A few decent journals publish special issues with

conference papers; however, this is clearly stated

19

Impact factor

� Impact factor in 2009 is

IF = N_cites / N_papers

� N_papers: number of papers published by the journal in 2007 and 2008

� N_cites: number of citations to these papers, in articles appeared in 2009, in all indexed journals

� For engineering journals� IF>1 is good

� Max values are typically 3-4

20

Impact factors 2008 (ISI)

� IEEE Trans. on Automatic Control 3.293

� Automatica 3.178

� International Journal of Control 1.130

� IET Control Th & Appl 1.070

� IEEE Trans. on Signal Processing 2.335

� IEEE Signal Proc. Letters 1.203

� Signal Processing (EURASIP) 1.256

21

Other measures

� Impact factor on 5 years

� Immediacy index: N_cites/N_papers from the

same year—not relevant in engineering

� Cited half life: median age of articles from a

journal, cited in the current year

� Eigenfactor

� Warning: different databases give different values

of the performance indices

Hirsch index (h-index)

� Appropriate for researcher evaluation

� Basic idea: it’s important that articles are cited, not only published

� A researcher has Hirsch index h if h of his papers are cited at least h times (and the other papers are cited less than h times)

� Good especially for researchers with some experience

� Advantage (and drawback): it’s a single number

h-index illustration

� Order the papers on

decreasing number of

citations

� Plot citation numbers

� Draw the bisector

� Count points above

bisector

Graph source: wikipedia

Timeliness

� Sometimes you are interested in a (relatively)

quick publication

� It’s difficult to find proper statistics on the time

taken by the publication process

� Browse the journal and see for a few articles the

relevant dates: “received March 3, 2008; revised

January 12, 2009”

� Compare with publication time and you’ll estimate

the duration of the publication process

Examples

� Journals dedicated to letters may offer a

publication time of about 6 months, e.g. IEEE

Signal Processing Letters

� Some journals are slower, but post on their site a

first electronic version of the article right after

acceptance

� A few reputed journals are very slow, e.g. IEEE

Transactions on Information Theory, publication

time 2 years

26

2. When do you start writing ?

� Different schools of thought: you should start

writing when you

� had a presumably good idea

� gathered evidence seeming to show the idea is good

� completed all proofs, experiments, etc., that will be

included in the article

27

Early start

� Write as soon as you start an investigation

� Pros:

� writing notes or even whole sections helps to clear

your mind, set up a single system of notations

� it will be much easier to write the final paper

� Cons:

� writing too many details may get you confused

� you’ll throw away most of the texts

� Good if you are able to organize your notes

28

Middle start

� Write when you know the general contents of the

paper

� This will make writing easier and will lead to

fewer versions and corrections

� In the process of writing it will become clear if

there are some gaps in the paper

� It is also possible to discover that in fact you have

all necessary material

29

Late start

� Write when you have no other choice

� Pros:

� you have all the material

� you can dedicate full time to writing

� you should expect no surprises

� Cons:

� it may be difficult to structure the information

� you may be under the pressure of a deadline

30

So, when to start ?

� Start as early as you can, especially if you have

coauthors

� The time lost by throwing away some of the old

versions is compensated by the quality of the

final paper

� Sometimes, you actually gain time, by gaining

more insight to the problem and hence finding

easier good results

31

How to start ?

� Most people have troubles when starting a paper

� Remember that anything you write releases some pressure—you have less to write

� No matter when you start, it’s better to start with the most familiar part

� Easy starts� statement of the problem

� proof of some technical results

� figures and tables

32

What first, what last ?

� Other starting points

� notation section

� a general bibliography

� tentative paper and section titles

� Where not to start

� introduction (maybe few notes are good)

� abstract

� conclusions

33

Editing tools: Latex or Word ?

� Word: articles with text, tables and figures

� Latex: (much) better for formulas

� Personal preference: Latex, by far; the papers simply look better !

� Articles are submitted typically in pdf form

� However, when the article is accepted, you’ll have to give the sources

� Most journals accept both Latex and Word, but it’s better to check from the beginning

34

3. General structure of an article

� Title, authors, affiliation

� Abstract

� Introduction

� The problem

� Solution

� Experimental evidence

� Conclusions

� Bibliography

Body of the paper}

35

Good titles

� The title is first read in an article

� It must be informative and, if possible, attractive

� A good title is a very short abstract of the paper

� It contains the main keywords that describe� the problem

� your original contribution or at least your approach

� Basic title: “Method X for Problem Y”

� One or two eye-catching words or a good acronym help

36

Good titles: max information

� The following titles tell everything about the paper

� Root Locations of an Entire Polytope of Polynomials: it

Suffices to Check the Edges

� Edge Theorem for MIMO Systems

� Protein is Compressible

� A Plurality of Sparse Representations Is Better Than

the Sparsest One Alone

37

Good titles: catchy

� Some catchy titles

� Greed is Good: Algorithmic Results for Sparse

Approximation

� A WISE method for designing IIR filters

� The period three means chaos

� 19 dubious ways to compute the exponential of a

matrix

� However, a bad paper with a catchy title is easier

to reject—unsupported arrogance is punished

38

Bad titles: too general

� These titles just give the general problem, but

don’t say anything about the solution

� “On Factorization of Trigonometric Polynomials”

� “On Distributed Averaging Algorithms and Quantization

Effects”

� May be good only if it’s the first article on that

topic, but even then they can be improved

39

Words to avoid in the title

� Avoid the words “new”, “novel”, “improved”

� “New Results on Stability of Discrete-Time Systems

With Time-Varying State Delay”

� “A Novel Method for Designing…”

� You have an original contribution, so of course

the results are new and the method is novel

� When you’ll improve on the “novel method”, how

will you call it: “an improved new method” ?

40

What is wrong in these titles ?

� Titles taken from IEEE Trans. Auto. Control, 2009

� New Expressions of 2x2 Block Matrix Inversion and

Their Application

� On the Value Functions of the Discrete-Time Switched

LQR Problem

� Efficient Routing Algorithms for Multiple Vehicles With

no Explicit Communications

� Some Properties of Conservative Port Contact

Systems

41

What is wrong in these titles ?

� From Continuous-Time Design to Sampled-Data

Design of Observers

� Data Transmission Over Networks for Estimation and

Control

� Modified Anti-Windup Compensators for Stable Plants

� New Results on Modal Participation Factors:

Revealing a Previously Unknown Dichotomy

� Further Results on Incremental Input-to-State Stability

� Some graph-theoretic approaches to certain facilities

layout models

42

Authors—names

� Consider adding a middle initial, like in “Bogdan

A. Dumitrescu”

� It’s helpful in differentiating authors in databases,

especially if you have a common last name

� Use the initial of your second forename, of your

father’s name

� Women: consider continuing using you maiden

name for publishing after marriage

43

Authors—order

� If there are several authors, what’s the order ?

� Normal procedure: authors are listed in

decreasing order of contribution

� Alphabetical order is used in mathematics

� Team leader or supervisor is often last

� First position in authors list is important: don’t

give it away if you made most of the work

44

Corresponding author

� The corresponding author submits the paper and

is the liaison with the journal in all matters

regarding the article

� In some journals, the corresponding author is

indicated

� Typically, the corresponding author

� has the most significant contribution to the article

� or/and is the team leader

45

Affiliation

� Give all details of your professional address

� Avoid giving home address instead

� Avoid yahoo or gmail email address

� Remember: a bit of your status is given by the

institution for which you work

� However, a good paper gets to be published, no

matter the authors and their affiliation

46

Abstract

� The abstract has usually 100-200 words

� It must contain only essential information� the problem (1 sentence)

� the nature of your contribution (1-3 sentences)

� the benefits of your contribution (1-2 sentences)

� Aim to short, precise sentences

� Many people decide reading the article based on the abstract: state clearly your contribution

� Write the abstract when the paper is almost ready

47

Abstract—example 1

� An example (Mahmoud 2000), minimalist

� The problem:In this paper, we address the problems of robust H

∞

performance analysis and control synthesis for a class of discrete-time systems with norm-bounded parameter uncertainty and unknown constant state delay.

� The contribution:

Through finite-dimensional algebraic Riccati equations,

we provide a necessary and sufficient condition for

designing a memoryless state-feedback controller which

stabilizes the discrete time-delay system under

consideration and guarantees an H∞-norm bound

constraint on the disturbance attenuation for all

admissible uncertainties and unknown delays. An

example is worked out to illustrate the developed theory.

� No benefits !

49

Abstract—example 2

� A bit too long and too emphatic (Stoica et al 2000), but goodThe parameter estimation of moving-average (MA) signals from second-order statistics was deemed for a long time to be a difficult nonlinear problem for which no computationally convenient and reliable solution was possible. In this paper, we show how the problem of MA parameter estimation from sample covariances can be formulated as a semidefinite program that can be solved in a time that is a polynomial function of the MA order. Two methods are proposed that rely on two specific (over)parametrizations of the MA covariance sequence, whose use makes the minimization of a covariance fitting criterion a convex problem.

The MA estimation algorithms proposed here are computationally fast, statistically accurate, and reliable. None of the previously available algorithms for MA estimation (methods based on higher-order statistics included) shares all these desirable properties. Our methods can also be used to obtain the optimal least squares approximant of an invalid (estimated) MA spectrum (that takes on negative values at some frequencies), which was another long-standing problem in the signal processing literature awaiting a satisfactory solution.

51

Keywords

� Some journals require a few keywords after the

abstract

� Start with keywords defining the topic, then

narrow the scope to keywords related to the

problem and your specific contribution

� Example:

� Discrete systems, robust control, uncertain

parameters, delay factors, H∞

performance

52

Introduction

� First section of the paper should contain

� historical perspective

� problem statement

� previous work

� original contribution

� paper outline, notations

� Each of these may have 1-3 paragraphs

� The order may be different, some of the points

can be merged

53

Introduction—historical perspective

� Show how your problem has appeared

� Stress its importance for the potential readers

� Cite a few landmark papers in which the problem

was defined and shaped

� You may start with a general sentence on the

field, then narrow the description to your problem

54

Introduction—problem statement

� Purpose: clear, but short description of your

specific problem

� Very few formulas, only those strictly necessary

to describe the problem

� Don’t describe in detail the problem if it takes too

much text

� Illustrate with a scheme or figure, if possible: this

helps the reader to understand quickly

55

Introduction—previous work

� List the main existing solutions to your problem

� Try to organize the previous articles into classes

� Cite relevant articles for each class

� Hint at possible deficiencies of the cited methods, especially if you improve them

� However, praise previous work: being generous is often rewarded

� Have the cited articles at hand, to refresh your memory

56

Introduction—original contribution

� This is an essential part of the introduction� saying what is new

� suggesting why it’s better

� Describe your contributions clearly, referring to previous work to show the improvements

� Use mainly words, no formulas

� Don’t anticipate the technical results, especially the experimental ones

� You may organize the contributions in list form

57

Introduction—outline, notations

� The last paragraph of the introduction may contain a brief outline of the paper

� It’s a substitute for a contents

� Letters may not need an outline

� You can merge the outline with the previous paragraphs, describing each section as you advance in the introduction

� Main notations can be grouped here, but also given later

58

Introduction—some rules

� Be careful not to repeat sentences from the abstract or the conclusions

� You can repeat the ideas, but try to vary the form

� Be less technical—more plain language (but don’t make it trivial !)

� Write the introduction after you have shaped the paper

� Work more on the introduction, here is where your “literary” skills are most needed

59

Body of the paper

� It contains mainly your original contribution, so

you are free to choose the best way to express it

� Try to find the simplest way for the reader to

understand your ideas

� Don’t describe how the idea came to you, the

reader does not want to replicate your efforts, but

to understand as quickly as possible

60

Make a plan

� When attacking the first “final” version, you must have a presentation plan

� Decide� what is the exposition order

� how the ideas flow from one to the next

� where do you place the proof of each idea

� how to organize experimental evidence (tables, figures)

� Don’t be afraid to throw away some of the old text if it does not go according to the plan

61

The skeleton-flesh technique

� Write first a skeleton of the paper� section-subsection titles

� sketches of main results

� figures and tables

� notes, comments, etc.,

� Write informally, only for remembering

� Try to put all main ideas there, even if formulated in a very short form

� Then…

62

Add flesh

� Replace the short notes with full text versions, in

the order you feel easier

� Don’t polish too much

� Try to have the plan in mind all the time and

check occasionally if you still follow it

� Whenever you consider fit, add “bones” to the

skeleton

63

Theorems, propositions and lemmas

� If you have theoretical results, it is helpful to

structure them formally as theorems, etc.

� Ideally, a theorem is self-supported, i.e. all

necessary information is in its body

� Structure the result upon importance

� lemma: technical auxiliary result, used e.g. for

demonstrations

� proposition: standalone result, not especially important

� theorem: standalone significant result, non trivial proof

64

Comments, remarks

� Significant results should be commented

� Elaborate on their significance !

� The comments can be structured formally as

“comments” or “remarks”

� Dedicate a comment to each aspect, don’t mingle

them

� Try to be precise, even though some comments

refer to intuition offered by the result

65

Use examples

� Illustrate your results with simple examples, they

help immensely

� Simple does not mean trivial or artificial !

� The best type of example is a typical model, to

which you apply your method

� It is especially nice if you can carry an example in

several stages, adding features as you advance

66

Experimental results

� At least two purposes:

� to illustrate your theoretical results

� to show their benefits

� Compare the results obtained with your approach

with previous methods

� It should be clear that your method is better at

least in some scenarios

� Organize the experimental results such they are

easy to understand

67

Design examples

� A detailed design example is almost always a

good illustration of a method

� Get design data from previous literature or from a

practical application

� Show clearly that your design is better

� Try to find simple comparisons, based e.g. on few

numbers (criteria, performance indices, other

quality measures)

68

How much data ?

� How much experimental data should you give ?� enough to support your claims,

� but not so much that it’s hard to follow

� How to organize ?� relevant values (average, deviation, etc.) in the case of

many runs depending on some arbitrary factors (e.g. simulated noise)

� typical scenarios: one representative scenario out of many you have tried (always mention the extent of your experiments, even if you present only a few)

69

Graphs and tables

� Graphs and tables are best means for presenting

experimental results

� Ideally, each graph or table should illustrate a

single property/behavior of your method

� Aggregate information is acceptable if unitary in

some way (e.g. errors and execution times for

approximation methods solving a problem)

� At most 4-5 curves in the same graphs, with

easily distinguishable lines (and legend !)

70

Captions

� Each figure or table must have a caption

� Ideally, the caption is self consistent: it explains completely the figure/table

� However, the caption should be not very long

� Refer to the text if needed: “Execution times for the three methods compared in Example 2”

� Conversely, each figure/table must be referred to in the text; the discussion in the text is normally longer than the caption

71

Interpretation of the results

� Sometimes you can draw some conclusions out

of the experimental results, other than the simple

“my method is the best”

� Try to go from specific to general

� Do not attempt to explain unexpected results, if

they are scarce, just state their existence

� Do not make far-fetched claims

72

Conclusions

� The final section of the paper is another abstract

� However, now the reader has gone through your

article, so don’t repeat sentences from the

abstract or the introduction

� Point out your main contributions, referring to

specific results given in the article (theorems,

experiments, etc.)

� Give a general conclusion resulting from the

experimental evidence

73

Conclusions ctd

� Point out the major advances with respect to

previous work, as resulting from the article

� Last sentences can be dedicated to future work

that you have in mind

� This is good for claiming your interest and

showing that the problem has more research

potential

� However, don’t forecast any results, just outline

the direction of the future research

74

Bibliography

� You must do a thorough bibliographical research and the article must show it

� A good bibliography may contain 10-40 entries

� Take care to cite� papers that started the problem

� most recent papers on the topic

� all articles that are relevant to your approach

� Cite journal articles instead of conference papers

� Cite books only for standard results

75

Sources

� Most of the bibliographical search must be done

before the writing

� You must make sure that your idea is original

� Bibliographical sources: article databases

� Scopus (subscription needed)

� ieeexplore (IEEE members can search, subscription

needed for articles)

� Google scholar (free search)

� other databases…

76

Informal sources

� After finding interesting titles and (maybe)

abstracts

� google the title, maybe the authors have put the article

on their web page

� ask a friend from a university with subscription to

databases

� find the email of an author and ask a pdf of the article:

you’ll be surprised how many authors reply kindly

77

How to search ?

� Search combinations of relevant keywords

� After finding an interesting article, search� the articles in its bibliography

� articles by the same authors

� articles citing this article

� In the beginning it’s difficult to see quickly if an article is relevant, but you’ll get it in time

� Try to organize the articles in categories, it will be useful later

78

Citation rules

� You must refer in the text to all entries in the

bibliography: you must show that you have used

those papers, not that you have read them

� Cite whenever you take a result from another

article. Never say “it is well known…”

� Give details if possible:

� “it is shown in [4, Th.3] that…”

� “we take the data from [7], Example 1”

79

Other citation rules

� Avoid copying sentences or paragraphs, even if

you quote and give the source

� Avoid grouping many citations

� “there are many methods for designing filters [1-25]”

� “there are many methods for designing FIR [1-14] and

IIR [15-25] filters”

� If you cite papers in groups, put at most 2-3

papers in a group

80

Final citation rules

� Golden rule: you cite a paper because it contains

information that is important for your article

� You don’t cite a paper because

� it’s famous, but you haven’t read it

� it’s recent, and you need recent citations otherwise

your article may look outdated

� it’s recent and its authors might be your reviewers

� it’s fancy to cite that old German or Russian paper

� it’s cited in other papers you have used

81

Acknowledgments

� Short section or paragraph in which the authors

� mention a source of financing for their research

� thank to other researchers that have helped making

the article better by suggesting bibliography, proof

reading, even giving raw ideas

� thank to students or other personnel for running

experiments or for other non-creative jobs

� thank to the reviewers, if they indeed contributed with

good suggestions—only in a revised version

82

Appendices

� Appendices come at the end of the article

� They may contain

� Long or technical proofs that can be skipped at a first

reading

� Short reviews of known results, usually from another

field or topic, that are used in the article

83

4. Style

� The purpose of a scientific article is to convey

information directly and explicitly

� Writing should be clear and unambiguous

� Hence, articles should be “style-less”

� However, a personal touch, not impeding on

clarity, is welcome

� We discuss here only a few issues regarding

basic problems of writing style

84

Style golden rules (Hengl & Gould 2002)

� TAKE A READER'S VIEW: write for your audience not for

yourself

� TELL A STORY: keep a clear focus in the paper and present

only results that relate to it

� BE YOURSELF: write like you speak and then revise and

polish

� MAKE IT SIMPLE: use simple(st) examples to explain complex methodology

� MAKE IT CONCRETE: use concrete words and strong verbs,

avoid noun clusters (more than three words), abstract and

ambiguous words

� MAKE IT SHORT: avoid redundancy, repetition and over-

explanation of familiar techniques and terminology

� TAKE RESPONSIBILITY: make a clear distinction between

your work and that of others

� MAKE STRONG STATEMENTS: "We concluded... “, not "It

may be concluded... "

� BE SELF-CRITICAL: consider uncertainty of conclusions and their implications and acknowledge the work of others

86

Personal vs. impersonal

� Who is telling the story ?

� “we” (first person)

� “the authors” (third person or impersonal)

� There are journals and conferences

recommending to avoid the use of “we”, as

showing a subjective position

� “Science is impersonal” !

� Science is about truth, not opinions

87

Example

� Which one do you prefer ?

� we prove Johnson’s conjecture

� Johnson’s conjecture is proved

� the algorithm was implemented in Matlab

� we implemented the algorithm in Matlab

� it results from the experiments that…

� we conclude from the experiments that…

88

Why “we” ?

� Common sense is for “we”

� “we” shows clearly that the action was performed by

the authors

� it is a claim of responsibility, so it is stronger

� in a mathematical proof there is hardly place for “we”,

but in experimental sciences there are choices to be

made

� “we” is warmer—more appeal to the reader

89

When is “we” ok ?

� Use “we” whenever referring to an action

performed by the authors

� we have implemented the test…

� we have obtained the following results

� we have proved the theorem using…

� When “we” is not proper ?

� “from (4) and (6), we have a=b” (the equality holds and

that’s all, we don’t “have” anything)

90

Participative “we”

� Sometimes “we” is meant to include the reader

� “We” = authors + reader

� It is debatable if this helps

� the reader may feel more involved

� it may be confused with “we, the authors”

91

Single author: “I” or “we”

� Following the logic of “show clearly who performs

the action”, “I” should win

� However, “I” is seldom used

� “I” is maybe too strong and too personal

� My opinion: I have used “we”, partly because I

didn’t dare to use “I”

92

Active vs. Passive Voice

� Active: subject does action

� Passive: action is done by subject

� Passive: action is done

� Examples

� the passive voice should be avoided

� avoid the passive voice

� it is shown in Figure 3 that…

� Figure 3 shows that…

93

When passive voice is good

� Use active voice:

� sentences will become clearer and shorter

� it is easier to understand

� usually, it does not decrease objectivity

� Passive voice may be good

� when the agent is not important and may be omitted

� to emphasize the object of the action

� However, most passive constructions have a

good active equivalent

94

Tenses

� Follow your common sense in choosing tenses

� Present is the time of writing (and of reading !)

� You have to use present for whatever you think is

perennial

� Present perfect is used to describe your actions

that have led to the results

� Past is for actions before the time of your

research (so, mostly other peoples’ actions)

95

Time scale

Past Near past Present Future

Time of

other

people’s

research

Time of

your

research

Time of

your

writing and

of others

reading

?

tt

96

Tenses—present

� Present is the basic tense:

� “Our main goal in this paper is…”

� “The algorithm provides a solution…”

� “Let R denote the covariance matrix”

� “The OS algorithm derives from (27) and consists of

the following steps”

� “The second example investigates the parameter

estimation performance”

97

Tenses—present perfect

� For your actions when conducting the research

� “We have also considered MA signals with zeros well

inside the unit circle”

� For other actions, when appropriate

� “Assume that N data samples have been collected”

� In the conclusions

� “Two novel methods for the estimation of the

parameters of a moving average signal have been

introduced”

98

Tenses—past

� For referring to other research

� “A similar idea was used in [14]”

� “This idea, which was utilized in [18] and [21] for FIR

filter design…”

� However, use present perfect if you refer to

collective efforts (still going on, possibly)

� “To “factorize the unfactorizable,” researchers have

tried to correct the estimated MA covariance

sequence”

99

Tenses—future

� Easy rule: use future only about “future work”

� Occasionally, you may use future with reference to actions that appear later in your article� m is an integer whose choice will be discussed shortly

� such an assumption means no restriction for the second-order statistics that will be considered throughout this paper

� Otherwise, avoid future� Hence, we will obtain estimates of the MA parameters

by minimizing the following criterion: …equation…

100

Words

� Avoid long sentences

� Use the right word

� Don’t use fancy words

� Be consistent: name each notion in a single way

� If you give a “method”, name it “method” in the whole

paper, not “procedure” or “algorithm”

� If a is first referred to as “coefficient”, don’t name it

later “constant”, “element” or “value”

101

Hyphenate to avoid confusions

� Many qualifiers before a noun may be confusing

� A gradient descent bounded region method

� Hyphenate to make it clearer

� A gradient-descent bounded-region method

� Alternatively, change topic

� A bounded region method using gradient descent

� Or change topic and hyphenate

� A bounded-region method using gradient descent

102

British vs. American English

� Many small differences (see wikipedia: American and British English differences)� optimization vs. optimisation

� color vs. colour

� Ph.D. vs. PhD

� Try to be consistent

� However, it’s much more important to use proper English—many grammar or spelling mistakes will make your paper look bad, no matter the contents

103

Comma

� A comma can change completely the sense

� Such errors are spotted easier when rereading a

whole paragraph or section

� Which is correct ?

� “The authors wish to acknowledge their co-workers,

Superman and Batman.”

� “The authors wish to acknowledge their co-workers,

Superman, and Batman.”

Formulas

� Punctuation in formulas: like formulas would be

words. Example: ”taking into account that

it results that...”

� Some journals avoid punctuation in formulas,

which is a pity

104

1

( ) ( ) ( ),N

i

i

y t h y t i e t=

= − +∑

Symbols as words

� In the beginning of a proposition, don’t treat

symbols as words of their own

� Write ”The velocity v was measured” instead of

”v was measured”

� Write ”Equation (3) shows”, not ”(3) shows”

� In the middle of a sentence, it can be accepted

� However, it is good to write sometimes ”the

velocity v” just to remind what the symbol denotes

105

106

Judgment words

� Avoid judgments not supported by doubtless

evidence

� Think twice before writing “obviously”, “clearly”, “it

is well known”, “easily”

� You must keep an objective position

107

Acronyms

� Acronyms are helpful, but don’t abuse

� Define the acronym at its first occurrence

� Redefine if used much later (or don’t use at all)

� Try to make acronyms easier to remember by

changing words order, adding or omitting letters

� WISE—weighted integral of the squared error

� RoC—region of convergence

108

Other writing issues

� See excerpts from Kristin Cobb’s course at

http://www.stanford.edu/~kcobb/writing

(Cobb_Sciwri_style_notes.ppt)

109

Plagiarism

� Plagiarism is not only punishable, but also stupid

� You’ll be caught, especially in a good journal, and punished with publication interdiction for a certain period, plus bad publicity

� A few rules for paraphrasing (from Cobb !)� Use your own words

� Work from memory

� Draw your own conclusions

� Do not simply re-arrange the original author’s words

� Do not mimic the original author’s sentence structure

110

Other sins

� Self plagiarism: to copy paragraphs from your previous papers� easy to detect, reviewers tend to search your old

papers; your paper will be probably rejected

� Fabrication: to invent data supporting your theory� hard to detect, but once detected, you’re normally out

of the research community: red card

� Omission: not to report data against your theory� easier to detect, but you may claim not making that

class of experiments: yellow card

111

Is the paper ready ?

� A paper can be always improved, but at some

point you have to submit it

� Even if the paper is not perfect, try to eliminate

ALL typographical, mathematical and

grammatical errors

� Take a break (2-3 days at least), then read the

version you want to submit as coldly as you can

� If you made many corrections, then take an other

break and repeat

112

A good quote

� M.A.Morrison, “Tips on Scientific Writing”:

� Professionals do not submit error-ridden documents. You can almost guarantee that your paper will antagonize readers, reviewers, and editors by leaving technical errors in it. Eliminating technical errors from a paper requires time, effort,patience, and persistence. It is hard work that you must do. Runeach draft through a spell checker. Check your figures. Check your tables. Check your references. Get a friend or two to proof it for you. Do whatever is necessary. But never submit a sloppy, error-ridden paper. You've invested precious time and energy in your work; your work deserves the best presentation you can give it.

113

Last details before submission

� Prepare pdf file in required format

� Prepare a few keywords, often from a list given by the journal. Choose carefully, they determine who will manage the review process

� Fill copyright transfer form, if needed

� Write a cover letter, if required� Usually only: “Dear Editors, please consider our article

‘Title’ for publication in journal X”. Signed: Y, author

� Maybe also: a sentence or two regarding your original contribution or a claim on the benefits

114

After submitting the paper…

� Once the paper is submitted, you should be

prepared to wait 2-4 months, even more, for the

results of the review process

� You can work on the same topic or another, the

only forbidden act is to submit the same paper or

a slightly different version to another journal

115

Double submission

� Do not send similar manuscripts to different journals, hoping that one is accepted� they may go to the same reviewer !

� if one is accepted, you’ll withdraw the other ?

� Can you send a shorter version to a conference ?� yes, but better before submitting the article

� cite (or mention) the conference submission in the article

� take care that the article contains significant new information

116

Upgrading rules example

� From the rules of IEEE Signal Processing Society

� “It is acceptable for conference papers to be used as

the basis for a more fully developed journal

submission. Still, authors are required to cite related

prior work; the papers cannot be identical; and the

journal publication must include novel aspects”

5. Review process

� You have sent the paper to a journal

� What happens there ?� A member of the editorial board (AE—associate editor)

chooses reviewers; this takes a week or so

� The reviewers evaluate the paper and send their reports to the AE (6-8 weeks normally, but often more)

� The AE makes a decision and sends it to you, together with reviewers’ reports (one more week)

� You will usually know who is the AE, but the reviewers are anonymous

118

Editorial board

� The typical editorial board� Editor-in-chief (usually one)

� Associate editors (many: 20-50, even more)

� Administrative staff

� Your paper goes to an AE, chosen by the EIC or by a publication manager

� Choice is dictated by keywords, title, abstract, author affiliation

� Sometimes you may send paper directly to an AE

119

AE activities

� The AE reads quickly your paper, then

� either starts the review process

� or proposes immediate rejection, if the paper does not

meet the technical standard of the journal (it’s visibly

bad) or the topic is not appropriate

� The AE chooses 2-3 reviewers, even more (I had

5 reviewers at a paper and know of a max of 6)

� Based on reviewers’ reports, the AE makes a

decision

120

Reviewer report

� Each reviewer writes a report containing

� a general assessment of your paper

� objections to the method, the planning of the

experiments, the organization of the paper

� improvement suggestions

� The reports may be extremely diverse, see the

two examples

� Also, the reviewer grades your paper on

originality, technical merit, writing, English

Grading example (IEEE TSP)

� Suitability of topic

� 1. Is the topic appropriate for publication in these transactions?: Yes / Perhaps / No

� 2. Is the topic important to colleagues working in the field?: Yes / Moderately so / No

Grading (2)

� Contents� 1. Is the paper technically sound?: Yes / No

� 2. Is the coverage of the topic sufficiently comprehensive and balanced?: Yes / Important information is missing or superficially treated / Treatment somewhat unbalanced, but not seriously so / Certain parts significantly overstressed

� 3. How would you describe technical depth of paper?: Superficial / Suitable for the non-specialist / Suitable for the generally knowledgeable individual working in the field / Suitable only for an expert

� 4. How would you rate the technical novelty of the paper?: Novel / Somewhat novel / Not novel

Grading (3)

� Presentation

� 1. How would you rate the overall organization of the paper?: Satisfactory / Could be improved / Poor

� 2. Are the title and abstract satisfactory?: Yes / No

� 3. Is the length of the paper appropriate?: Yes / No (If not, recommend how the length of the paper should be amended)

� 4. Are symbols, terms, and concepts adequately defined?: Yes / Not always / No

� 5. How do you rate the English usage? : Satisfactory / Needs

improvement / Poor

� 6. Rate the Bibliography: Satisfactory / Unsatisfactory

Grading (4)

� Overall rating� 1. How would you rate the technical contents of the

paper?:

� 2. How would you rate the novelty of the paper?:

� 3. How would you rate the "literary" presentation of the paper?:

� 4. How would you rate the appropriateness of this paper for publication in this IEEE Transactions?:

� All these are graded on a scale from 1 to 10, with Excellent 8-10, Good 5-8, Fair 3-5 and Poor 1-3.

125

AE decision

� Typical decisions (IEEE style)

� A (accept as it is)

� AQ (accept with minor changes)

� RQ (revise and resubmit)

� R (reject)

� AE decision is usually an “average” of reviewers’

recommendations

� What do you do in each case ?

126

Accept

� A after the first round of review means that� either you’re a genius

� or it’s a bad journal (since reviewers are careless)

� You may have to correct few details or typos, but the paper is practically published

� From now on you’ll have only to� send all files corresponding to the final version

� transfer copyright

� correct the proofs when they will be ready (after some months)

127

AQ

� AQ means that your paper is essentially good, but may be improved, especially in form rather than in contents

� Usually, AQ means some of the following� few paragraphs should be slightly reformulated

� some equations, proofs, etc., need minor corrections

� new experiments have to be made, but basically with the methods you have already used

� new bibliography should be added, but without much impact on your method

128

RQ

� RQ means that your paper is basically correct,

but your proof (theoretical and/or experimental) is

doubtful or can be significantly improved

� RQ may mean that

� the structure of your paper has to be changed

� some proofs have to be reformulated

� new experiments, involving new methods, are needed

� new bibliography is required, which may put your

methods in a new angle

129

Reject

� Several types of reject

� paper is flawed (don’t dream of resubmission)

� paper is not acceptable now, but may be reconsidered

if authors work hard (you’ll have to resubmit it)

� If you’ll resubmit, the paper will probably go to the

same AE, who will probably get the same

reviewers

� So, try to answer ALL reviewers’ suggestions

� The resubmission is treated as a new paper

130

Reject DON’Ts

� Even if the reviews seem blatantly unfair, don’t

complain immediately to the AE. Cool off first !

� Don’t hope that the AE will trust you more than

the reviewers. You must have a hard case to

change AE’s mind

� Don’t challenge the reviews in matters of opinion,

but only if you can prove them wrong with facts

� Don’t complain to the EIC—he will support the AE

131

Reject DOs

� Take the good part of it: you have 2-3 expert

opinions on your paper

� Remember that reviewers would be happy to

read a good paper, so, if they didn’t like your

paper, there must be some reason

� Try to take maximum advantage from reviewers’

comments: improve the paper !

132

6. Revising a paper

� See how much time you are allowed for revising

� Start by understanding what the reviewers want� Print the paper as it was seen by the reviewers

� Print reviewers’ comments

� Read the comments one by one, marking the affected paragraphs in the paper

� Tag the comments: very important/not so important, difficult/easy

� Don’t get angry if the comment seems stupid: try to get the reader’s viewpoint—maybe your text is not so clear

133

Preparing a revision—the hard work

� After getting reviewers’ points, do first the difficult

tasks

� Read new bibliography and see how it relates to your

contribution

� Complete/change proofs

� Run new experiments, compare with other methods

� Think how all these will affect the structure of your

paper

� Don’t touch the paper in this stage

134

Preparing a revision—the revised paper

� You are now ready to write the revised paper

� Write the new text with a different color: the reviewers

will spot it easily

� Each time you have made the modifications that answer

a comment, mark it as solved

� Try to make the modifications in a logical order, e.g. from

the beginning of a section to its end

� Don’t be afraid to make corrections/modifications not

required by the reviewers, but keep them rather small

135

Take a break—2-3 days

136

Preparing a revision—the reply

� Write a letter to the AE describing the changes

� Structure: many paragraphs of the form

� Copy of reviewer comment (or clear reference to it)

� Description of modification, arguments, etc.

� Be specific: give page, eq. numbers, describe

modifications as clearly as you can

� Try to modify the paper as an answer to that

comment. Reviewers appreciate even small steps

taken to implement their suggestions

137

Strategies for satisfying the reviewers

� Best strategy: answer all comments by making

modifications in the paper

� If you cannot answer to a comment (because it’s

difficult or it ruins your theory), you may gamble

� try to refute the comment, without modifying the paper

� answer thoroughly the other comments, trying to get

the approval of two reviewers

� hope that, with 2 AQs and 1 R, the AE decides

publication (it’s not necessarily so)

138

Reply DOs and DON’Ts

� Thank the AE and the reviewers

� Don’t use lengthy arguments

� Don’t try to look smarter than the reviewer

� Be polite, be polite, be polite

� Don’t be overly polite, it might look strange

� Remember that the best reply is a correct one

139

Preparing the revision—the end

� Check that you have answered all comments and

read again the paper

� Submit the new version, taking care to see where

to upload the reply

� A new review round starts

� The outcome will be again an AE decision

� Take care: some journals don’t accept RQ twice.

They reject the paper at the second RQ

140

Withdrawing a paper

� At any point in the review process you can

withdraw a paper by writing to the AE

� Reason ?

� You have discovered a major flaw

� You cannot do what the reviewers request

� In both cases, think again !

� In the second case, think twice more

141

You have published a paper ! What next ?

� Be happy !

� Let people know that you have published

� put the article in your publication list on your web page

� put also a preliminary version there (beware of

copyright issues !) or a link to the journal website

� cite it in future papers, if appropriate (a reviewer feels

safer if the author is not a “nobody”)

� Don’t forget the writing experience, but don’t rely

completely on it for the next paper

142

Improve your style

� When reading articles, keep an eye for the style

and writing technique

� Imagine how you would tell the story

� Grade writing in other papers

� When you say “this is nicely/badly written”, try to

realize what are the reasons of your grade

� Write on a regular basis, not necessarily for

immediate publication

143

However… never forget:

You can write a good scientific article

only if

you have something new to say

144

Online bibliography

� Search these (among others)� K.Cobb, Scientific Writing,

http://www.stanford.edu/~kcobb/writing

� E.R.Firestone, S.B.Hooker, ”Careful Scientific Writing: A Guide for the Nitpicker, the Novice, and the Nervous”, 2001

� T.Hengl, M.Gould, ”Rules of Thumb for Writing Scientific Articles”, 2002

� M.A.Morrison, “Tips on Scientific Writing”, 2004

� M.E.Tischler, “Scientific Writing Booklet”