1

Research Methods:Technical Writing

James Gainjgain@cs.uct.ac.za

Research Methods

Outline

Approaching WritingGetting PublishedSome Key Elements ofTechnical StyleYour ReportA Research Paper

Research Methods

Why Learn to Write Well?

It takes lots of practice, so why bother?Because it is one of the most valuable life-longskills

Most CS careers require writing:Research - proposals, research notes, literaturesurveys, paper reviews, conference and journalpapers, thesesIndustry - code comments, documentation,reports, memos

The purpose is communication notobfuscation

Research Methods

How to begin?

Bottom-upDescribe details and link them togetherLeads to unstructured mess

Top-downStart with structure and flesh outLeads to shifting structure as you progress

Bi-directionalWrite notes as you do research (bottom-up)Then structure your thesis/paper around a message (top-down)Then fill in the structure with details (bottom-up)

2

Research Methods

High-level Issues

Your writing should have a messageAn argument (hypothesis) for which yourresearch provides evidenceMessage must be reflected in the title, abstract,introduction, conclusion and body of yourwriting

Aiming to be understood is not sufficient:Write so that you cannot be misunderstoodAssume your audience is intelligent but:

(a)ignorant and (b) given to misunderstandingState key ideas transparently, prominently andoften

Research Methods

Outline

ApproachingWritingGettingPublishedSome KeyElements ofTechnical StyleYour ReportA ResearchPaper

Research Methods

Getting Published -The FlowChart

Submit

Referees

Editor’sDecision

Accept

In Print at Last!

Reject

Revise

StillRelevant

TrySomewhere

Else

Yes

No

RevisedManuscript

Try Again!

Research Methods

Submission

Submission:

• Don’t bother too muchwith ‘Instructions toAuthors’

• Never submit tomultiple destinationssimultaneously

Submit

Referees

Editor’sDecision

Accept

In Print at Last!

Reject

Revise

StillRelevant

TrySomewhere

Else

Yes

No

RevisedManuscript

Try Again!

3

Research Methods

Refereeing

Refereeing:

• Single or Doubleblind

• Review quality isoften proportional toreview length

Submit

Referees

Editor’sDecision

Accept

In Print at Last!

Reject

Revise

StillRelevant

TrySomewhere

Else

Yes

No

RevisedManuscript

Try Again!

Research Methods

Acceptance

On Acceptance:

1. Minor changesand formatting

2. Galley proofs

3. Receive journalcopies or off-prints

Submit

Referees

Editor’sDecision

Accept

In Print at Last!

Reject

Revise

StillRelevant

TrySomewhere

Else

Yes

No

RevisedManuscript

Try Again!

Research Methods

Revision

Revision Options:• Treat as a rejection

• Make at least 80% ofthe suggestedchanges (in a collageformat)

• Argue the toss (withthe editor not thereviewers)

Submit

Referees

Editor’sDecision

Accept

In Print at Last!

Reject

Revise

StillRelevant

TrySomewhere

Else

Yes

No

Try Again!

RevisedManuscript

Research Methods

Rejection

On Rejection:

• Damage dependson the reviewingdelay and comments

• May have to submitto a less prestigiousdestination

Submit

Referees

Editor’sDecision

Accept

In Print at Last!

Reject

Revise

StillRelevant

TrySomewhere

Else

Yes

No

Try Again!

RevisedManuscript

4

Research Methods

Outline

Approaching WritingGetting PublishedSome Key Elements of Technical StyleYour ReportA Research Paper

Source: W. Hopkins,“Guidelines on Style forScientific Writing”, SportsScience, 3(1), 1999

Research Methods

The Basics

Submit by the deadlineKeep to the length restrictions

Do not narrow the marginsDo not use 6pt font

On occasion, supply supporting evidence(e.g. experimental data, or a written-outproof) in an appendix

Always use a spell checker

Research Methods

Visual Structure

Give strong visual structure to your paperusing

sections and sub-sectionsbulletsitalicslaid-out code

Find out how to draw pictures, and use themCan the reader understand the paper usingthe diagrams (and captions) alone?

Research Methods

Citations

Serve to:Acknowledge the work of othersDirect the reader to additional sources ofinformationAcknowledge conflicts with other resultsProvide support for the views expressed in thepaperBroadly, place a paper within its scientificcontext, relating it to the present state of the art

An unsupported statementSure sign that either a reference is needed or asupporting argument

5

Research Methods

Citation Styles

There are many styles. Choose one and apply itconsistently.Example: ACM Style

Journal - Anderson, R.E. Social impacts of computing:Codes of professional ethics. Social Science ComputingReview 10, 2 (Winter 1992), 453-469.Conference - Mackay, W.E. Ethics, lies and videotape, inProceedings of CHI '95 (Denver CO, May 1995), ACM Press,138-145.Book - Schwartz, M. Guidelines for Bias-Free Writing. IndianaUniversity Press, Bloomington IN, 1995.Citing in the text - [1] [3, 15]

Other styles include Harvard, IEEE

Research Methods

Exercise: Citations

Place ACM-style citation labels in the following textwhere required:

“The field is well researched and Bechmann and Milliron et al.provide useful surveys. Typically, deformations are specified bymanipulators, including parametric hyperpatches, points, curves,twisting frames and 2-1/2 D surfaces.”

Solution:“The field is well researched and Bechmann [1] and Millironet al. [2] provide useful surveys. Typically, deformations arespecified by manipulators, including parametrichyperpatches [3, 4], points [5], curves [6, 7], twisting frames[8] and 2-1/2 D surfaces [9].”

Research Methods

Viewpoint Usage

Rule:Never use the 1st person singular (‘I’)

Third person is preferredNot - “I found out when I ran pilot experiments that theinitial design suffered from my personal bias.”Rather - “On running pilot experiments it was found that theinitial design suffered from experimenter bias.”This sometimes necessitates passive voice (subject last)

Use of 1st person plural (‘We’)Use where the sentence would otherwise become toocontortedEven if you are the only author

Research Methods

Exercise: 3rd Person

Convert to a technical viewpoint:“As I approached the road that cut through the New RiverMesa, I noticed that there were seven layers. Looking atthe lowermost layer it seemed to me to be an arkosicsandstone.”

Solution:“Where the road cut through the New River Mesa, sevenlayers were noticeable. The lowermost of these layersseemed to be an arkosic sandstone.”

6

Research Methods

Use the Active Voice

The passive voice is “respectable” butit DEADENS your writing. Avoid.

We can see that...It can be seen that...

This might seem like atype error

It might be thoughtthat this would be a

type error

We wanted to retainthese properties

These properties werethought desirable

We ran 34 tests34 tests were run

YESNO

“We” =you and

thereader

“We” =the

authors

Research Methods

Use Simple, Direct Language

The ball moved sidewaysThe object under study was

displaced horizontally

The garbage collector wasslow

It could be considered thatthe speed of storage

reclamation left somethingto be desired

Find outEndeavour to ascertain

YearlyOn an annual basis

YESNO

Research Methods

Reminder: Tense

Tense shows position in time (past, present, future)Types:

Simple (most basic)Continuous (ongoing)Perfect (completed)Perfect continuous (ongoing actions that will be completed atsome definite future time)

will have exploredwill be exploringwill/shall exploreFuture

has exploredis exploringexplore/sPresent

had exploredwas exploringexploredPast

PerfectContinuousSimple

Research Methods

Tense Usage

Present Simple and Perfect predominate in scientificwriting:

The work exists now and is timely but may have started inthe pastExample - “From-point visibility algorithms are less costlycomputationally than from-region approaches”

Except:Use past tense to report results.

• “in our experiments we found that …”But use present tense to discuss them.

• “a simple explanation of these findings is that …”

7

Research Methods

Exercise: Conciseness

Reword the paragraph to make it concise:“Virtually all experienced writers agree that any written expression thatdeserves to be called vigorous writing, whether it is a short story, an articlefor a professional journal, or a complete book, is characterized by theattribute of being succinct, concise, and to the point. A sentence--nomatter where in the writing it occurs--should contain no unnecessary orsuperfluous words, words that stand in the way of the writer's directexpression of his or her meaning and purpose. In a very similar fashion, aparagraph--the basic unit of organization in English prose--should containno unnecessary or superfluous sentences, sentences that introduceperipheral content into the writing or stray from its basic narrative line. It isin this sense that a writer is like an artist executing a drawing, and it is inthis sense that a writer is like an engineer designing a machine. Goodwriting should be economical for the same reason that a drawing shouldhave no unnecessary lines, and good writing should be streamlined in thesame way that a machine is designed to have no unnecessary parts,parts that contribute little or nothing to its intended function.”

Research Methods

Solution: Conciseness

“Vigorous writing is concise. A sentence should contain nounnecessary words, a paragraph no unnecessary sentences,for the same reason that a drawing should have nounnecessary lines and a machine no unnecessary parts.”34 wordsBe careful not to overdo it. Some concepts need to beexplained in detail.

Research Methods

Flow of Ideas (Cohesion)

At a sentence levelOne sentence linked to the next

At a paragraph levelFirst sentence sets the topicNo unlinked ideas in the paragraph

At a section levelOutline firstDon’t repeat or contradict other sections

At a document levelCreate a logical and cohesive outline supporting themessageSet the draft aside for a while, get others to read it

Research Methods

Outline

Approaching WritingGetting PublishedSome Key Elements of Technical StyleYour ReportA Research Paper

8

Research MethodsTell a story

Project Write-up

This is what determines your mark!(Very Last) Abstract

1. (Last) Introduction: Aims, importance, outline2. (First & ongoing) Background3. (Second) Theory/Algorithms4. (Third) Application of Theory/Algorithm

Implementation5. (Fourth) Experiment: Design + Results + Discussion

of Results6. (Last) Conclusion — Tie up with aims:

“we said we would and we did”, except (oops) some didn’twork, and (wow) we found an amazing unexpectedthing, but now we would do this … (future work) Research Methods

Outline

Approaching WritingGetting PublishedSome Key Elementsof Technical StyleYour ReportA Research Paper

Source: S. Peyton Jones,“How to write a great researchpaper”, Microsoft Research,Cambridge

Research Methods

Writing a Paper

The purpose of writing a researchpaper is to communicate your ideas toyour peers

This is more limited than the projectresearch report or dissertation or thesis

Each paper must have a central ideaWith evidence to support it

Research Methods

The Idea

A re-usable insight, useful to the readerFigure out what your idea isMake certain that the reader is left in nodoubt about the idea or contribution

Be 100% explicit:“The main idea of this paper is....”“In this section we present the main contributionsof the paper.”

Many papers contain good ideas, but donot distil what they areThe reader is interested in ideas not artefacts

9

Research Methods

Here is a problemIt’s an interesting problemIt’s an unsolved problemHere is my ideaMy idea works (details, data)Here’s how my idea compares toother people’s approaches

Your narrative flow

I wish Iknew howto solve

that!

I see howthat works.Ingenious!

Research Methods

Structure

TitleAbstractIntroductionThe problemMy ideaThe detailsRelated workConclusionsReferences

(1000 reader)(4-8 sentences, 100 readers)(1 page, 100 readers)(1 page, 10 readers)(2 pages, 10 readers)(4 pages, 3 readers)(1 page, 10 readers)(0.5 pages, 20 readers)(1 page, 10 readers)

Research Methods

Abstract: structure

Write the abstract lastUsed by program committeemembers to decide which papers toreadFour sentences [Kent Beck]

State the problemSay why it’s an interesting problemSay what your solution achievesSay what follows from your solution

Research Methods

Abstract: example

Many papers are badly written and hard tounderstandThis is a pity, because their good ideas maygo unappreciatedFollowing simple guidelines can dramaticallyimprove the quality of your papersYour work will be used more, and thefeedback you get from others will in turnimprove your research

10

Research Methods

Introduction: structure

Describe the problemProviding context

State your contributionsExplicitly

And that is all

Use an exampleto introduce the

problem

Bulleted list ofcontributions

Research Methods

Introduction: state yourcontributions

Write the list of contributions firstThe list of contributions drives the entire paper:

The paper substantiates these claims

Reader thinks:“Wow, if they can deliver on this … I’d better read on”

Do not leave the reader to guess what yourcontributions are!

“In this paper we …”“We explain precisely what … surprisingly this has not beendone before”“… articulating this is one of our main contributions”

Research Methods

Introduction: contributionsshould be refutable

YES!NO!

We have built a GUI toolkit inWizWoz, and used it to implement atext editor (Section 5). The result ishalf the length of the Java version.

We have used WizWoz inpractice

We prove that the type system issound, and that type checking isdecidable (Section 4)

We study its properties

We give the syntax and semantics ofa language that supportsconcurrent processes (Section 3). Itsinnovative features are...

We describe the WizWozsystem. It is really cool.

Research Methods

Introduction: No “rest of thispaper is...”

Not:

Instead, use forward references fromthe narrative in the introduction.

The introduction (including thecontributions) should survey the wholepaper, and therefore forward referenceevery important part

“The rest of this paper is structured as follows.Section 2 introduces the problem. Section 3... Finally, Section 8 concludes”.

11

Research Methods

Related Work

TitleAbstractIntroductionRelated WorkThe problemMy ideaThe detailsConclusionsReferences

Relatedwork

Your reader Your idea

We adopt the notion of transaction fromBrown [1], as modified for distributedsystems by White [2], using the four-phaseinterpolation algorithm of Green [3]. Ourwork differs from White in our advancedrevocation protocol, which deals withthe case of priority inversion as describedby Yellow [4].

Research Methods

Related Work: later is better

Problem 1:

the reader knows nothing about theproblem yet; so your (carefully trimmed)description of various technical tradeoffs israther incomprehensible

Problem 2:

describing alternative approaches getsbetween the reader and your idea

But delaying related work isunconventional I feel

tired

I feelstupid

Research Methods

The Body

TitleAbstractIntroductionThe problemMy ideaThe detailsConclusionsReferences

(1 page, 10 readers)(2 pages, 10 readers)(4 pages, 3 readers)

Research Methods

The Body: Presenting the Idea

The idea:“Consider a bifircuated semi-lattice D, over ahyper-modulated signature S. Suppose pi is anelement of D. Then we know for every such pithere is an epi-modulus j, such that pj < pi.”

Sounds impressive ... butSends readers to sleepIn a paper you MUST provide the details,but FIRST convey the idea

12

Research Methods

The Body: Idea 1st, Details 2nd

Explain it as if you were speaking tosomeone using a whiteboardConveying the intuition is primary, notsecondaryOnce your reader has the intuition, shecan follow the details (but not viceversa)Even if she skips the details, she stilltakes away something valuable

Research Methods

The Body: Putting the reader first

Avoid the Journey:Do not recapitulate your personal journey ofdiscovery. This route may be soaked with yourblood, but that is not interesting to the readerInstead, choose the most direct route to the idea

Use Examples:Introduce the problem, and your idea, usingexamples and only then present the generalcase

Research Methods

The details: evidence

Your introduction makes claims

The body of the paper provides evidence tosupport each claim

Check each claim in the introduction,identify the evidence, and forward-reference it from the claim

Evidence can be:Analysis and comparison, theorems,measurements, case studies, experiments

Research Methods

Structure

TitleAbstractIntroductionThe problemMy ideaThe detailsRelated workConclusionsReferences

(1 page, 10 readers)(0.5 pages, 20 readers)(1 page, 10 readers)

13

Research Methods

Related work

Fallacy:To make my work look good, I have to makeother people’s work look bad

Giving credit to others does not diminish thecredit you get from your paper

Warmly acknowledge people who have helpedyouBe generous to the competition. “In his inspiringpaper [Foo98] Foogle shows.... We develop hisfoundation in the following ways...”Acknowledge weaknesses in your approach

Research Methods

Credit is not like money

Failing to give credit to others can kill yourpaper

If you imply that an idea is yours, and thereferee knows it is not, then either

(a) You don’t know that it’s an old idea (bad)(b) You do know, but are pretending it’s yours

(worse)

Conclusion and Future Work:Be brief and too the point

Research Methods

In Conclusion

Technical writing is a skill that must be honedthrough practice

Different from other forms of writingDeliver a coherent message

Identify your key ideaUse examples

Make your contributions explicit