1 of 39

Research Methodology

연구방법론

11 - Writing the (IMRaD) Paper

2 of 39

RESEARCH METHODOLOGY · 연구방법론

8866356-02 · 2026년 2학기 · 충북대학원

09

Writing a Paper

One question · real data · a story the reader can follow

J. Zobel, Writing for Computer Science, Ch. 5 · course notes on IMRaD · courses.aaron.kr

3 of 39

WHERE WE ARE

From Reading Research to Doing It

DONE · REWRITING NOW

Literature Review

What others found. Themes, gaps.

STARTS THIS WEEK

Final Paper

What you found — from your own data analysis.

LAST CLASS

Conference talk

Present it like a real CS conference.

Your literature review becomes the Related Work section of your final paper. Nothing is wasted.

Research Methodology · Week 9 · Writing a Paper

4 of 39

PLAN FOR TODAY

Three Parts, Then the Final Paper Brief

Part

Topic

Activity

Time

A

Scope and story

1 · Sketch your Figure 1

20 min

B

IMRaD, section by section

2 · Five-sentence abstract

3 · Results or Discussion?

40 min

C

Draft to submission

4 · One-minute reviewer

20 min

—

Break

​

10 min

★

Final Paper requirements

Calendar · checklist · rubric

30+ min

Keep one question in mind all class: What data will I analyze?

Research Methodology · Week 9 · Writing a Paper

5 of 39

KEY WORDS

Words We Use Today

scope — what your paper includes, and what it leaves out

contribution — the one new thing your paper adds

IMRaD — Introduction, Methods, Results and Discussion

reproduce — repeat your study and get the same result

dataset — the data you collect, download or create

interpret — explain what a result means

venue — the conference or journal you submit to

referee — a reviewer who decides accept or reject

draft — an early version, written to be revised

overclaim — say more than your data shows

Research Methodology · Week 9 · Writing a Paper

6 of 39

A

Scope & Story

“Begin early: if it is possible to start writing, then writing should start.”

Zobel, Ch. 5 — The Scope of a Paper, Telling a Story

7 of 39

THE CORE IDEA

A research paper answers one question using evidence.

Everything in your paper should help answer that one question.

8 of 39

A · THREE TESTS

Is It a Paper Yet?

No question

→ no paper

No data

→ no research — just a project

No analysis

→ no contribution

“I built an app” is a project. “I measured whether my app helps users, and here is the data” is research.

Research Methodology · Week 9 · from course notes

9 of 39

A · SCOPE

Decide What Goes In — and What Stays Out

Which result is the most surprising?

Which one result could other researchers use?

Who are the readers? Specialists, or general CS?

Which venue? Page limit? Theory or experiments?

New idea halfway through? Write it down for the next paper. Finishing the old work matters more than adding new results.

Research Methodology · Week 9 · after Zobel Ch. 5, The Scope of a Paper

10 of 39

A · ZOBEL’S BEST TRICK

Sketch the Graph Before You Write Code

Question: Does compressing data make database sorting faster?

Before any experiment, guess the shape of the answer.

The sketch tells you what to measure (x and y), and what your code must output.

Research Methodology · Week 9 · Zobel’s external-sorting example · sketch drawn for this deck

11 of 39

MINI ACTIVITY 1 · 8 MIN · ALONE, THEN PAIRS

Sketch Your Figure 1

  1. Write your research question in one sentence.
  2. Name your dataset (or the data you will create).
  3. On paper, draw the one graph or table that would answer it. Label both axes or all columns.
  4. Draw what you expect to see. Then show your partner.

Partner asks: Can this data really make this graph? What would surprise you?

Research Methodology · Week 9 · Writing a Paper

12 of 39

A · TELLING A STORY

A Paper Is a Guided Tour

IMAGE PLACEHOLDER

Photo: a museum guide leading a small group through a gallery

Teach the past you — the you who started this project — everything needed to reach the present you.

Each “room” shows one new thing. The reader always knows why they are there.

Not a diary. Not “first we tried…, then we tried…”.

Research Methodology · Week 9 · after Zobel Ch. 5, Telling a Story

13 of 39

A · FOUR WAYS TO ORDER THE STORY

Choose a Shape for the Body

Structure

Order

Good for

Chain

Problem → old solutions → new solution → proof it is better

Most papers; most common

Specificity

The whole system first, then each part in detail

Systems with many parts

Example

A concrete case first, then the general method

Abstract or new ideas

Complexity

A simple case first, then the harder case

Step-by-step results

Most data-analysis papers use the chain: question → data → analysis → result → meaning.

Research Methodology · Week 9 · after Zobel Ch. 5, Telling a Story

14 of 39

A · PUT THE ANSWER FIRST

Write Like a Newspaper, Not a Mystery Novel

1 · The main result

2 · Context

3 · Details

Busy readers decide in a minute. Put important statements as early as possible.

Never: “as we show later, the results were surprising…” — say what was surprising.

Research Methodology · Week 9 · after Zobel Ch. 5

15 of 39

B

IMRaD, Section

by Section

What each part of your paper must do

16 of 39

B · THE MAP

Each Section Answers One Question

I

Introduction

Why does this matter?

M

Methods

What did you do?

R

Results

What did you find?

D

Discussion

What does it mean?

Our paper: Title · Abstract · I. Introduction · II. Related Work · III. Methods · IV. Results · V. Discussion & Conclusions · Acknowledgment · References

Research Methodology · Week 9 · course IMRaD notes

17 of 39

B · TITLE & KEYWORDS

Many Readers Stop at the Title

WEAK

AI and Software Development

BETTER

Do AI Code Assistants Reduce Bug-Fix Time? An Analysis of 1,200 GitHub Pull Requests

Good titles are specific, informative, short and searchable.

Keywords name your themes — not your method. Not “timing experiments”.

Research Methodology · Week 9 · course notes; Zobel Ch. 5 · example title invented

18 of 39

B · ABSTRACT · 150–200 WORDS

Five Sentences, Five Jobs

#

Job

Example sentence start

1

The broad area

“AI code assistants are now used by many developers…”

2

The specific problem

“However, their effect on bug-fix time is unclear…”

3

What exists, and its limits

“Earlier studies used small lab tasks…”

4

What you did

“We analyzed [N] pull requests from [source]…”

5

How you evaluated + results

“Fix time was [X]% shorter when…”

Be specific: not “much faster” but “60% faster”. No citations, no “we review the literature”.

Research Methodology · Week 9 · after Zobel Ch. 5, Abstract

19 of 39

MINI ACTIVITY 2 · 10 MIN · ALONE

Write Your Five-Sentence Abstract — Before the Results

  1. Use your Figure 1 sketch from Activity 1.
  2. Write one sentence for each of the five jobs.
  3. Where you have no result yet, write a placeholder: [X]%, [N] records.
  4. No AI for this one. Your own words, even if simple.

Save it. In four weeks, fill in the brackets with your real numbers.

Research Methodology · Week 9 · Writing a Paper

20 of 39

B · I. INTRODUCTION

The Abstract, Expanded

1 · BROAD

The research area

2 · CONTEXT

What others found

3 · GAP

What is still unknown

4 · QUESTION

What this paper does — and finds

  • About one paragraph for each of the abstract’s five sentences.
  • Say clearly what is new and what you found.
  • Don’t start too broad: not “Since the invention of computers…”
  • Keep deep literature for Related Work.

Research Methodology · Week 9 · Zobel Ch. 5, Introduction; course notes

21 of 39

B · II. RELATED WORK

Your Literature Review, Made Shorter and Sharper

AVOID

“[1] studied X. [2] studied Y. [3] studied Z.” — one paper at a time, no connection.

DO

Group by theme. Compare. End with the gap your study fills.

Background = what the reader must know to understand you.

Previous work = other answers to your question, and their limits.

Research Methodology · Week 9 · Zobel Ch. 5, Literature Review; course notes

22 of 39

B · III. METHODS

The Replication Test

If another student cannot repeat your study from this section, your Methods section is not finished.

Say exactly

Example

Data

Source and link, date collected, number of records, what one record contains

Cleaning

What you removed or changed, and why (“412 rows with missing labels removed”)

Analysis

Which test or model, which tool and version (Python 3.12, pandas, scikit-learn)

Measures

What each number means and how it was computed

Write paragraphs, not bullet lists.

Research Methodology · Week 9 · course notes; Zobel Ch. 5, Body

23 of 39

B · IV. RESULTS AND V. DISCUSSION

What You Found vs. What It Means

Results

  • Tables and graphs
  • Describe them briefly
  • Only the findings that matter
  • Don’t repeat every number

“Fix time fell from 41 to 29 minutes (Fig. 2).”

Discussion

  • Did you answer the question?
  • Expected or surprising? Why?
  • Compare with Related Work
  • Limitations — honestly

“The gain may depend on task size, because…”

Research Methodology · Week 9 · course notes; Zobel Ch. 5, Body

24 of 39

MINI ACTIVITY 3 · 5 MIN · PAIRS

Results (R) or Discussion (D)?

  1. The model reached 87% accuracy on the test set (Table II).
  2. This suggests that short commit messages are harder to classify.
  3. Repositories with CI had 2.3 times fewer failed builds.
  4. Unlike [4], we found no effect of team size.
  5. Our data covers only Python projects, so the result may not hold for Java.
  6. Response time rose sharply above 500 users (Fig. 3).

Bonus: which sentence would be an overclaim if it said “proves”?

Research Methodology · Week 9 · all sentences invented

25 of 39

B · THE FIGURE RULE

Figures Must Do Work, Not Decorate

Weak: no units, no meaning, shadows, 7 colors, title “Results”.

Strong: labeled axes, units, n, error bars, one highlight.

Every figure: numbered, captioned, mentioned in the text, and you tell the reader what to notice.

Research Methodology · Week 9 · both charts made for this deck with invented data

26 of 39

B · CONCLUSIONS, REFERENCES, APPENDICES

End With What Readers Should Remember

Conclusions

Main result + why it matters. Limitations. Future work.

No new figures, tables, citations or arguments.

References

IEEE style. Only works cited in the text. Every one checked.

Add DOIs when possible.

Appendix

Bulky material: long tables, code links, AI use log.

Only if needed.

Zobel’s tip: write “Conclusions” — the inferences you draw. Nothing to conclude? Call it “Summary”.

Research Methodology · Week 9 · Zobel Ch. 5; course notes

27 of 39

B · HOW LONG?

A 4–6 Page Conference Paper, Roughly

Abstract

150–200 words

I. Introduction

300–500

II. Related Work

300–500

III. Methods

400–700

IV. Results

400–700 + figures and tables

V. Discussion & Concl.

300–500

About 2,000–3,000 words in total. These are guides, not rules — the venue’s page limit decides.

Research Methodology · Week 9 · course guidelines, adapted from the course notes

28 of 39

B · IN COMPUTER SCIENCE

IMRaD Inside, Your Own Headings Outside

ZOBEL’S SORTING PAPER

I. Introduction

II. External sorting

III. Compression for databases

IV. Sorting with compression

V. Experimental setup

VI. Results and discussion

VII. Conclusions

II–III = background (“related work”)

IV–V = methods

VI = results + discussion

Headings can name your topic. The elements must all be there.

Research Methodology · Week 9 · Zobel Ch. 5, Body

29 of 39

C

Draft to Submission

How to start writing — and how papers get rejected

30 of 39

C · THE FIRST DRAFT

Write Badly First. Fix It Later.

“It saves time to scribble … whole pages as quickly as I possibly can … Sentences thus scribbled down are often better ones than I could have written deliberately.”

— Charles Darwin, Autobiography (quoted by Zobel)

STUCK? USE SHORT SENTENCES

“Large files must be sorted in blocks. Each block is sorted in memory. Sorted blocks are merged.”

WHY IT WORKS

Not elegant — but the ideas are on the page. Good writing comes from revision.

Research Methodology · Week 9 · Zobel Ch. 5, The First Draft

31 of 39

C · ZOBEL’S PROCESS

From Notes to a Finished Paper

1

Brainstorm results

2

Choose section titles

3

20–200 words per section

4

Write the body

5

Revise the intro

6

Abstract last

Material that fits no section? Your structure is wrong — or the material doesn’t belong.

Then read it as a referee: answer objections, or admit them.

Research Methodology · Week 9 · Zobel Ch. 5, From Draft to Submission

32 of 39

ZOBEL

“The writing defines the research, and the one cannot proceed without the other.”

Reading is never complete. Start writing this week — with your data, not after it.

33 of 39

C · YOUR MEMORY

Keep a Research Log From Today

Data sources + dates

Decisions + why

Experiments + outcomes

Code versions (Git)

Papers you read

Ideas + expectations

Your Methods section is written from this log. Your AI use log goes here too.

Research Methodology · Week 9 · Zobel Ch. 5, From Draft to Submission

34 of 39

C · GETTING IT WRONG

Five Ways a Reviewer Rejects You in Minutes

Problem

What the reviewer sees

Irrelevance

No clear question; results unrelated to the conclusion; wrong venue

Inadequacy

Experiments too small or artificial for the claims; missing literature

Incomprehensibility

Sentences the reader cannot follow, from the first line

Ugliness

Messy graphs and tables, font changes, many small errors

Ignorance

Pages on basic ideas; old or wrong facts; key papers not cited

Zobel’s most common reason for rejecting good papers: experiments too weak for the claims.

Research Methodology · Week 9 · after Zobel Ch. 5, Getting It Wrong

35 of 39

MINI ACTIVITY 4 · 6 MIN · PAIRS

One-Minute Reviewer: Which Problem?

  1. “With the explosioning of documents on the internet, systems of finding documents that are the answers of users their queries have become important.”
  2. A paper about Web search tests its method on 2,000 documents.
  3. The experiments measure string-search speed; the conclusion says the method protects data privacy.
  4. Six pages explain the difference between RAM and a hard disk, citing 1980s sources.
  5. The graph points are drawn as small cartoon ducks.

Then: which one is the biggest risk for your paper?

Research Methodology · Week 9 · examples adapted from Zobel Ch. 5

36 of 39

C · BEFORE YOU SUBMIT

The Final Check

I can state my question in one sentence.

Every figure is mentioned and explained in the text.

I have real data, and I say where it came from.

I state my limitations honestly.

I interpreted my results, not just showed them.

Every reference is checked against the original.

Someone could reproduce my work from Methods.

I followed the template and disclosed AI use.

Research Methodology · Week 9 · course notes; Zobel’s “Writing Up” checklist

37 of 39

AFTER THE BREAK

Your Final Paper:

requirements & calendar

Keep your Figure 1 sketch and five-sentence abstract.

You will use them in the next 30 minutes.

38 of 39

REFERENCES

Sources for Today

  1. J. Zobel, Writing for Computer Science, 3rd ed. London: Springer, 2015 — Ch. 5, Writing a Paper.
  2. Course notes: “Writing Research” and “Writing a Paper (IMRaD Structure)”, courses.aaron.kr.
  3. IEEE A4 conference paper template (Causal Productions, V2), adapted for this course.
  4. C. Darwin, Autobiography — quoted in Zobel, Ch. 5.

Example titles, sentences, numbers and charts in this deck are invented for teaching.

Research Methodology · Week 9 · Writing a Paper

39 of 39

THANK YOU

Does anyone have any questions?

courses.aaron.kr

Please keep this slide for attribution

CREDITS: This presentation template was created by Slidesgo, including icons by Flaticon, infographics & images by Freepik