Reading and Reviewing Papers › Module 2: Reading a paper
The parts of a paper
Lesson 1 of 3 · 2 min
Most research papers have the same parts. Knowing what each one is for tells you where to look.
| Part | What it tells you |
|---|---|
| Title and abstract | The question, the method and the main result, in short |
| Introduction | Why the problem matters, what is missing, and what this paper adds (often a list of contributions) |
| Related work | Earlier work and how this paper differs |
| Method | Exactly what was done: data, model, setup |
| Experiments / Results | What happened, usually in tables and figures |
| Discussion | What the results mean, and why they might have come out this way |
| Limitations | Where the results may not hold |
| Conclusion | The main points again, and future work |
| References | Every source used |
| Appendix | Extra details, proofs, more results |
- Title and abstract
- What it tells you: The question, the method and the main result, in short
- Introduction
- What it tells you: Why the problem matters, what is missing, and what this paper adds (often a list of contributions)
- Related work
- What it tells you: Earlier work and how this paper differs
- Method
- What it tells you: Exactly what was done: data, model, setup
- Experiments / Results
- What it tells you: What happened, usually in tables and figures
- Discussion
- What it tells you: What the results mean, and why they might have come out this way
- Limitations
- What it tells you: Where the results may not hold
- Conclusion
- What it tells you: The main points again, and future work
- References
- What it tells you: Every source used
- Appendix
- What it tells you: Extra details, proofs, more results
Many fields call the core structure IMRaD: Introduction, Methods, Results and Discussion.
Where the important parts hide
- The contributions list at the end of the introduction is often the clearest summary of the paper.
- The limitations section is where honest authors tell you what their result does not show. Read it every time.
- In machine learning papers, important details, such as how the hyperparameters were chosen, are often in the appendix.
You do not read a paper from start to end
Unlike a novel, you jump around. You read the abstract, then the figures, then the conclusion, then decide whether the rest is worth your time. The next lesson gives you a method for this.