How to Write a Great Research Paper - 2016
Details
Title : How to Write a Great Research Paper Author(s): Microsoft Research Link(s) : https://www.youtube.com/watch?v=VK51E3gHENc
Rough Notes
Great talk on the writing aspect of scientific communication. 7 simple, actionable suggestions:
Don't wait: write Just start writing. The plan should be:
- Develop idea (months to some years)
- Write paper
- Do research. The typical plan switches 2 and 3 but writing first forces us to be clear, focused. Earlier on you will catch concepts that you thought you knew but actually did not.
Writing it down also opens it up for critique, reality checks and collaboration.
- Identify the key idea Aim to convey a useful and re-usable idea, and aim to infect the reader with this idea. The notion that you need a fantastic idea before you can even write a paper is a fallacy. Do not be intimidated. We only discover ideas are good later on. Your paper should have one clear, sharp idea. Be 100% explicit on what the idea is, for e.g. "The main idea of this paper is…", "In this section we present the main contributions of the paper"
Tell a story Imagine explaining at a whiteboard:
- Here is a problem.
- It is an interesting problem.
- It is an unsolved problem.
- Here is my idea.
- My idea works (details, data).
- Here's how my idea compares to other people's approaches.
Common structure for conference papers: Title, Abstract (4 sentences), Introduction (1 page), The problem (1 page), My idea (2 pages), The details (5 pages), Related work (1-2 pages), Conclusions and Future Work (0.5 pages).
- Nail your contributions In the Introduction, describe the problem and your contributions. Do not write like "Computer programs have bugs. It is very important to eliminate these bugs [1,2]. Many researchers have tried [3,4,5,6]. It is very important". Instead write "Consider this program, which has an interesting bug. [Brief Introduction]. We will show an automatic technique for identifying and removing such bugs." Do not also let the reader guess the contributions, state them explicitly. No "Rest of this paper is structured as follows. Section 2 introduces the problem. Section 3… Finally, Section 8 concludes." Instead use forward references, e.g. "We give the syntax and semantics of some … (Section 3)." Contributions should be refutable (falsifiable). The body of the paper should have evidence for each claim - check each claim in the introduction, identify the evidence and forward reference it from the claim. The evidence can be analysis and comparisons, theorems, measurements and case studies.
- Related work comes later Put related work at the end, since the reader does not yet understand the problem and your highly compressed description of related work gets between the reader and your idea, making your idea to understand.
- Put your reader first Don't make them feel stupid, don't add sophistical technical detail to just say later on that does not matter. Conveying intuition is primary not secondary. Use examples.
- Listen to your readers Get your paper read by as many friendly guinea pigs as possible. Explain carefully what you want, that you want feedback like "I got lost here" is more important than "There is a typo here". Reviews are a free gift, be thankful for them.