Today I watched a video by Prof. Peter Carr of University of Minnesota on how to read research papers. Here are the notes I took.
Most of the papers are organized like this:
- Results and Discussion
For computer science papers, especially systems papers, the layout can be a little different. After the introduction section there is usually a section that explains the design of the system. The experimental and results section are usually merged into one.
But regardless of the format of a paper, the last thing you want to do is to read it from start to end. Do NOT read a paper linearly.
Prof. Carr divided his paper reading method into two phases.
Phase 1: survey the paper
This phase should takes 5 to 30 minutes.
- read the title and keywords
- read the abstract
- read the conclusion
Phase 2: read the article
This phase might take between 1 hour to a few days
- look at the tables and the figures, including the captions
- read the introduction (for some background)
- read the results discussion (this is the heart of the paper)
- read the experimental/design (learn how the authors did the work)
At any point of the read process, if you feel that you have lost interest, feel free to stop reading. After reading the paper, it is a good idea to write down some notes (preferably in a notebook instead of in the margins) so you do not have to read the paper again.