How to Read an Academic Paper?
It is the first thing everyone has to learn in graduate schools, or even before. And it is not so easy as it appears. I still remember when I read the first paper, I read word by word, sentence by sentence, only overwhelmed by terms and notations. By reading more and more papers, I learned how to read a paper efficiently. I realized that reading a lot of papers is not the only way to learn. There is a shortcut.
- Read the abstract.
A good abstract addresses three questions. What is the problem? What method does the authors propose to solve the problem? And what is the performance of the proposed method?
I always read an abstract with the three questions in mind. If you feel hard to read something with questions in mind, try to mark the sentences that answer each question (see the above figure as an example). By reading the abstract, you will know whether this paper is of your interest or relevant to your research. If not, you are done with this paper. And this saves much of your time.
2. Read the introduction.
The introduction is an extended abstract, addressing more questions in more details. Why is the problem important? What did previous researchers do to solve the problem (usually there is a “related work” section in computer science papers)? What is the proposed method in more details? How significant does the proposed method outperform previous methods? Why?
The introduction section is very important for reviewers to evaluate the paper. For readers in general, the introduction provides more information so that you can decide whether to look into every detail of the paper before getting overwhelmed by notations and equations.
3. Read the remaining parts
If you are interested in the paper, you will have to understand the notations, theorems if there are, and other details. Hope your interest will power you up to overcome these difficulties. If you are not in a graduate school yet, you may consider whether to apply for one now since this is the most fundamental requirement to do research.
