How to read a paper?

Time:2022-8-9

How to Read a Paper?

Author: S.Keshav

Paper link:https://dl.acm.org/doi/pdf/10.1145/1273445.1273458

Link to this article:https://www.cnblogs.com/zihaojun/p/15713151.html


1. Introduction

Reading papers is an important job for scientific researchers. A scientific researcher may spend hundreds of hours reading papers a year, but few people teach how to read papers. Students often waste a lot of time when they learn to read papers. This paper presents the author'sthree times reading, this method enables readers to have an overall grasp of the article before reading the details of the paper, and can adjust the time spent reading the paper according to their own needs.

2. Three times reading method

When reading a paper, you should not just read it word by word from beginning to end, but you should read it three times.

  • For the first time, have a general understanding of the article
  • Grab the main content of the article a second time and skip the details of the article
  • The third time to understand the article in depth

2.1 First time reading

For the first time, take 5-10 minutes and complete the following steps:

  1. Read the article title, abstract, introduction carefully
  2. Read article chapter headings, ignoring other content
  3. Take a look at the math content to see what relevant theoretical foundations are used (whether you can understand it)
  4. Read the conclusion
  5. Take a look at the references, which ones have already been read

After the first reading, you should be able to answer the following 5Cs:

  1. Category: What type of article is this article? Overview or Method? Partial theoretical proof or experimental verification?
  2. Context: What field is this article about? What theoretical basis was used?
  3. Correctness: Are the article's assumptions reasonable?
  4. Contributions: What is the main contribution of the article?
  5. Clarity: How was the article written?

After reading it for the first time, you should have a general understanding of the article, so as to decide whether to read further.

  • It is possible that after reading it once, you will find that you are not interested;
  • Or don't know enough about the field;
  • It is also possible that the article is not well written, the assumptions are unreasonable, etc.

After mastering this method, when writing an article, you can also consider it from the perspective of the first reading – maybe the reviewer will only read it once and decide to reject you. Most of the people who read your paper may Just read it once and throw it aside.

  • Therefore, when writing an article, you need to carefully consider the parts involved in the first reading, so that readers can understand what the article is about within five minutes.
  • A diagram summarizing the article's approach might be useful.

2.2 Second reading

On the second reading, it may also take an hour. This time, you need to grasp the main content of the article, skip the details such as the proof of the article, and quickly understand the logic, arguments, arguments and conclusions of the article.

  • Pay attention to whether there are any errors in the charts of the article. If there are any errors, this article may be rushed out in a hurry, and the quality is questionable.
  • Mark articles of interest that you haven't read in the bibliography for later reading.

Sometimes I may not understand the main content of the article after reading it for the second time. It doesn’t matter, it may be because I don’t understand the relevant direction, I didn’t understand the mathematical proof, the expression of the article is confusing, or it is because it is late at night and I am too tired.

When you don't understand, you have three options:

  • Put it aside and pray that you don't need to read this article to be successful.

  • After a period of time, if you know more about the relevant direction and background, and then come back to read it, you may be able to understand it.

  • Regroup and read it a third time!

2.3 The third reading

To fully understand a paper, you need to read it a third time, especially when reviewing. The core of the third reading isRecreate this work in your mind: Based on the author's assumptions, how would you do this work?

On the third reading, you need to:

  • Pay attention to every detail of the article
  • Consider whether each assumption is reasonable
  • Think, if you were to write, how would you express these ideas?
    • Contrast with the author's method and draw on the author's strengths
  • Make an effort to come up with and document more innovations in preparation for future work.

The third reading can take many hours, even a seasoned researcher may take an hour or two.

After the third reading, you should be able to mentally reconstruct the entire article, pointing out the strengths and weaknesses of the article.

3. How to do literature research?

When you need to investigate a relatively unfamiliar field, how to use the three-pass reading method mentioned above? How to start, what papers to read?

  • First, use some keywords to search for 3-5 papers that seem to be the most relevant, read the first time, and see if you can find a review paper in this field.
  • If you can find it, congratulations, you can study the review and read the literature cited by the review.
  • If you can't find it, look for the articles or authors that are commonly cited by these articles, and visit the author's personal homepage to see if there is any new work.
  • Go to the top conferences or top journals in the field to find relevant articles, and read these articles with a second pass method.
  • If these articles all mention some articles that have not been collected, you can collect them until you are satisfied.

some thoughts

When entering a field at the beginning, don’t be quick, try to understand the method of the article thoroughly, and actively look for related papers. After laying a solid foundation and building a knowledge system, reading papers will naturally become faster and faster. After all, everything is difficult at the beginning!

The most important thing is to persevere. Reading a dissertation is a long-term learning process. Be prepared for long-term battles, and you will gain a lot in the process.

Come on, come on↖(^ω^)↗

Stanford boss Professor Wu Enda also specifically talked about the methods of reading papers and doing literature research in CS230:https://www.bilibili.com/video/BV195411Y7ms?from=search&seid=2053463791157624953&spm_id_from=333.337.0.0

It's quite similar to the method in this article (I like Mr. Wu Enda's accent and teaching style very much)


Author: Zhao Zihao University of Science and Technology of ChinaLDS LaboratoriesMaster's student

You are welcome to criticize and correct in the comment area!