How can I read paper efficiently

Fish and sheep gather and organize
Qubit Report | Public account QbitAI

The beginning of salting lies in reading newspapers.

On the one hand, capture the latest research trends and stimulate your own research inspiration. However, if you don't do a good job in literature research, your brilliant ideas for reinventing wheels will be. This experience is not very good.

So here comes the question:How should I read a newspaper?

Hundreds of new articles can be published every day from arxiv alone. During the summit, the articles came from the bank like the Qiantang River flood. How can we not get lost in the smoke of papers and receive effective information efficiently?

Here are essay reading skills from Srinivasan Keshav, Professor of Computer Science at Cambridge University, and ACM and IEEE FellowTriple theory, Teach you how to read papers efficiently, and say goodbye to the dilemma of using massive amounts of energy, but having little impact.

Salt loving is a new must-have, and salt-loving veterans can relate to it too.

First time: quick preview, capture the summary

How long does it take to read a new newspaper for the first time?

5-10 minutessufficient.

Not all paper is full of dry goods, so it's more efficient to make an impression the first time you meet before deciding whether to continue.

The specific operations are as follows:

1. Read the title, abstract, and introduction carefully.

2. Ignore the content first and read each subtitle in the article.

3. If there is math content, search it first to determine its theoretical basis.

4. Read the conclusion.

5. Search the references and review them if you have already read them.

After reading it for the first time, you need to ask yourself the following questions:

1、classification: What type of paper is it?

2、background: What other papers does it have to do with? On what theoretical basis should the problem be analyzed?

3、accuracy: Does the paper's hypothesis look correct?

4、Contribution: What is the main contribution of the paper?

5、clarity: Is the paper well written?

With these answers in mind, you will know whether or not you should read this paper carefully.

PS: Here is also a little technique for writing a paper:The structure is as clear as possible and the main points are as important as possibleSo that the reviewer can get it the first time.

Second time: grasp the most important points and briefly ignore the details

If you decide an article is worth reading, you can queue it up for the second reading.

For the second reading you need to look carefully at the contents of the paper. The time invested is approximate1 hourOver.

However, don't dwell on terms you haven't seen before and don't get obsessed with the details of the deduction of evidence: write them down and skip them first.

There are two tips in this reading:

1. Look closely at the diagrams in the paper. Pay attention to the details: are the axes correctly marked? Is the conclusion statistically significant? Often in the details you can see the difference between really good work and hydrology.

2. Highlight the references in the paper that you haven't read, and then continue reading.

After the second reading you should be able to grasp the contents of the paper and summarize the main points of the full text.

However, sometimes you can't read the paper this way even after reading it: the paper may have a field you are not familiar with and there are too many unfamiliar terms. The author may have used evidence or experimental techniques that warrant it You did not know, also this article The paper may not be well written.

So let's go to the last step.

The third pass: revise the paper, paying attention to details

To fully understand the paper you need to start a third reading: follow the author's ideas and reproduce the contents of the paper in your head.

By comparing the reproduced results with the actual paper, you can easily see the innovative points of the paper, find the implicit hypotheses in the text, and grasp the potential problems and citations hidden in the experiment and technical analysis.

The third time, the most important thing is highlighted three times:detaildetaildetail

Find out each hypothesis in the author's statement, challenge it yourself, and put forward your own thinking. This will give you a deeper understanding of the evidence of the thesis and the technology involved.

One more thing: How do I research literature?

When reading newspapers, do you think of the fear of being dominated by literature research?

Professor Srinivasan Keshav also has a "three step method" to teaching you.

First, use academic search engines (like Google Scholar) to find 3-5 articles with the latest citations in related fields.

Read the relevant working part to understand how these documents work. If you are lucky, this content can directly help you complete the literature review.

The second step is to find the articles cited or repeated author names in the references of those articles.

Visit the websites of these metrics, check out recent publications, and see which top conferences they've attended.

The third step is to visit the major conference websites and browse their recent conference recordings.

Quickly identify high quality papers by first reading the "triple theory".

Summarizing the articles found in this step and the frequently cited articles in the second step essentially forms the first edition of your literature review.

Finally, the three-step process can be carried out iteratively.

I wish you all the best with salting ~