Before starting and now during my PhD I have looked up for information on how to read papers efficiently. I have already read many resources, tried different approaches, and read a lot of papers, which makes me think I am getting better at it.

I created this post with two goals in mind:

  • To have a summary of the main points I have learned so far on ‘how to read papers efficiently’
  • To have a list of resources (references at the end) on ‘how to read papers efficiently’

The summary is probably more helpful to me than to anybody else, but I think you can get advantage mainly of the list of resources (references) if you are interested in improving your efficiency on reading papers.

How to read a paper

There are many videos and resources talking about this but I would summarize the method in these steps:

  1. Read title, abstract, pass quickly through the figures and subtitles and read the conclusion. Check if you are still interested to continue to read.
  2. Read the rest of the paper in the order most interest you (introduction for previous work, methods for techniques used, results for results ^_^). Check if you want to continue reading or this level of understanding is enough for now.
  3. Read again the parts that you want to understand better (maybe it is the math, maybe the algorithms, etc; now you can work out the math or try to implement the algorithms to improve your understanding). Check again if you want to continue reading.

The previous steps are very flexible, but the most important thing to keep asking yourself while reading is “How will this paper help me in my research?”.

For better advice than mine, please look at [4] and [5].

Time to spend reading papers daily

According to [2], there are people who spend 3 hours weekly on reading, some others 3 hours daily.

Resources from where to get interesting papers

  • Journal table of contents, Google Scholar Alerts, Journal Clubs, Journal Scan Club
  • Reddit, Twitter, Facebook, Discord,

Information overload

  • Google Scholar Alerts is a great tool, but must be used carefully to avoid information overflow [3]

References

  1. How many papers do you read per week and what is your position in the lab?
  2. Scientific literature: Information overload
  3. How to keep up to date with the literature but avoid information overload
  4. How to Read a Paper Efficiently (By Prof. Pete Carr)
  5. Stanford CS230: Deep Learning Autumn 2018 Lecture 8 Career Advice Reading Research Papers