Playbook: How to Read Research Papers

By Amna, TKS Boston

🔑
Purpose: To help you become more efficient when doing research.
Originally made during the TKS Global AI Hackathon 2020

Step 1: Exploration Phase ⛵

Start by identifying a general topic that you're interested in. Think about what sparks your curiosity.

You can start by looking through the following resources:

‣
Science Based
‣
Tech Based

Tip: Wikipedia and Golden are also a great source to get a broad understanding of a topic before you start to dig deep.

Step 2. Refinement Phase 🥽

When you have narrowed down a research area, you'll need to validate it by asking yourself: How feasible is this? Has this been done before? What are the latest developments in this field? These questions are often best answered by digging into research papers directly.

Why read papers?

“Be skeptical. But when you get proof, accept proof.” – Michael Specter

Research papers go through comprehensive review processes before getting published that check the validity of the methods, results and conclusions. With millions of contradicting sources on the internet reading papers is an easy way for you to make conclusions yourself based on what the data shows.

How to find research papers:

🔑
Pro-tip: Make sure that the research you are looking at is RECENT. Especially for AI, a LOT happens each year, so keep in mind how outdated the papers you are looking at are.

Start by looking up key words in Google Scholar.

‣
How to access papers that you dont have access to:
‣
Tips for searching on google scholar

How to EFFICIENTLY Read a Paper:

“There is an art of reading, as well as an art of thinking, and an art of writing.” - Clarence Day

Research papers are filled with technical jargon that can be intimidating. But with these tips you can read a paper and get its key takeaways in 10 minutes or less without any expertise in the area.

🔑
Pro-tip: Don't read a paper from start to finish. See my tips below.
‣
Papers are typically outlined as follows:

Here is the order in which you SHOULD read them instead:

  1. Start with the abstract. You can also skim over the headings in the rest of the paper. The overview will give you a sense of whether this paper is worth reading more.
  2. Read the discussion/conclusion. These are the biggest takeaways from the research, and may be all you need to know.
    • for some papers the discussion and conclusion may not make sense without the introduction, in that case go to the introduction and come back
  3. Go back to introduction if you are interested in digging deeper or the conclusion did not make sense. It will give you more background into the field of research, and justify the experiments performed in the paper.
  4. Read the methods and results if you are skeptical about the findings of the paper.
‣
Also, look over these:

Step 3: What Solutions Already Exist? 🛠️

If your research was done for the purpose of coming up with a solution to a problem, you need to determine what solutions already exist. Make sure that before you start to build anything you do your research on what already exists.

  • If you are building an app, you can search on apple's app store or google play.
  • You can also look for up and coming products that exist on Product Hunt

IF your solution already exists you can do 2 things:

  1. Refine the product that already exists by determining what its pitfalls are and where it can be improved.
  2. Brainstorm another solution.