Thursday, March 7, 2013

Writing Reviews

The following guide suggests a steps for writing a paper review. The structure should include the following
  • Summarize the paper
  • State the contributions (what they are , what was missed)
  • Novelity: Whats new , related work missed , related work invalidates the contribution
  • Make it constructive : "paper would be stronger if it does so and so"
  • Criticize the paper, not the work itself
  • Avoid flat assertions: "The description left me worried" rather than "not correct"
  • Its OK to be humorous  

Reference:

Writing reviews for systems conferences (March 2007) by Timothy Roscoe

How To Read A Research Paper

The following is a summary for a paper written by S. Keshav from University of Waterloo. It presents a methodology for reading research papers with the aim of learning, reviewing or writing a literature survey.


The paper discusses the three-pass method. The first-pass allows getting a general idea about the paper. The second-pass should allow grasping of the paper content. The third-pass allows understanding the paper in-depth.

The first-pass: (Estimated : 5 Minutes)
- Read title,abstract,introduction
- Read section and subsection headings
- Glance at the mathematical content
- Read conclusion
- Glance over references

The second-pass: (Estimated : 1 hr)
- Ignore details of proofs
- Understand figures,diagrams,graphs
- Mark relevant unread references for background information

The third-pass: (Estimated : 5 hr beginner, 1 hr experienced)
- Fully understand the paper
- Virtually be able to re-implement the paper
- Identify and challenge all assumptions
- Able to present the idea yourself
- Able to reconstruct the entire structure of the paper from memory

Reference: 
S. Keshav. 2007. How to read a paper. SIGCOMM Comput. Commun. Rev. 37, 3 (July 2007), 83-84.