How to write a good paper

These days, I've been working on an ASPLOS paper assisted Olivier.

When I was an undergraduate student, I was shocked that some of the paper on top conference were written by Chinese, since most Chinese students major in sciences are not good at English. Then how can they complete the ''master piece''?

Some of the following notes may help reveal the secrets:

1. Note regularly: either writing report about your investigation to your superadvisor or keeping regular notes about the program with references. This will definitely save a lot of repeated efforts when you begin writing.

2. Good management of time: Olivier emphasized a lot of times that writing a good paper is an issue of managing your time efficiently. Start early of writing since there is no way completing a good paper in the last week. In some cases, you may get stuck in the experiment issue. So just keep writing as a routine when you set up your mind of start the project. And then strike a balance between experiment and writing issues when you are in different stages of the project.

3. Mastering the related tools: such as Latex for writing, Inkscape for drawing figures, GNUplot for performance figures, Bibtex for adding references. Google and Wiki are always the best teacher, but mastering the tools before writing will absolutely save you a lot of the debugging time.

4. Review your paper in reviewers or your audiences'shoes from time to time. You're the specialist in the content, but they are not. So take it for grant that they know the backgrounds at the same level as you. Otherwise, there is no meaning writing the paper. In short, see things in a layman's eyes but do them in a professional way.

5. Extended reading. In related fields, not just focus on your own areas, then it will be easier for you when it calls for some application background knowledge.

6. Compromise tradeoff between preciseness and clearness. It's so hard to achieve both. So in most cases, you have to make tradeoff decisions.

7. Learn to present your interesting experiment results. Make sure that they do make sense and doesn't looks trivial.

8. Format for the references:

-- Name of the authors: first initial of the surname with last name

-- Use the abbr. of a conference to save space

-- page nbs for all papers

-- month and year of publications

-- use the conference reference, i.e., use "PLDI" or "ASPLOS" instead of "ACM SIGPLAN Notices"

Maybe there are some more key points, but I just cannot come up with them. TO be added.

Comments welcome!
本文出自 “梦的脚印” 博客，请务必保留此出处http://daisy8867.blog.51cto.com/1043582/619882