Some thoughts on constructing your papers

Make sure you are very clear on what you are doing. What task, data, algorithm, experiment, and related work will be used? Decide this as quickly as possible, so you know what your "story" will be! Get feedback from me as often as you can.

In the paper itself, you have only 4 pages to make your whole sales pitch. Use the space intelligently!

• Keep the abstract short. Make sure you make it absolutely clear what your contribution is. That is, what is it that you have to say in the paper that other researchers will want to know? They won't want to know with you did, they will want to know what you did for them. Don't try to over-inflate its importance or hide it in the paper. Say it up front and give the reviewers an opportunity to evaluate it.
• Describe the problem you are solving, its significance, and the current state of knowledge about this problem. This and the abstract and title should be about one page.
• Describe your solution to the problem and any variations you hope to use in experimental comparisons. Throw in a diagram if you can. This should take about one page also.
• Lay out your results, including any figures or graphics that help clarify the results. This should also take about one page, maybe one and a half.
• Conclude, discussing the significance of your results and making suggestions about what your results imply for future research in the area. With the references, this should get you to the end of the four pages.
• For examples of short, good papers, consider reading papers from the NIPS proceedings, which used to be 7 pages (single column), I believe. There's an art to it. Don't try to say too much, but make a small, significant, justifiable point. Then, present the evidence to back it up. Say what you need to do to make the skeptical reader excited about what you're doing.