How to present a research paper
Table of Contents
Note from Simon Peter's Data Center class.
Why is talking research useful?
- order your thoughts.
- communication to other researchers.
- gather feedback.
- establish relationships.
- build a career (eventually).
What makes a good research talk?
- centered around the audience.
- provide intuition.
- make audience want to read the paper.
- should not
- present every detail.
- show off how smart you are.
Anatomy of a talk
- motivation (20%)
- key ideas (70%-80%)
- evaluation (0-10%)
- Do not present results without an explanation.
Beginning of your talk
- Two minutes before audience coming in.
- use them!
- present an abstract of your talk
- problem/motivation
- approach/idea
- experiments/results
- broader meaning/impact
- Answer these questions in mind:
- what is the problem?
- why is this talk interesting, why should I listen?
Communicating the key idea
- pick a goal for your talk (like how you organize your paper).
- have the keyword in your head
- organize the whole talk around these points.
- no more than three (better: one).
- be very explicit.
- repeat and repeat and repeat
- Do NOT be shallow, be deep
- avoid overviews
- do not ramble
- go to meat quickly
Examples
- Example is the main weapon in a presentation.
- Make your own examples and avoid the papers' examples.
- Have motivating example at the beginning.
- Illustrate the idea in action.
- Images say more than 1000 words. (should be 16x16 words, XD)
What to omit
- Do not present outlines, it's a waste of time.
- Do not present excessive related work.
- but mention it, or have backup slides.
- Do not present too many technicalities.
- Audience won't follow anyway.
- Put details in backup slides in case someones asks.
- Do not exaggerate with animations.
- Animations are good but hard to follow.
- Do not clutter your slides with graphics.
How to present
- Be self-confident.
- Make eye contact.
- Watch audience.
- Finish on time.
Misc
- standard staff
- avoid errors
- face the audience
- make jokes but only related ones.
- check your laptop and avoid wasting time.
- Practice
- At least three times before you present it publicly.
- Add a summary page at the end. (Simon's feedback on my presentation on Sip-ML).