Conference talk crash coursePosted: July 25, 2012
If you got a paper published in a research conference, someone needs to give an oral presentation, conventionally with digital slides, at the conference. This can be intimidating, but remember that this is a good problem to have. Your work went through a rigorous and selective process, and was chosen to be one of the few topics showcased at the conference. Some of the most esteemed members of the international research community will set aside time to hear your ideas. Exciting!
Conferences typically operate on a compressed time frame that only gives presenters a few short weeks to prepare their presentation. Generally, in a student-faculty collaboration, the student gives the talk, since they are in greater need of notoriety and public speaking experience. So, it’s not uncommon for a student to need to learn how to give conference talks, and develop their presentation, in a very limited amount of time.
Here is my crash course on how to do that:
- Read this page in its entirety.
- If you haven’t been to a CS research conference before, watch at least 2-3 videos of academic research talks. You can find many on the FOCS website. The goal of this exercise is to observe the tone and pacing of conference talks, and reflect on what seems to work, or not work. Keep the advice in (1) in mind, and critique the extent to which the talk was clear, interesting, and received well.
- Pick a presentation software package and write a rough draft of your slides. I prefer using LaTeX to create slides, since that lets me copy math notation and figures directly from the paper manuscript. The most popular LaTeX packages for slides seem to be Beamer, Prosper, and seminar. You could also use a GUI program such as PowerPoint, OpenOffice Impress, or Apple Keynote, but you may need to reformat your math and figures, which has the potential to become a time-wasting distraction. This decision is not crucially important, so don’t spend too long on it. Whatever you use, generate a PDF file, since that is more likely than a proprietary format such as PPT to be seamlessly compatible with the podium computer.
- Follow the Generic Conference Talk Outline in (1). That outline is a conservative, uncontroversial structure that will work fine. It’s a “nobody ever got fired for buying IBM” sort of deal. In my opinion it’s a bit cliché and tired, but your first talk is not the place for avant garde experimentation.
- Check the conference website to see how long your talk should be. 20 minutes is common. Budget about 3 minutes for audience questions and interjections, and 1 minute for the logistics of transitioning to the next speaker, leaving about 16 minutes for you to present content.
- Based on that time limit, and the heuristic of 2 minutes/slide, you will be limited to about 8-10 content slides, not counting front matter such as title, outline, or acknowledgement slides.
- Iterate. Get feedback on your slides from your coauthors, adviser, or peers. Conference talks are expected to be polished, and to achieve that you’ll need to go through several rounds of revisions.
- Practice. Go through the entire talk, speaking out loud, at least twice. Pay attention to your running time, pacing, and areas where you tend to stumble. Speakers almost always run longer than they expect; speech is slower than internal dialogue. Ideally, you will have a chance to practice at least once in private and once in front of a critical audience.
- Try to have fun.