[ Home | Lab | New Students | Courses | Research | Publications | Activities ]

For a technical talk, you need to be able to get into sufficient detail that the audience can walk away knowing the subject moderately well. They won't become specialists, but they should know what technical challenges there are and how you are addressing them. If people who attended your talk can summarize your talk confidently and well, that is the best success you can hope for!

1. When you are trying to report on other people's work and it was not clear to you, say what you understood and what was not clear. Don't dwell on what was not clear unless you can point to what is missing from the other work. Saying something is not clear is always tricky, because the authors presumably think it is and might be offended. If you can raise a technical objection that is a lot better. To convert a remark about unclarity into a remark about technical inadequacy, try to give an example. Say what interpretations are possible in the paper that you are reviewing. If one of the interpretations is acceptable probe it further or find another example where it is not acceptable. In the end, if there is an interpretation that works, then maybe the shortcoming of the explanation of the paper rather than its basic idea.
2. Use Latex for any serious presentation though sometimes you may have to use Google Slides or Powerpoint because of what your collaborators work with.
3. Put your name and the date on each presentation. Number all pages so I can easily give comments on a draft and the audience can easily ask questions about specific slides.
4. Zero in quickly on the new stuff—that is, have a short introduction.
5. Give an outline. Ideally, the outline should be meaningful. For example, you might say "there are three main pieces and I will discuss them sequence", but not enumerate section headers "and in the end I will conclude."
6. Remove captions and move to frametitle, except perhaps in rare cases.
7. Avoid overly dense figures. If you can eliminate some detail, do.
8. Minimize mathematical symbols. For example, instead of defining X as "accuracy" and using X, you might as well use "accuracy" directly.
9. Leave some blank space at the bottom of your slides. Often, especially in large rooms such as for the major conferences, the bottom 10–15% of your slides won't be visible to the audience, especially those sitting at the back.
10. While you don't want too much text on your slides, place enough text that someone reading your slide can hope to guess at your intended meaning. There are two motivations for this. One, some people will read your slides, e.g., from handouts or off the Web. Two, sometimes the audio at a conference can be messed up and you don't have a chance getting through if your slides are not self-contained.
11. Use a pointer if you can. When you use a laser pointer, don't let it wander when you aren't using it. People pay a lot of attention to a pointer, even an erratic one or perhaps especially an erratic one.
12. As for papers, distinguish different kinds of terms with different type-faces.
13. Use sentence fragments in bullet items; only rarely are whole sentences OK.
14. When giving a talk, repeat any questions that are asked from the audience. Often, other attendees can't hear the question—and, you might be the only one with a mic. Plus, you should paraphrase the question to match more closely what you are going to answer.
15. Know reasonably standard pronunciations of the words that you are using. (I am beginning to assemble a list of tricky words.)
16. It is a good idea to skip over details when giving a presentation. This is one of the hardest things to carry out, though. Be careful not to let yourself be drawn into topics you decided you wouldn't care to present.
17. For a time-limited talk (aren't they all?) be aware of the time you have left. Pay attention to your conference session chair and request sufficient time warnings so you can be sure to include your key conclusions and leave time for questions.
18. Some people include a slide at the end that says "Questions?"—I find that quite rude. When you are addressing your peers or your graduate committee, you don't say "any questions?" to them. They will ask questions when they want to. Say nothing or say thanks.
19. For slides that are part of a series on a topic, use sequence numbers in the title, as in "Foo: 1" and so on. If you just say "Foo" it doesn't tell the audience that more is coming or where you are in the sequence. And, "...contd" and its variants are too long.