How to Give a Good Talk
The speaker approaches the head of the room and sits down at
the table. (You can't see him/her through the heads in front of
you.) S/he begins to read from a paper, speaking in a soft
monotone. (You can hardly hear. Soon you're nodding off.)
Sentences are long, complex, and filled with jargon. The
speaker emphasizes complicated details. (You rapidly lose the
thread of the talk.) With five minutes left in the session, the
speaker suddenly looks at his/her watch. S/he announces -- in
apparent surprise -- that s/he'll have to omit the most
important points because time is running out. S/he shuffles
papers, becoming flustered and confused. (You do too, if you're
still awake.) S/he drones on. Fifteen minutes after the scheduled
end of the talk, the host reminds the speaker to finish for the
third time. The speaker trails off inconclusively and asks for
questions. (Thin, polite applause finally rouses you from
[Paul Edwards on “How to give a talk”]
What’s wrong with this picture?
No visual graphs
Large print busy slides
No moving about
Focus on main points
Lost in details
Finish on overtime
Respond to audience
What will I talk about today?
– Giving a good talk is important!
– Don’t annoy your audience!
– Where are you presenting?
– Who is your audience?
– What is your central message?
– How are your slides put together?
– How do you answer questions?
– oral exam or thesis defense
– conference talk
Giving a Good Talk:
It’s important! [Tamara G. Kolda, 2002]
More people will see your talks than will
read your papers
The audience will form their impressions
of you based on your talks
Early in career, treat every talk like an
Start as early as you can—no later than 1
year before your PhD
How might I annoy my
audience? [David Patterson, circa 1983]
not be neat (in good order)
not covet brevity (concise)
not write large
not use color
not make eye contact
not skip slides in a long talk
Know what your surroundings will be like:
size of room, microphone, equipment, etc.
Know your audience and tune your
message to that audience
Get to the point—early and often
Organize your slides so that they
effectively deliver your central message
Answer questions skillfully
Where are you presenting?
Know your surroundings
[Patrick Winston on “How to Speak”]
Look at the hall: you want the place to be
How many people will be there? Nature of talk
changes with size of audience!
– 20 people: discussion is possible
– 50 people: performance is expected
Find out who has talked before
Schedule talk for 11am
Mood: people reading newspapers will make you
feel bad! (Don’t allow it.)
Who is your audience?
Know your Audience
[Tamara G. Kolda, 2002]
One of the biggest mistakes speakers make is not
knowing their audience!
Will your audience include …
– Specialists in your sub-field? In your field?
– Researchers in the computer/mathematical sciences
– Engineers and scientists?
– Faculty and postdoctoral researchers? Graduate
Delivering the central message
[Tamara G. Kolda, 2002]
What did you do? Why is it important?
What’s the one-sentence summary of your talk that the
audience should walk away with?
Tune your message to your audience
– Symbolic knowledge improves statistical techniques for cross-
language topic detection and generation of short summaries to
represent foreign-language news articles
– Generation of English headlines for foreign news stories is
enhanced when our algorithms use linguistic knowledge
Repeat the message over and over again throughout the
Keep the content of the talk focused on the central
[Tamara G. Kolda, 2002]
Sample outline of a research talk:
Title slide: credit to co-authors and funding agencies
Up-front “carrot” (attention-getter)
Outline (unless 10-15 min talk)
What you did
– new algorithm, theorem, proof, computational paradigm
Why is it important
– numerical results
Summary and future work
Minimize background material
At least 2/3 of talk should be original work
Identify those who have done related work
and spell their names correctly!
Hint: People love to hear their own names.
Describe motivating applications that will
later tie into your results
What you did
Emphasize your simple message repeatedly
Back it up with details of algorithm and theory
Use pictures and diagrams as much as possible in
lieu of wordy explanation
Keep notation to a minimum and avoid too many
Never use equation numbers—repeat the
equation if necessary
Illustrate your points via simple examples
Like this …? present an algorithm?
How do you
Or like this …?
Finding the largest among five integers
Tables and Figures
– Don’t make font too small
– Use color for emphasis
– Be sure axes are clearly labeled
– Use color to differentiate lines
– Don’t just copy verbatim out of a conference
Why is it important?
Think BIG PICTURE!
Emphasize an application
What makes it a hard problem?
Why should people care?
Summary and Future Work
Repeat what you did
Repeat why it is important
Future work is important for recent PhDs
because it shows you are thinking beyond
your thesis problem
Include contact info at the end
– email, web page
Repeat the question.
Demonstrate knowledge of standard problem solving.
Draw a diagram.
Specify an analogy.
List the assumptions.
List the ideas and tools that seem relevant.
Respect the questioners and their questions
Inevitably, someone will tell you your work has already
been done by someone else!
Oral Exam or Thesis Defense
Observe and try to emulate excellent speakers
Ask in advance what examiners will ask!
Memorize a few key sentences.
Get there early and set up!
Cycle in on what you have done.
Try to convey a sense of quiet confidence.
Conference talk [Mark Hill, 1992]
Title/author/affiliation (1 slide)
Forecast (1 slide)
Outline (1 slide)
– Motivation and Problem Statement (1-2 slides)
– Related Work (0-1 slides)
– Methods (1 slide)
Results (4-6 slides)
Summary (1 slide)
Future Work (0-1 slides)
Academic Interview [Mark Hill, 1992]
Take a 20-minute conference talk.
Expand the 5 minute intro to 20 minutes
Do the rest of the conference talk, minus the
summary and future work.
Add 10 minutes of deeper stuff from your thesis.
Do the summary and future work from the
conference talk in a manner accessible to all.
Add 10 ten minutes to survey all the other stuff
you have done (to show your breadth).
Save 5 minutes for questions (to show that you
Know your audience
Create a simple message and repeat it several
Allow plenty of time to prepare your talk
Don’t block the slides during the talk
Speak slowly, clearly
Don’t run over on time
Have fun and learn from your mistakes
Thanks goes to …