How to Give a Good Research Talk

Based on a presentation by:

• Simon Peyton Jones
  • Functional Programming God
  • Implementation of lazy languages
  • Used in Haskel classical compiler which he largely wrote
• John Hughes
  • Author of the famous paper: Why Functional Programming Matters
• John Launchbury
  • Creator of the Crypto language

Research is About Communication

Communication is done using talks and papers. It's a really good idea to give a talk before writing a paper to crystalize your ideas, get feedback, communicate, get credit for your research

• Give a talk about ANY IDEA
• Don't be shy about presenting your ideas
• Invest time, learn skills, practice
• The more you practice the better it is

What Your Talk Is For

• Your paper = The beef
• Your talk = The beef advertisment

Talk Is Not

• The talk is not to impress your audience
• You don't want to introduce your audience to your topic
• You don't want to present all details (implementation, proof, step by step)

Talk Is

• The talk is to give your audience an intuitive feel about your idea
• Push them to read your paper
• Get them excited

Audience

Don't assume that your audience is ideal (have read all your previous paper, is awake, understand all details about your field, excited to hear you) The actual audience is just the inverse: WAKE THEM UP

What to Put in Your Talk

• Motivation: 20%
• Your key idea: 80%

Motivation

You have 2 minutes or less to motivate audience before they sleep. They are thinking: Why should I care? what is problem? what is interesting about that problem?

Key Idea

• You must have ONE idea
• Make that explicit
• Arrange all the talk around this ONE idea
• Don't start a novel by writing "it was a dark night, ..." and see where this takes you, similarly don't start a presentation without being crystal clear about the ONE idea that you want to communicate
• The idea "What I did in the summer" is not good

What to Include

• Logical sequence
• Going through a table of contents is the most boring thing just like reading a dictionary. You must not spend all the talk giving an overview of your paper or your idea. You have to go deep and provide details. "Narrow, deep beats wide, shallow"
• The biggest tool to get deeper is using EXAMPLES. Examples will also help you measure the size of the information you are communicating. If necessary, omit the general case, but not the example.

What to Leave Out

• Outline
  • No Outline in the beginning
  • Maybe after the motivation, brief, maybe don't call it outline
• Related work
  • You must know it. Not knowing it is the best way to get your papers rejected. It should be addressed adequately between the introduction and the related work section of your paper.
  • Acknowledge co-author (title slide), and pre-cursors (as you go along)
  • Don't mention things without mentioning how it is related to what you have done.
  • Never say the work done by X is stupid or this previous solution is lame.
• Technical details
  • Omit technical details like all typing rules.
  • Have backup slides with all details is important
• Apology
  • Don't sound apologetic

Presenting Your Work

• Practice the night before
• When you become very good maybe you can write the slides the night before the talk
• Be enthusiastic about the talk
• Being afraid is normal: breath deeply, prepare the first few sentence
• Being seen and being heard are very important: Point at the screen, not your laptop, watch audience, be interactive
• Welcome Questions
• Don't reveal your points one by one
• Minimize animations usage
• Finish on time

Hope

The general standard is so low that you don't have to be outstanding to stand out

-- CherifAndraos - 30 Aug 2007