Pre-Paper Talks
The idea is to recommend that we do a paper planning exercise early (rather than later) to map out the path to a paper. The idea is to use a “pre-paper talk” to force us to write down the plan in enough detail, to have a plan in written form for discussion, and to use an informal talk as a way to iterate on the plan.
This is a work in progress, we are experimenting.
There are two things here:
- To organize your thoughts and plan in a “document” - but have this document be a slide deck, rather than another kind of document. I think this makes it easier to write it out, easier to use as a conversation tool, and easier to someone to check quickly.
- Provide an opportunity to talk through the plan in a structured way. Since goal #1 is so big, it’s kindof a backwards talk (start with the slides)
In this case, the slides are a separate goal - this is unnusual (usually, slides are there to support the speaker giving a presentation). The slides might even be the primary goal.
Since all papers are “Intro, Related Work, Methodology, …” an outline at that level shouldn’t be necessary.
You should write the pre-paper talk as if the work has been done, and you know what they results are. Assume the results are awesome (you succeeded at making the algorithm, the experiments turn out cleanly, …) - although, you should note in places what the contingencies are.
Slides you should have… These might seem redundant, but they ask you to look at the paper in different ways.
- Sentence - this isn’t necessarily the first slide to talk about, but have the slide with the sentence by itself - for when it comes up.
- Lede (this is a newspaper term) - the way you intend to lead the reader into the paper. It is motivation and setup. Kindof like the “first paragraph” (but not necessarily in paragraph form).
- Research Question - not all work can be framed this way - but what is the bigger question you are asking, and what is the more specific thing you are going to do, and why is that the right thing. What are you trying to achieve? This overlaps with many things. But I think explicitly writing it out can be useful.
- Accomplishments/Results - this is an outline of what you expect to have done (the way you describe the outcomes of the work - at the level of the intro). These should connect to the question (how do the results achieve what you were trying to achieve.)
- Overview of the work - what did you actually do (to achieve the results) - at a high level. The idea is to describe the “what you did” at the level you might put it in the intro - and to decide what of these details are important.
- Contributions - think of this as the “explicit contributions” paragraph of the intro, but in bullet point form. Another way to think about this: what do I expect the reader to learn.
- Related Work: Overview - how do you plan to organize related work?
- Related Work: Context - what is similar? how will you differentiate it?This is a few key things. Often these are the ones that will make it into the intro.
- Methods or Technical Description - depending on the paper type, this is the “technical meat” of the paper. Outline what you expect to have.
- Results - what results do you intend to show, and how do you intend to show them. (this is an overlap with #4 - but it is more about the details of what you will present and how will you present it)
- What could go wrong? - what are the biggest risks you are aware of? (of course, the biggest risks you aren’t aware of are scarier)
- What critiques do you think reviewers might have? - predict what “angry reviewer #2” will say.
- What are the key diagrams and figures? - thinking about “how will I tell the story visually to a reader who is skimming” is a good exercise. Also, identifying figures that might take a lot of effort and preparation is good.
- what is your plan/schedule? - spelling out milestones, specific outcomes on specific dates
- overview of the paper - this probably shouldn’t be last, but… the above are a list of pieces, it really doesn’t say what the “narrative flow” of the paper will be. I’m not sure how you capture this concisely as an outline (without it just becoming “Intro, Related Work, Methods, Results, Discussion”)
This is a case where bullet lists may be preferable to narrative prose. It should be quickly skimmable. Pictures are good. Even if it is sketches. “I expect to show a chart like this”.
Having these slides lets us switch between things quickly - it might not be the best for a “presentation” - in fact, you might want to have a separate “how I would tell the story as a talk”.
Relation to Sentence, Paragraph, Page
In Writing Advice: The Paper "Recipe" I recommend a formula: sentence, paragraph, page.
In a sense, the pre-paper talk slides are a different way to do the sentence paragraph page exercise. And the talk is a structure way to discuss.
An advdantage to “formally” writing things down is that it forces you to confront the wording issues.
- Put the sentence on a slide. I think it is such a valuable exercise that you should have a sentence, and remind yourself what it is. In the extreme, you put the sentence on every slide so you can assess how each thing connects to it. But for the pre-paper talk, we might not want to get hung up on the wording of the sentence (the pre-paper).
- We replace the paragraph with the “intro slides”. This doesn’t help expose the wording issues as directly. But you will find these when you start to write.
- We replace the page with the outline - there is a question of what the right level of outline is. The slides can deal with this explicitly.