🎙️ Podcast Link 🎙️
Writing good conference papers in fields like robotics takes planning ahead of time, but there’s also lots you can do close to the final submission deadline.
In this new video, I provide a range of specific tips focused primarily on the final paper writing stages when you have a complete or near-complete draft paper, and are refining and tweaking it for submission.
I cover concepts ranging from the value of clear signposting for the reader, to appropriate use of specific types of technical words, to compelling presentation of results, to minimizing cognitive load on a reviewer and more…
Please 🙏 if you think your colleagues, fellow students, family or friends may find this useful!
Complete list of 🕒 timestamps
🕒 (0:00) Introduction
🕒 (0:40) A Useful Reviewer Stereotype
🕒 (0:57) Don’t Try to Satisfy Unreasonable Reviewers
🕒 (1:17) Overview of the Video
🕒 (1:55) Control the Narrative Around Your Contributions
🕒 (2:42) Make Specific Claims About Benefits
🕒 (3:33) Explain Why Improvements Are Relevant
🕒 (3:58) Make Sure Motivating Examples Are Relevant
🕒 (5:25) Make Results as Directly Comparable As Possible
🕒 (6:29) Give Competing Techniques an Advantage
🕒 (7:08) Apples to Apples Comparisons Where Possible
🕒 (8:19) Minimize Unnecesary Distractions
🕒 (9:05) Clear Signposting is Key
🕒 (10:07) Clear Structure and Flow
🕒 (10:36) Minimize Unnecessary Cognitive Load
🕒 (11:28) Careful Usage of Words with Strong Meanings
🕒 (12:03) Sparing Usage of Descriptive Words
🕒 (12:29) Specific Performance Claims
🕒 (12:43) Don’t Let Grammar and Typos Get in the Way
🕒 (13:27) Avoid Rambling Discussion About Results
🕒 (14:18) Future Work Shouldn’t Be Trivial
🕒 (15:02) Single Message Paragraphs
🕒 (15:31) Sentence by Sentence Sanity Check
🕒 (15:50) A Paper That is Too Easy to Understand is Great!
Full Video Notes
- Intro: Reference the general paper writing video, this is more easy specifics
- General principle: treat the reviewer like a tired, grumpy, inattentive, but not fundamentally nasty reader. You’re writing for the reviewer at this stage, not the general audience. Continually
- Explain what it is you’re doing
- Explain why it matters
- Throughout: Explain why you’re presenting each bit of the paper
- Not knowing your contributions and their relative weighting, and controlling the narrative
- Explicit differentiation from past work in introduction
- Inspecific claims e.g. faster / compute
- Unjustified / inappropriate motivation (bio-inspired, AVs but not really)
- Clear Signposting: not clearly signposting what is included for completeness and what is novel
- Unclear paper structure: structural relationships, flow
- Letting the reader get lost
- Unnecessary Distractions: e.g. old model, obviously flaws with other implementations and results
- Inappropriate benchmarks and comparisons (reimplemented instead)
- Indeterminate results tradeoffs
- Waffling results sections
- Tempting Future work: that should have been done already
- Figure captions without takeaway messages.
- Unnecessary cognitive load on the reader – “this”
- Inappropriate superlatives, adjectives and adverbs
- General grammar and styling: grammar, Style, typos
- Multi-themed paragraphs
General principle: you want to be erring on the side of making things too easy to understand and appreciate.