Lessons from first submission - (1) Writing
Posted on :: First Submission Research Skills Academic Writing AI for Writing

From Text Dump to Paper: My Writing Transformation Journey

I recently submitted my very first paper as a first author, and the journey taught me a lot. I want to share the lessons I've learnt and archive my experiences through a series of posts, starting with this one.

Among the many hurdles, writing stood out as a significant challenge. When I first wrote my draft, I found myself staring at a 40-page "text dump"—not even a complete draft, missing experiments and related work. I had to condense it into a strict 25-page limit paper.

Guided by my advisor Jeehoon and colleagues (Jaewoo, Minseong, Janggun, and Seungmin), I went through a complete overhaul, learning to consolidate disparate ideas into dense, organically flowing sections.

In this post, I'll share the principles and tools that helped me navigate this challenging process, from fundamental writing advice to the surprising utility of large language models, all forged in the crucible of a tight deadline and an overflowing draft.

General Principles of Academic Writing

Jeehoon provided me with general principles of academic writing, which I diligently applied and found to be a constructive guide for my writing.

  • Structure your writing rigidly—even to the point where it might feel a bit stiff—to make the text predictable and followable for readers.
  • Always state the conclusion (or summary) first. This applies at any hierarchy within the text, whether it's at the beginning of a paragraph or the opening of a (sub)section.
  • Avoid writing a "conclusion" at the end of sections. The main point should already be established in the initial summary.
  • Present examples before generalizations to help readers grasp high-level ideas, then introduce the generalized theory afterward.
  • Minimize redundancy
    • For me, especially the introduction, background, and related work sections often contain similar content. The introduction should only provide high-level information to establish context, identify the gap, and state the innovation. The background should offer only the minimum information necessary to understand subsequent sections. The related work section should focus on clear comparisons with existing work in the domain that hasn't been discussed elsewhere.
  • Write concisely and without embellishment. Defensiveness or self-congratulatory language is counterproductive. And objectively describe your work, clearly comparing it to existing work and acknowledging any areas where it falls short.

Derek Dreyer's Advice on Talks and Papers

Revisiting How to write papers and give talks that people can follow, PLMW@ICFP 2022 was incredibly helpful.

Leverage Large Language Models (LLMs)

Finally, utilize an LLM as your writing assistant and your fastest feedback tool. Incorporating the principles above into my prompts and always getting feedback from the LLM first ensured they were enforced. This way, I could produce a much more polished version of my writing before presenting it to my colleagues.

Table of Contents