Eacl writing style
Skill brycewang-stanford/Awesome-Journal-Skills/EACL-Skills/skills/eacl-writing-style
Journal-specific Claude Code/Codex skill packs covering mainstream journals — AER, QJE, Nature, Cell, 管理世界, 经济研究 & 200+ more — your fast track to getting published. ｜ 覆盖主流期刊的 Claude Code/Codex 期刊技能包，从选题、识别策略到表格规范与审稿回复全流程，助你快速发论文。
npx -y skills add brycewang-stanford/Awesome-Journal-Skills --skill eacl-writing-styleAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
What its author says it does
Copied from the file, not written here
Use when revising an EACL paper's prose for a task-first first page, a concrete page-one example, scoped LLM-era claims, quantified error analysis, an anonymity-safe voice, honest Limitations, and compression into the 8-page long or 4-page short content budget without pushing the argument into appendices.
SKILL.md
3.4 KB, as published. Nobody here has run it
EACL Writing Style
Use this to revise an EACL paper so its contribution is legible fast and its claims are scoped.
EACL reviewers read many papers in a short window; the ones that land put the task and the
result on the first page, quantify rather than assert, and name their limits. Pair this with
the worked example in ../../resources/worked-examples/01-introduction.md.
The EACL first-page arc
- Task — the specific problem, in the first breath, not "great progress in NLP."
- Gap — why current methods fall short, each reason nameable.
- What we do — the contribution, stated plainly.
- Measured result — a number tied to a table, with variance.
- Honest scope — what the result does and does not cover.
Habits to cut, habits to keep
| Cut | Keep |
|---|---|
| "Achieves strong performance" | "Improves F1 by X (95% CI ...) over baseline B" |
| Generic "prior work is limited" | A specific failure per cited approach |
| A concrete example only on page 5 | A worked example on page 1 |
| Unscoped "our method generalizes" | "On the six languages tested; see Limitations" |
| Roadmap standing in for an argument | A one-line roadmap after the argument |
Scope the LLM-era claim
- If the paper uses or evaluates LLMs, bound the claim to the models, prompts, and settings tested, and disclose contamination risk. An unscoped "LLMs can/cannot do X" invites the reviewer to name the counterexample.
- Report prompts and decoding as part of the method, not as trivia (see
eacl-reproducibility).
Quantify the error analysis
- A page-one or early-section error analysis with counts ("40% of errors are agreement errors; examples in Table 3") is worth more than adjectives. EACL rewards papers that show where and why a system fails, especially across languages.
Anonymity-safe voice
Anonymity check before submission:
- no author names, affiliations, or acknowledgements
- no "as we showed in our EMNLP 2025 paper" -> use third-person citation
- no links that identify authors (personal repos, named grant pages)
- self-citations phrased neutrally
Multilingual clarity
- Name languages and scripts explicitly; render diacritics correctly in the PDF and later in the
Anthology metadata (
eacl-camera-ready). - Do not let an aggregate multilingual score stand in for per-language honesty — a table beats an average.
Compression discipline
- The content pages carry the argument; appendices carry detail. If cutting for length pushes
a core claim into an appendix, cut something else instead (see
eacl-supplementary). - The Limitations section is free space and read — use it to state scope, not to hide results.
Output format
[First-page arc] Present / Missing elements: <task/gap/what/result/scope>
[Overclaims] <phrases to scope, with fix>
[Evidence pairing] <claims lacking a table/number>
[Anonymity] <any leak>
[Multilingual honesty] <aggregate-hiding issues>
[Compression] <what to cut so the body carries the claim>