My PhD advisor rewrote himself in bash (2010)
Shell and editor-based tools for flagging passive voice, weasel words, and duplicate phrases in academic writing prompt broader questions about how far automated style enforcement should go. Commenters weigh the benefits of concise, standardized prose—especially in technical and scientific texts—against the loss of nuance when adverbs, qualifiers, and subjective terms are stripped away. Alternatives such as Vale, classic Unix style checkers, and modern AI-based reviewers are compared, with many seeing them as useful prompts for human judgment rather than prescriptive arbiters of “good” writing.
Tooling and Alternatives
- Many like the shell scripts for quick checks (passive voice, weasel words, duplicates), especially outside Emacs.
- Others prefer integrated tools: Emacs modes, vale.sh, and classic Unix
diction/style. Vale is praised for CI integration and configurable style rules. - A web UI port of the scripts appears useful but had early regex/context bugs; these were iteratively fixed.
- Several argue LLMs could do richer, context-aware editing than brittle bash/regex, though one person reports LLMs missing obvious typos.
Adverbs, Weasel Words, and Nuance
- Some strongly endorse stripping adverbs and “weasel words” (“quite”, “very”, “surprisingly”) to reduce fluff and vague claims.
- Others counter that this often changes meaning: “quite difficult” ≠ “difficult”; “various methods” carries extra information.
- There is extensive debate on “quite” in English: meanings range from “somewhat” to “completely” or sarcastic negation, depending on dialect and tone—making it risky in technical writing.
- Critics suggest replacing beholder words with data and explicit comparisons (e.g., “3% vs expected 10%”) rather than emotional framing.
Technical Writing vs. Readability
- Some readers of technical docs and papers plead for maximal concision and structure; flowery prose and soft qualifiers are seen as time-wasting.
- Others insist slight “fluff” and qualifiers can prevent over-precision and better reflect messy reality (“very close” vs “close”).
- A humanities-oriented view warns that rule-based scripts lack semantic understanding and may push blind prescription; tools should flag, not auto-edit.
Discipline-Specific Issues
- In mathematics, there is a debate over whether “monotonically increasing” is redundant or clarifying compared to “increasing”; commenters give conflicting definitions and examples.
- One thread contrasts expectations in medicine vs “hard sciences” about error bars and claims like “surprisingly low” rates.
Academia, Training, and Resources
- Several note that in many fields, writing quality ranks below speaking skills and politics, though poor writing can still hurt reviews.
- Multiple books and talks on mathematical/scientific writing are recommended as primary training, with linters as a final polish.