SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsacademic-writing

academic-writing

Writing & research

View repository

Updated 10-21 16:21

About this skill

Academic Writing Skill (Formatting & Polishing + Humanization)

For default output paths, see the repository root directory README.md（{{DRAFT_DIR}}, default my-thesis/）。Academic Integrity and Legitimate Usesee the root README.md's"Design Boundaries and Academic Integrity"。

Mode Selection

Mode Applicable Scenarios Output
Formatting/Polishing "Polish this paragraph", "rewrite as an academic paragraph", "terminology consistency", "formula standardization" Standardized academic paragraph, saved as my-thesis/第X章_主题_年月日.md
Humanization "reduce AI flavor", "humanize", "remove AI traces", "rewrite more naturally" Engineering/operations research style rewriting, optionally run ai_detector/text_analyzer, save to my-thesis

Select the mode based on the user's wording; when unclear, if the user provides a draft paragraph and requests "polishing/standardization", use formatting/polishing; if they emphasize "human-like/natural/remove AI", use humanization.

I. Formatting/Polishing Mode

Core Points

  • Paragraph-level polishing: output continuous text paragraphs that can be inserted directly into the paper, without Markdown headings (#, ##).
  • Output an MD file each time: named 第X章_主题_年月日.md, saved to the working directory my-thesis/。
  • Academic style: complies with degree thesis standards, terminology see references/terminology.md, formulas see references/formula-templates.md。
  • Human writing tendencies: appropriately varied sentence lengths, diverse transitions (in view of this, by contrast), academic hedging (in most cases, existing evidence indicates), avoiding "first, second, finally" templates and typical AI sentence patterns.
  • Default humanizer-zh self-check: before output, perform a de-AI-trace self-check on the paragraph, against references/humanizer-zh-patterns.md(delete filler phrases, break formulaic structures, vary rhythm, trust the reader, delete punchy lines; scan through with a quick checklist).

Workflow

  1. Receive draft paragraphs or technical points
  2. Semantic analysis, identify chapter affiliation and polishing needs
  3. Terminology standardization (consult terminology.md), formula standardization (consult formula-templates.md), logic optimization
  4. humanizer-zh self-check: go through the references/humanizer-zh-patterns.md quick checklist, correct AI traces
  5. Output pure paragraph Markdown and save to my-thesis/

For detailed steps and examples, see references/workflow.md。

Thesis Chapter Adaptation

  • Chapter 3: Trafficability analysis (vehicle dynamics, soil mechanics, trafficability quantification)
  • Chapter 4: Path planning (A*, heuristic function, cost function)
  • Chapter 5: Transportation scheduling (VRP/PDP, mathematical models, optimization methods)
  • Chapter 6: Experimental validation (experimental design, result analysis, performance comparison)

II. Humanization Mode (Engineering/Operations Research, Chinese and English)

Academic Integrity

  • Purpose: To improve the naturalness and academic feel of text drafted with AI assistance based on one's own research and ideas.
  • Legitimate use: Revise AI-drafted text based on one's own viewpoints; do not submit content that cannot represent one's own intellectual contribution.
  • Principle: Authentic academic expression, not evading detection.

Workflow

  1. Analyze text: Optionally run scripts/ai_detector.py（--lang zh or en), detect sentence length uniformity, mechanical transition words, abstract boilerplate, lexical diversity, etc.
  2. Apply rewriting strategies：Firstfollow the patterns of references/humanizer-zh-patterns.md and the 5 core rules to do one pass of de-tracing and naturalization,thenoverlay this skill's engineering/operations research terminology and references/rewriting-principles.md(sentence rhythm, reducing abstract scaffolding, eliminating mechanical transitions, academic tone and hedging, grounding in specificity); terminology and references/terminology.md、references/engineering-patterns.md aligned.
  3. Rewrite and explain: In Chinese, optionally attach 3–5 [Revision Notes]; by default save to my-thesis/, named the same as the formatting mode.

Disciplinary Norms Summary

  • General applicability: VCI/CI, subsidence, slope, geological hazard risk, etc.; expressions can be quantified.
  • Path planning: A*, heuristic function, cost function, grid map, etc.
  • Vehicle scheduling: CVRP, PDP, capacity constraints, time windows, etc.
  • Experiments and simulation: benchmark algorithms, comparative experiments, MAE/RMSE, computation time, etc.

See references/engineering-patterns.md、references/rewriting-examples.md。

Scripts and Tools

  • ai_detector.py: AI writing pattern detection,python scripts/ai_detector.py input.txt --lang zh(or en), supports --detailed、--json。
  • text_analyzer.py: sentence length distribution, TTR, transition word density, etc.,python scripts/text_analyzer.py input.txt, or original.txt revised.txt --compare。
  • Optional Chinese dependencies:jieba(see scripts/requirements.txt）。

Self-check list (after humanization)

  • Sentence lengths vary (short/medium/long combined)
  • Mechanical transition words reduced or replaced
  • Abstract boilerplate replaced with specific concepts or terminology
  • At least one specific method/parameter/scenario or academic hedge
  • Original meaning, citations, and data fully preserved
  • Disciplinary wording conforms to engineering/operations research conventions
  • humanizer-zh: No three consecutive sentences of the same length, not too many dashes, no three-part piling up, connectives (moreover/however, etc.) not piled up; self-checked according to the references/humanizer-zh-patterns.md quick checklist

Output conventions (common to both modes)

  • Polishing/rewriting resultsmustbe output to my-thesis/, naming format 第X章_主题_年月日.md。
  • Do not change the original meaning, key data, or citations; do not fabricate references or data.

Reference Documents