| name | dr-cook:chinese-polisher |
| description | Polish and refine academic Chinese writing for publication in Chinese-language journals. Use when improving Chinese grammar, vocabulary, sentence fluency, or domain-specific terminology in Chinese manuscripts, abstracts, or grant sections. Triggers on: Chinese polish, Chinese academic writing, 中文润色, 学术中文, 中文写作, 中文修改, 学术写作润色, polish Chinese, improve Chinese writing, academic Chinese, Chinese manuscript. Does NOT trigger on translation tasks: en-to-zh (English to Chinese translation) or zh-to-en (Chinese to English translation).
|
chinese-polisher
1. Overview
chinese-polisher refines Chinese academic text for publication in Chinese-language journals. It operates in three sequential passes: Pass 1 corrects grammar and syntax — sentence structure, connector word usage, and redundancy removal. Pass 2 elevates vocabulary and register — replacing colloquial words with academic equivalents, upgrading informal connectors to formal register, and standardizing domain-specific terminology (TCM terms via references/tcm-terminology-zh.md). Pass 3 improves sentence fluency — varying length and structure, ensuring paragraph coherence, and aligning logical connectors to actual relationships (因果/转折/递进/并列). The module accepts text from upstream modules via context_output.raw_text or directly from user input.
2. Parameters
Required
| Parameter | Values | Description |
|---|
text | string | Chinese text to polish. Paste directly or inherit from context_output.raw_text. |
Optional
| Parameter | Values | Description |
|---|
domain | tcm | bioinformatics | clinical | pharmacology (default: tcm) | Determines which domain-specific terminology rules apply during Pass 2. Default is tcm — most Chinese academic writing in this suite is TCM-focused. |
target_journal | string | If specified, adapt register and formality to the journal's style preferences. |
polish_level | light | standard | deep (default: standard) | Depth of revision applied. light corrects grammar and obvious errors only (Pass 1 only). standard applies all three passes. deep applies all three passes and additionally restructures sentences for maximum fluency, suggesting alternative phrasings where beneficial. |
preserve_meaning | bool (default: true) | If false, allow more aggressive restructuring that may alter sentence framing while preserving the core argument. |
output_format | tracked_changes | clean | both (default: both) | tracked_changes shows original → revised pairs for each changed sentence. clean shows polished text only. both shows tracked changes first, then the full polished text. |
Parameter collection rule
If domain is not available from upstream context and has not been stated by the user, use the default (tcm) and proceed. Do not ask for optional parameters unless the user has explicitly mentioned a journal name or polish depth. If text is absent and no upstream raw_text exists, ask the user to paste the Chinese text.
3. Workflow
Step 1 — Check upstream context.
Inspect context_output first. If a prior module wrote raw_text, use it directly. Inherit domain, target_journal, and language without re-asking. If no upstream raw_text exists, ask the user to paste the Chinese text.
Step 2 — Verify language.
Check the pasted text for CJK characters. If the input is predominantly Latin script — defined as CJK characters comprising < 20% of total non-whitespace, non-punctuation characters (exclude Latin technical terms, gene names, and numeric notation from the denominator) — redirect: "检测到输入为英文，请使用 english-polisher 模块。" Do not attempt polishing.
Step 3 — Confirm polish level and domain.
Apply defaults if unspecified: polish_level: standard, domain: tcm. Record confirmed values for the output header.
Step 4 — Load reference files.
If domain: tcm, load references/tcm-terminology-zh.md (term pairs, syndrome rules, formula names). For all domains, load references/academic-chinese-style.md (vocabulary table, connector logic, structural patterns).
Step 5 — Execute three-pass polish.
- Pass 1 — Grammar and syntax: Fix misplaced or dangling modifiers. Remove redundant subject pronouns (avoid stacking 它/这/其 across adjacent sentences — restate the subject noun). Correct connector misuse: 所以→因此 (formal causality). Limit "的" modifier chains to ≤ 3 layers. Add explicit subject noun phrases where missing (e.g., "本研究") when the sentence subject differs from the preceding sentence. For
polish_level: light, stop here and proceed to Step 6.
- Pass 2 — Vocabulary and register: Apply the colloquial-to-academic vocabulary table from
academic-chinese-style.md (用→采用, 做→进行, 看→观察, etc.). Upgrade informal connectors to formal register: 但是→然而, 还有→此外, 然后→进而. If domain: tcm, apply TCM term standardization from tcm-terminology-zh.md: add 证 suffix to syndrome names, use full formula names (六味丸→六味地黄丸), correct syndrome component order. Remove embedded English words where an established Chinese equivalent exists. For polish_level: light, skip this pass.
- Pass 3 — Sentence fluency: Vary sentence length — alternate short (10–20 characters) with medium (25–40 characters); break sentences exceeding 60 characters without an internal comma. Audit logical connectors against actual relationships: causal → 因此/表明; contrast → 然而/尽管如此; additive → 此外/同时; progressive → 进而/在此基础上; summary → 综上所述. For
polish_level: deep, additionally suggest alternative phrasings and review paragraph-level topic sentence placement ("本研究…" / "结果显示…" / "分析表明…"). For polish_level: light, skip this pass.
Step 6 — Present output.
Return the polished result under the header [Chinese Polish — Domain: TCM | Level: Standard] (substituting actual values). Apply output_format rules: if tracked_changes or both, list changed sentences as 原文：… / 修改：… pairs before the full polished text; if clean, omit the pairs. Include [字数：原文 XX 字 → 修改后 YY 字] at the end. Close with: "需要我调整润色强度、针对特定段落重新润色，或协助翻译成英文吗？"
4. Output Format
[Chinese Polish — Domain: TCM | Level: Standard]
## 修改对照
原文：本研究通过对60例患者进行了观察并对其进行了数据收集和分析后发现...
修改：本研究观察60例患者，收集并分析数据，结果发现...
原文：气血不足，肾虚，脾虚的患者预后较差。
修改：气血不足、肾虚证及脾虚证患者的预后较差。
[additional 原文/修改 pairs...]
---
## 润色后全文
[full polished text]
---
[字数：原文 847 字 → 修改后 821 字]
- In
tracked_changes mode: show only the 原文/修改 pairs section, omit the full polished text.
- In
clean mode: show only the 润色后全文 section, omit the pairs.
- In
both mode (default): show pairs first, then the full polished text, then the word count.
- If no change was made to a sentence, do not include it in the pairs section.
5. context_output
Reads from upstream
| Field | Source | Usage |
|---|
raw_text | paper-writer, grant-writer, rebuttal-writer, zh-to-en (Chinese source), paper-reviewer | Primary input text for polishing; used without asking the user to paste |
target_journal | any upstream module | Triggers journal register adaptation in Pass 2 and Pass 3 |
language | any upstream module | Verified to be zh; if en detected, redirect to english-polisher |
domain | any upstream module | Loads the correct terminology reference file in Step 4 |
parameters.* | any upstream module | Inherits domain, target_journal, and polish_level if previously set |
Writes to output
{
"module": "chinese-polisher",
"domain": "<inherited or default: tcm>",
"target_journal": "<inherited or null>",
"language": "zh",
"raw_text": "<polished text — overwrites upstream raw_text>",
"summary": "<brief description: e.g., Standard Chinese polish, TCM domain, 847 characters, 12 grammar corrections, 8 vocabulary replacements>",
"status": "success | partial | failed",
"error_message": null,
"parameters": {
"domain": "<inherited or default: tcm>",
"polish_level": "<collected or default: standard>",
"preserve_meaning": true,
"output_format": "<collected or default: both>"
}
}
Downstream readers of raw_text: citation-checker, cover-letter-writer, english-polisher. Downstream readers of summary: cover-letter-writer.
6. References
See references/ for:
tcm-terminology-zh.md — Standardized TCM term pairs, syndrome naming rules, herb and formula naming conventions, and the six external pathogenic factors; loaded when domain: tcmacademic-chinese-style.md — Colloquial-to-academic vocabulary table, connector logic table, structural patterns for Chinese academic prose, and common error patterns; loaded for all domains
