Chinese doc optimizer

Chinese Technical Document Optimizer (中文技术文档优化器)

Make Chinese technical writing read like a native speaker wrote it — not like it was translated from English by an engineer in a hurry.

Why This Exists

Chinese technical documents suffer from five systemic problems that English-native LLMs and ESL engineers consistently produce:

1. 翻译腔 (Translation-ese) — "在这个功能中，我们实现了对数据的处理" vs "该功能处理数据"。Subject-less Chinese is natural; explicit 我们 + 在...中 is English-calqued.
2. 欧化句式 (Europeanized Syntax) — Long modifier chains before nouns, passive voice overuse, unnecessary 的 stacking.
3. 术语不一致 (Term Drift) — Same concept called 用户数据 in one paragraph and 使用者信息 in the next.
4. 废话 (Verbosity) — 需要注意的是，我们应该注意到... → just say it.
5. 标点混用 (Punctuation Chaos) — Chinese commas (，) mixed with English commas (,), full-width vs half-width confusion.

Workflow

Step 1: Receive the Text

Accept any Chinese technical document. Common formats: README, API docs, code comments, tech specs, changelogs, error messages, UI copy.

Step 2: Five-Pass Diagnosis

Read first, tag issues, then rewrite. Do NOT modify anything in the initial passes.

Pass 1: 翻译腔 (Translation-ese)

Look for these patterns:

• 在...中/下/方面 → remove the wrapper, use direct statement
• 我们 as subject → Chinese technical writing rarely needs explicit 我们
• 你可以... → redundant; imperative or declarative works
• 被... (passive voice) → Chinese prefers active voice
• 当...时 → often replaceable with simpler temporal structure
• 对于/关于... → calque of "regarding/concerning"
• 提供了...的功能 → 支持... or just the verb

Pass 2: 欧化句式 (Europeanized Syntax)

• Long noun-modifier chains (3+ 的 in one phrase) → break up or restructure
• Abstract verb inflation: 进行+verb, 做出+noun, 实现+noun → use the verb directly
  • 对数据进行分析 → 分析数据
  • 做出决定 → 决定
  • 实现功能 → just describe what it does
• Sentences mirroring English clause order → reorder to Chinese topic-comment structure

Pass 3: 术语一致性 (Term Consistency)

• Identify all technical terms and variants
• Flag inconsistencies
• Choose the most standard term and unify throughout

Pass 4: 废话与冗余 (Redundancy)

• 需要注意的是/值得注意的是/值得一提的是 → delete, just say it
• 当然/其实/基本上/一般来说 → delete unless they carry real meaning
• 功能/模块/系统 appended to every noun → delete where context makes it obvious
• 进行/实现/完成 + verb → use the verb alone
• Duplicate information across adjacent sentences → merge

Pass 5: 标点与格式 (Punctuation & Format)

• Chinese text MUST use Chinese punctuation: ，。、；：？！""（） not ,.;:?!""()
• Exception: code, variable names, URLs, file paths keep English punctuation
• Numbers and English words in Chinese text: add a space before and after (使用 React 构建 not 使用React构建)
• Code blocks use English punctuation
• Full-width characters for Chinese, half-width for English

Step 3: Generate the Optimized Version

Rewrite applying all fixes. Principles:

• Preserve all technical facts. Never change what the text says — only how.
• Preserve structure. Don't reorganize unless the structure itself is broken.
• Code, URLs, file paths are UNTOUCHABLE.
• Prefer simpler. 简洁 > 华丽.
• Default to declarative, active-voice, subject-optional Chinese.

Step 4: Generate the Change Log

Output a structured change log organized by category with counts and specific before→after examples:

## 修改说明

### 翻译腔修正 (X处)
- "在用户登录流程中" → "用户登录流程"

### 句式优化 (X处)
- "这是一个用于验证用户密码是否满足安全要求的模块" → 拆分为两句

### 术语统一 (X处)
- "用户数据" / "使用者信息" → 统一为 "用户数据"

### 冗余删除 (X处)
- 删除 "需要注意的是" / "当然" 等冗余表达

### 标点修正 (X处)
- 英文逗号 → 中文逗号

Pattern Reference

Translation-ese → Natural Chinese

❌ Translation-ese	✅ Natural Chinese
在本文档中，我们将介绍...	本文档介绍...
你可以通过以下命令来安装	安装命令：
这个功能被设计用来...	该功能用于...
当你点击按钮时	点击按钮后
对于大多数情况来说	大多数情况下
它提供了对数据的处理功能	它处理数据 / 支持数据处理
需要注意的是，这个配置...	注意：该配置...
为了能够实现...	为实现...
关于这个 API 的详细说明	该 API 的详细说明
存在着一些问题	存在问题
进行了修改	已修改

Abstract Verb Inflation

❌ Inflated	✅ Direct
对数据进行处理	处理数据
做出响应	响应
实现登录功能	登录
进行配置	配置
开展测试	测试
执行部署	部署

标点 Quick Rules

Context	Punctuation
Chinese sentences	， 。 、 ； ： ！ ？ " " （ ）
English within Chinese doc	, . ; : ! ? " " ( )
Code / filenames / URLs	Never change
Numbers next to Chinese	Add space: 需要 3 个参数

Output Format

Present the result in two sections:

## 优化后文档

[Complete optimized text, ready to paste back into the project]

---

## 修改说明

[Change log organized by category, with counts and specific before→after examples]

If the text exceeds 2000 characters, offer to process it in sections so the user can review incrementally.

Anti-Patterns (Never Do This)

• ❌ Do NOT translate Chinese to English or vice versa — output language = input language
• ❌ Do NOT modify code blocks, inline code, URLs, file paths, or API endpoints
• ❌ Do NOT add technical information not in the original
• ❌ Do NOT remove technical details to make it "simpler"
• ❌ Do NOT change document structure (section order, heading levels)
• ❌ Do NOT over-correct informal-but-natural Chinese into rigid formal style — 跑一下测试 is fine, do not turn it into 执行测试套件
• ❌ Do NOT correct regional vocabulary differences — both 服务器 (mainland) and 伺服器 (Taiwan) are valid; do not "standardize" between regional variants