humanize-chinese

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Humanize Chinese

中文文本人性化改写

Use this skill when you need to detect AI-like Chinese writing, rewrite it to feel less synthetic, reduce AIGC signals in academic prose, or convert the text into a more specific Chinese writing style.

当你需要检测类AI风格的中文写作、将其改写得更具自然感、降低学术文本中的AIGC特征，或是将文本转换为特定中文风格时，可使用本技能。

When to Use

使用场景

Use when the user says

去AI味

降AIGC

去除AI痕迹

让文字更自然

改成人话

, or

降低AI率

Use when the user wants a Chinese text checked for AI-writing patterns or suspicious phrasing
Use when the user wants academic-paper-specific AIGC reduction for CNKI, VIP, or Wanfang-style checks
Use when the user wants Chinese text rewritten into a different style such as
```
zhihu
```
,
```
xiaohongshu
```
,
```
wechat
```
,
```
weibo
```
,
```
literary
```
, or
```
academic
```

当用户提及「去AI味」「降AIGC」「去除AI痕迹」「让文字更自然」「改成人话」或「降低AI率」时
当用户需要检查中文文本是否存在AI写作模式或可疑表述时
当用户需要针对知网（CNKI）、维普（VIP）、万方等平台的检测标准，对学术论文进行AIGC特征降重时
当用户需要将中文文本改写为「知乎」「小红书」「微信」「微博」「文学性」或「学术性」等特定风格时

Core Workflow

核心工作流

1. Detect Before Rewriting

1. 先检测再改写

Start by identifying the most obvious AI markers instead of rewriting blindly:

rigid
```
first/second/finally
```
structures

mechanical connectors such as

综上所述

值得注意的是

由此可见

abstract grandiose wording with low information density
repeated sentence rhythm and paragraph length
academic prose that sounds too complete, too certain, or too template-driven

If the user provides a short sample, call out the suspicious phrases directly before rewriting.

先识别最明显的AI特征，而非盲目改写：

生硬的「第一/第二/最后」结构
机械式衔接词，如「综上所述」「值得注意的是」「由此可见」
信息密度低的空泛宏大表述
重复的句子节奏和段落长度
过于完整、绝对或模板化的学术文本

若用户提供的文本样本较短，可先指出可疑表述再进行改写。

2. Rewrite in the Smallest Useful Pass

2. 精准改写，避免全盘重生成

Prefer targeted rewrites over total regeneration:

remove formulaic connectors rather than paraphrasing every sentence
vary sentence length and paragraph rhythm
replace repeated verbs and noun phrases
swap abstract summaries for concrete observations where possible
keep the original claims, facts, citations, and terminology intact

优先选择针对性改写，而非完全重新生成文本：

删除公式化衔接词，而非逐句改写
调整句子长度和段落节奏
替换重复的动词和名词短语
尽可能将抽象概括替换为具体表述
保留原文的论点、事实、引用和专业术语

3. Validate the Result

3. 结果验证

After rewriting, verify that the text:

still says the same thing
sounds less templated
uses more natural rhythm
does not introduce factual drift
stays in the correct register for the target audience

For academic text, preserve a scholarly tone. Do not over-casualize.

改写完成后，需验证文本：

核心含义保持不变
模板感降低
节奏更自然
未引入事实偏差
符合目标受众的语体风格

对于学术文本，需保留学术语气，避免过度口语化。

Optional CLI Flow

可选CLI工作流

If the user has a local clone of the source toolkit, these examples are useful:

bash

python3 scripts/detect_cn.py text.txt -v
python3 scripts/compare_cn.py text.txt -a -o clean.txt
python3 scripts/academic_cn.py paper.txt -o clean.txt --compare
python3 scripts/style_cn.py text.txt --style xiaohongshu -o out.txt

Use this CLI sequence when available:

detect and inspect suspicious sentences
rewrite or compare
rerun detection on the cleaned file
optionally convert into a target style

若用户本地克隆了源工具包，以下示例可供使用：

bash

python3 scripts/detect_cn.py text.txt -v
python3 scripts/compare_cn.py text.txt -a -o clean.txt
python3 scripts/academic_cn.py paper.txt -o clean.txt --compare
python3 scripts/style_cn.py text.txt --style xiaohongshu -o out.txt

若CLI可用，可遵循以下流程：

检测并检查可疑句子
改写或对比原文
对清洗后的文件重新检测
（可选）转换为目标风格

Manual Rewrite Playbook

手动改写指南

If the scripts are unavailable, use this manual process.

若无法使用脚本，可遵循以下手动流程。

Common AI Markers

常见AI特征

numbered or mirrored structures that feel too symmetrical
filler transitions that add no meaning
repeated stock phrases
overly even sentence length
conclusions that sound final, polished, and risk-free

过于对称的编号或镜像结构
无实际意义的填充性过渡语
重复的套话
过于均匀的句子长度
过于完美、无风险的结论表述

Rewrite Moves

改写技巧

delete weak transitions first
collapse repetitive phrases into one stronger sentence
split sentences at natural turns instead of forcing long balanced structures
merge choppy sentences when they feel robotic
replace generic abstractions with concrete wording
introduce light variation in cadence so the prose does not march at a constant tempo

先删除无效过渡语
将重复表述合并为更凝练的句子
按自然逻辑拆分长句，避免强行构建平衡结构
合并过于零散的机器人化短句
用具体表述替换空泛抽象内容
适当调整节奏，避免文本节奏一成不变

Academic AIGC Reduction

学术文本AIGC降重

For papers, reports, or theses:

keep discipline-specific terminology unchanged
replace AI-academic stock phrases with more grounded scholarly phrasing
reduce absolute certainty with measured hedging where appropriate
vary paragraph structure so each section does not read like the same template
add limitations or uncertainty if the conclusion feels unnaturally complete

Examples of safer direction changes:

```
本文旨在
```
->
```
本文尝试
```
or
```
本研究关注
```

具有重要意义

值得关注

有一定参考价值

研究表明

前人研究发现

已有文献显示

Do not invent citations, evidence, or data.

针对论文、报告或毕业论文：

保留学科专属术语不变
将AI生成的学术套话替换为更贴合实际的学术表述
适当使用谨慎性表述，降低绝对化语气
调整段落结构，避免各章节呈现模板化特征
若结论过于完美，可补充局限性或不确定性内容

示例改写方向：

```
本文旨在
```
->
```
本文尝试
```
或
```
本研究关注
```

具有重要意义

值得关注

或

有一定参考价值

研究表明

前人研究发现

或

已有文献显示

禁止编造引用、证据或数据。

Style Conversion

风格转换

Use style conversion only after the base text is readable and natural.

Supported style directions from the source workflow:

```
casual
```
```
zhihu
```
```
xiaohongshu
```
```
wechat
```
```
academic
```
```
literary
```
```
weibo
```

When switching style, keep the user's meaning stable and change only tone, structure, and surface wording.

仅当基础文本已具备可读性和自然感后，再进行风格转换。

源工作流支持的风格转换方向：

```
casual
```
（口语化）
```
zhihu
```
（知乎风格）
```
xiaohongshu
```
（小红书风格）
```
wechat
```
（微信风格）
```
academic
```
（学术风格）
```
literary
```
（文学风格）
```
weibo
```
（微博风格）

转换风格时，需保留用户核心含义，仅调整语气、结构和表层表述。

Output Rules

输出规则

Show the main AI-like patterns you found
Explain the rewrite strategy in 1-3 short bullets
Return the rewritten Chinese text
If helpful, include a short note on remaining weak spots

列出检测到的主要类AI特征
用1-3条简短要点说明改写策略
返回改写后的中文文本
如有必要，可补充说明剩余的薄弱环节

Source

来源

Adapted from the

voidborne-d/humanize-chinese

project and its CLI/script workflow for Chinese AI-text detection and rewriting.

改编自

voidborne-d/humanize-chinese

项目及其用于中文AI文本检测与改写的CLI/脚本工作流。

Limitations

局限性

Use this skill only when the task clearly matches the scope described above.
Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.

仅当任务明确符合上述范围时使用本技能。
不得将输出结果替代为特定场景下的验证、测试或专家评审。
若缺少必要输入、权限、安全边界或成功标准，请暂停操作并请求用户澄清。