chenweichiang/zh-metadiscourse-scale
- 类型：github
- 标识：chenweichiang/zh-metadiscourse-scale
- 链接：https://github.com/chenweichiang/zh-metadiscourse-scale
- 主题：engineering
- 主分类：evaluation
- 形态：benchmark
- 分类：academic-writing
- Stars：0
- 周增：+0
- 语言：Python
- 许可：NOASSERTION
- 最近提交：2026-09-26
- 简介：zhmd: a descriptive metadiscourse scale for Chinese academic writing. Not an AI detector. Counts ten Hyland-style interactive and interactional markers per 15k characters and reports machine-to-human rate ratios together with the conditions under which the scale fails. Companion code for the paper on 'AI flavour' as a kansei object.
- 上次采集：2026-09-26
- 首次采集：2026-08-20
- 学术状态：accepted
- 学术阶段：writing
- 学术主任务：integrity-ai-governance
- 学术方向：polish
- 学术用途：writing-typesetting, ai-writing-governance, research-integrity
- 能力分面：writing
- 学术相关度：80
- 学术判定：deterministic
- 学术证据：anchor:paper; anchor:academic; query:academic-writing; task:integrity-ai-governance:ai detector
- 学术复核时间：2026-09-26T13:30:02+08:00
- 待LLM分类：否
- 成熟度：experimental
- 场景：academic-writing、chinese、metadiscourse
- 学术辅助任务：paper-drafting
- 简介中文：zhmd：面向中文学术写作的描述性元话语量表。非 AI 检测器。每 15k 字符统计十项 Hyland 式交互与互动标记，并报告机器-人类比率及量表失效条件。为"AI 风味"作为感性（kansei）对象的论文提供配套代码。
- 来源文件：
- [GitHub Search]