What makes 日语会话测试批改 notable?

1.6K downloads. Trust evidence available

How should 日语会话测试批改 be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

日语会话测试批改

面向 A1/A2/B1（对应 JLPT‑N5~N3）师生对话场景的日语口语批改 Skill。基于 faster‑whisper 完成音频 ASR 转写，采用 ** 内容完整性 50%、准确性 20%、流利度 30%** 三维加权预评分，秉持鼓励优先原则，降低语法偏差对总分的拖累。内置专有名词白名单，防止人名、地名被 AI 误判扣分；设置 ASR 置信度分级，低置信场景下 AI 分数仅作后台参考，交由教师人工打分。生成老师详细版、学生简洁版两套反馈，支持 CSV 批量导出。本技能平台无关，对接 LMS 需自行开发适配器；AI 仅做预批改，全部错误必须经过教师复核确认，不会自动判错扣分。 Skill: 日语会话测试批改 Owner: bianmaxingkong Summary: 面向 A1/A2/B1（对应 JLPT‑N5~N3）师生对话场景的日语口语批改 Skill。基于 faster‑whisper 完成音频 ASR 转写，采用 ** 内容完整性 50%、准确性 20%、流利度 30%** 三维加权预评分，秉持鼓励优先原则，降低语法偏差对总分的拖累。内置专有名词白名单，防止人名、地名被 AI 误判扣分；设置 ASR 置信度分级，低置信场景下 AI 分数仅作后台参考，交由教师人工打分。生成老师详细版、学生简洁版两套反馈，支持 CSV 批量导出。本技能平台无关，对接 LMS 需自行开发适配器；AI 仅做预批改，全部错误必须经过教师复核确认，不会自动判错扣分。 Tags: latest:1.1.7 Version history: v1.1.7 | 2026-08-21T14:26:54.248Z | user

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.6k

Updated

Oct 10, 2026

Version

1.1.7

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1.6K downloads reported by the source. Last updated 10/10/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 10, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 10, 2026
Adoption signal: 1.6K downloadsadoption · observed Oct 10, 2026
Latest release: 1.1.7release · observed Aug 21, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s17f7de24jwye68xxaykgnjqwn8616wn:japanese-conversation-scorer

Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.
Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-bianmaxingkong-japanese-conversation-scorer/snapshot"

Documentation

CLAWHUB

66,276 characters of source documentation, loaded on request.

Extracted files

3 files captured from the source.

SKILL.md

---
name: japanese-conversation-grader
display_name: 日语会话测试批改
description: A1/A2/B1(约对应JLPT的N5/N4/N3)日语会话测试音频批改；基于faster-whisper实现ASR转写，系统自动转写学生回答并做初步内容匹配与三维诊断（内容完整性/准确性/流利度）；不确定内容不扣分，只输出可疑候选供教师复核。支持单人/批量批改、纠错反馈、成绩归档；AI预批改辅助，教师人工复核，满分10分。
use_when: 教师批改日语会话作业录音、会话测试打分、批量评分与反馈
platform_independent: true
triggers:
 - 日语会话测试批改
 - 日语会话作业评分
---

## 会话测试形式与录音要求

**对话模式：师生对话**
教师根据课文内容编写5～6道提问。提前录制提问音频，每题之间预留固定作答时长。学生跟随音频参考课文内容回答，结束后提交音频文件。

---

## 标准答案与题目标注

1. **标准答案存储路径**：`./answer/{COURSE_ID}_{ASSIGNMENT_ID}_answer.txt`
2. **题目标注格式要求**
 - 每道题以 `[Q1]`、`[Q2]`……格式标注，便于音频转写后逐题对齐
 - 每题标注参考回答要点（关键词/句式），不要求逐字匹配

---

## 专有名词白名单配置
路径：`./answer/{COURSE_ID}_{ASSIGNMENT_ID}_proper_noun.txt`
每行1个名词，读音变体用`|`分隔，无专有名词则文件可不提供。

规则：
- 白名单内人名、地名、作品名，**AI不自动判错、不自动扣分**；
- ASR识别小偏差直接过滤；
- 专有名词严重读错，仅允许教师人工复核后扣分。

---

## 工作流程概览

Step1 音频预处理 → Step2 音频质量检测 → Step3 ASR转写 → Step4 ASR可靠性检测（置信度分级）→ Step5 三维AI评分 → Step6 教师复核 → Step7 成绩录入与反馈生成

---

## Step 1：音频预处理

- 批量转为 wav 格式，统一采样率
- 存放路径：`./audio/{student_id}/`

---

## Step 2：音频质量检测

检测项：
- 有效语音比例
- 音量范围
- 背景噪声
- 是否存在多人声音

**异常处理**：音频质量不足 → 不评分 → 请求重新上传。

---

## Step 3：ASR 转写

### 转写工具规格（强制）

- **唯一转写工具**：`faster-whisper`
- **仅允许模型**：`large-v3 fp16` / `medium`；禁用 `small` / `base` / `tiny` 及全部其他规格
- **硬件调度**：有充足 GPU 用 `large-v3 fp16`；无 GPU 自动降级 `medium`
- **固定开启** `word_timestamps=True` 输出词级时间戳

```bash
# 音频预处理
ffmpeg -y -i "input.mp4" -vn -acodec pcm_s16le -ar 16000 -ac 1 /tmp/audio.wav

# faster-whisper 日语转写
faster-whisper /tmp/audio.wav --model large-v3 --language ja --output_format txt --word_timestamps True
```

### 异常处理规则

- 音频损坏/空文件/无法识别 → 标记「转写失败」
- 自动移入 `./error_audio/`，触发人工复核
- 跳过自动评分，不直接判定低分

---

## Step 4：ASR 可靠性检测

系统综合判断以下方面：
- 文本匹配程度
- 音频质量
- ASR 置信度
- 对齐情况

输出：**高 / 中 / 低置信度**（用于 Step 6 教师复核的投入分级）。

### 置信度分级判定标准

| 置信度 | 判定条件 |
| :--- | :--- |
| **高** | 转写完整率 ≥80% 且 ASR 平均置信度 ≥0.8 且 音频信噪比 ≥20dB |
| **中** | 转写完整率 ≥60% 且 <80%，ASR 平均置信度 ≥0.5 且 <0.8 |
| **低** | 转写完整率 <60%，或 ASR 平均置信度 <0.5，或音频质量检测不通过 |

> 置信度‑AI评分衔接规则
> 1. **高、中置信度**：CAF三维AI加权分数作为正式预评分，教师仅做异常修正；
> 2. **低置信度**：AI三维加权分数仅作为后台诊断参考，**不计入、不对外展示，最终成绩完全由教师人工给出**。

---

## Step 5：三维 AI 评分

### 能力层级说明

遵循 **CEFR（欧洲语言共同参考框架）**；本系统采用 A1、A2、B1 三个等级。

| 能力层级 | 操作标签 | 会话评价重点 |
| :--- | :--- | :--- |
| **A1** | 约对应 JLPT N5 | 能围绕熟悉题目作简单、完整回答 |
| **A2** | 约对应 JLPT N4 | 能围绕日常话题进行较完整简单交流 |
| **B1** | 约对应 JLPT N3 | 能表达经历、理由和简单观点，注意连贯性 |

> JLPT 没有口语考试；N5/N4/N3 仅作为词汇、语法和课程任务难度参考。

---

### 评分原则

1. 内容切题比语法完美更重要
2. A1/A2（N5/N4）阶段轻微语法偏差以反馈为主，不扣分
3. 不确定错误不扣分
4. 鼓励为主，微小误差不扣分

> 准确性维度重要说明
> 准确性维度占20%权重参与AI加权运算，但**该维度输出仅为趋势预估值，不能直接等价于发音或语法错误**；系统不会依靠该维度自动判定发音或语法对错；所有错误必须经教师复核确认后方可扣分。

### 评分前置过滤规则（评分前必须执行）

计算任何维度分数前，先筛除以下情况，**排除出扣分范围**：

1. **ASR 误识别**——不进入评分计算
 - 同音词差异、语气词误识别
 - 系统标记为"ASR疑似误识别"的项
2. **转写存疑**——不进入评分计算
 - 系统置信度 < 0.5 的识别结果
3. **学习者自然偏差**——反馈但不扣分
 - A1/A2 阶段轻微语法偏差
 - 句尾语气词添加或省略
4. **白名单专有名词**——不进入AI自动扣分计算
   - 匹配白名单内的人名、地名、作品名；识别出现字形、读音轻微偏差，AI不自动判定词汇错误
   - 专有名词的严重误读，仅可由教师人工复核确认后扣分
---

### AI 自动评分模型（三维加权·主线）

| 评价维度 | 权重 | 数据来源 | 评分要点 |
| :--- | :--- |

_meta.json

{
  "ownerId": "kn7cw3j55n4hxe9ze445qp81ks86150x",
  "slug": "japanese-conversation-scorer",
  "version": "1.1.7",
  "publishedAt": 1787322414248
}

skill-card.md

## Description:

Grades A1, A2, and B1 Japanese conversation-test audio by transcribing responses with faster-whisper, producing teacher-reviewed pre-scores across content completeness, accuracy, and fluency, plus teacher and student feedback reports.

This skill is ready for commercial/non-commercial use.

## Publisher:

[bianmaxingkong](https://clawhub.ai/user/bianmaxingkong)

### License/Terms of Use:

MIT-0

## Use Case:

External educators and language-program staff use this skill to batch grade Japanese conversation-test recordings, review ASR-backed pre-scores, and generate separate detailed teacher reports and concise student feedback.

### Deployment Geography for Use:

Global

## Known Risks and Mitigations:

Risk: Optional course-platform adapters may post student grades and feedback through APIs without a clearly required final confirmation or dry-run.

Mitigation: Use the default CSV or local-report workflow, keep tokens in local config or environment variables, and require a teacher-reviewed preview before any adapter posts grades or feedback.

Risk: ASR-backed pre-scores can be unreliable when audio quality, transcript alignment, or recognition confidence is low.

Mitigation: Treat low-confidence AI scores as internal diagnostics only and require the teacher to provide the final score before student-facing feedback is released.

## Reference(s):

- [ClawHub skill page](https://clawhub.ai/bianmaxingkong/skills/japanese-conversation-scorer)

## Skill Output:

**Output Type(s):** [text, markdown, shell commands, configuration, guidance]

**Output Format:** [Markdown and plain-text feedback reports with inline shell commands and CSV-compatible summaries]

**Output Parameters:** [1D]

**Other Properties Related to Output:** [Produces teacher-facing detailed reports, student-facing concise feedback, and optional CSV exports; low-confidence ASR results require teacher scoring.]

## Skill Version(s):

1.1.7 (source: server release evidence)

## Ethical Considerations:

Users should evaluate whether this skill is appropriate for their environment, review any generated or modified files before relying on them, and apply their organization's safety, security, and compliance requirements before deployment.

Github ReposUpdated 15h agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.

MCPOPENCLAW

Github ReposUpdated 7mo agoRank 70

CopilotKit

The Frontend for Agents & Generative UI. React + Angular