WenetSpeech-Yue

# WenetSpeech-Yue: A Large-scale Cantonese Speech Corpus with Multi-dimensional Annotation

<p align="center">
Longhao Li¹*, Zhao Guo¹*, Hongjie Chen²,
Yuhang Dai¹, Ziyu Zhang¹, Hongfei Xue¹,
Tianlun Zuo¹, Chengyou Wang¹, Shuiyuan Wang¹,
Xin Xu³, Hui Bu³, Jie Li², Jian Kang²,
Binbin Zhang⁴, Ruibin Yuan⁵, Ziya Zhou⁵,
Wei Xue⁵, Lei Xie¹
</p>

<p align="center">
¹ Audio, Speech and Language Processing Group (ASLP@NPU), Northwestern Polytechnical University <br>
² Institute of Artificial Intelligence (TeleAI), China Telecom <br>
³ Beijing AISHELL Technology Co., Ltd. <br>
⁴ WeNet Open Source Community <br>
⁵ Hong Kong University of Science and Technology
</p>

<p align="center">
📑 <a href="https://arxiv.org/abs/2509.03959">Paper</a> &nbsp&nbsp | &nbsp&nbsp
🐙 <a href="https://github.com/ASLP-lab/WenetSpeech-Yue">GitHub</a> &nbsp&nbsp | &nbsp&nbsp
🤗 <a href="https://huggingface.co/collections/ASLP-lab/wenetspeech-yue-68b690d287cde88389e5dba1">HuggingFace</a>
<br>
🖥️ <a href="https://huggingface.co/spaces/ASLP-lab/WenetSpeech-Yue">HuggingFace Space</a> &nbsp&nbsp | &nbsp&nbsp
🎤 <a href="https://aslp-lab.github.io/WenetSpeech-Yue/">Demo Page</a> &nbsp&nbsp | &nbsp&nbsp
💬 <a href="https://github.com/ASLP-lab/WenetSpeech-Yue?tab=readme-ov-file#contact">Contact Us</a>
</p>

<div align="center">
<img width="800px" src="https://github.com/ASLP-lab/WenetSpeech-Yue/raw/main/figs/wenetspeech_yue.svg" />
</div>

## Dataset
### WenetSpeech-Yue Overview
* Contains 21,800 hours of large-scale Cantonese speech corpus with rich annotations, the largest open-source resource for Cantonese speech research.
* Stores metadata in a single JSON file, including audio path, duration, text confidence, speaker identity, SNR, DNSMOS, age, gender, and character-level timestamps. Additional metadata tags may be added in the future.
* Covers ten domains: Storytelling, Entertainment, Drama, Culture, Vlog, Commentary, Education, Podcast, News, and Others.
<div align="center">
<img width="800px" src="https://github.com/ASLP-lab/WenetSpeech-Yue/raw/main/figs/data_distribution.png" />
</div>

### Metadata Format
We store all audio metadata in a standardized JSON format, where the core fields include `utt_id` (unique identifier for each audio segment), `rover_result` (ROVER result of three ASR transcriptions), `confidence` (confidence score of text transcription), `jyutping_confidence` (confidence score of Cantonese pinyin transcriptions), and `duration` (audio duration); speaker attributes include `speaker_id`, `gender`, and `age`; audio quality assessment metrics include `sample_rate`, `DNSMOS`, and `SNR`; timestamp information includes `timestamp` (precisely recording segment boundaries with `start` and `end`); and extended metadata under the `meta_info` field includes `program` (program name), `region` (geographical information), `link` (original content link), and `domain` (domain classification).

Json Example:
```
{
"key": "xg0054364_9798410_9801030",
"rover_result": "人多一齐食咁样先至知味",
"confidence": 0.879,
"jyutping_confidence": 0.909,
"duration": 2.816,
"meta_info": {
"region": "Hong Kong",
"program": "Cantonese radio drama "I'll Send You Flowers Next Year" featuring Kathy Chow, Jacob Tsui, and Law Wai-kit. A 2002 production by Radio Television Hong Kong (RTHK).",
"time_stamp": "9798.410_9801.030",
"link": "<link>",
"domain": "Drama"
},
"speaker_attributes": {
"spk_id": "xg0054364_SPEAKER_08",
"gender": "Male",
"age": "YOUTH"
},
"speech_quality": {
"sampling_rate": 16000,
"DNSMOS": 3.2549686431884766,
"SNR": 25.29012680053711
},
"timestamps": [
[["<eps>", [0.0, 0.26]], ["人", [0.26, 0.48]], ["多", [0.48, 0.64]], ["一", [0.64, 0.74]], ["齐", [0.74, 0.92]]],
[["食", [0.93, 1.15]], ["<eps>", [1.15, 1.39]], ["咁", [1.39, 1.53]], ["样", [1.52, 1.6]], ["先", [1.6, 1.75]]],
[["至", [1.75, 1.83]], ["知", [1.83, 2.04]], ["味", [2.04, 2.4]], ["<eps>", [2.4, 2.78]]]
]
}
```

### WenetSpeech Usage
You can obtain the original video source through the `link` field in the metadata file (`wenetspeech_yue_meta.json`). Segment the audio according to the `cut_point` field to extract the corresponding record. For pre-processed audio data, please contact us using the information provided below.

## Contact
If you have any questions or would like to collaborate, feel free to reach out to our research team via email: [e-mail] or [e-mail]

You’re also welcome to join our WeChat group for technical discussions, updates, and — as mentioned above — access to pre-processed audio data.

<p align="center">
<img src="https://github.com/ASLP-lab/WenetSpeech-Yue/raw/main/figs/wechat.jpg" width="300" alt="WeChat Group QR Code"/>
<em>Scan to join our WeChat discussion group</em>
</p>

<p align="center">
<img src="https://github.com/ASLP-lab/WenetSpeech-Yue/raw/main/figs/[e-mail]" width="300" alt="Official Account QR Code"/>
</p>

# WenetSpeech-Yue：面向粤语语音研究的多维度标注大规模语料库

<p align="center">
李龙浩¹*, 郭钊¹*, 陈宏杰²,
戴宇航¹, 张子瑜¹, 薛鸿飞¹,
左天伦¹, 王承友¹, 王水渊¹,
徐鑫³, 卜辉³, 李杰², 康健²,
张彬彬⁴, 袁瑞斌⁵, 周子雅⁵,
薛巍⁵, 谢磊¹
</p>

<p align="center">
¹ 西北工业大学音频、语音与语言处理课题组（Audio, Speech and Language Processing Group, ASLP@NPU）<br>
² 中国电信人工智能研究院（TeleAI）<br>
³ 北京爱数智慧科技有限公司（Beijing AISHELL Technology Co., Ltd.）<br>
⁴ WeNet开源社区（WeNet Open Source Community）<br>
⁵ 香港科技大学
</p>

<p align="center">
📑 <a href="https://arxiv.org/abs/2509.03959">论文</a> &nbsp&nbsp | &nbsp&nbsp
🐙 <a href="https://github.com/ASLP-lab/WenetSpeech-Yue">GitHub仓库</a> &nbsp&nbsp | &nbsp&nbsp
🤗 <a href="https://huggingface.co/collections/ASLP-lab/wenetspeech-yue-68b690d287cde88389e5dba1">HuggingFace集合页面</a>
<br>
🖥️ <a href="https://huggingface.co/spaces/ASLP-lab/WenetSpeech-Yue">HuggingFace演示空间</a> &nbsp&nbsp | &nbsp&nbsp
🎤 <a href="https://aslp-lab.github.io/WenetSpeech-Yue/">官方演示页面</a> &nbsp&nbsp | &nbsp&nbsp
💬 <a href="https://github.com/ASLP-lab/WenetSpeech-Yue?tab=readme-ov-file#contact">联系我们</a>
</p>

<div align="center">
<img width="800px" src="https://github.com/ASLP-lab/WenetSpeech-Yue/raw/main/figs/wenetspeech_yue.svg" />
</div>

## 数据集
### WenetSpeech-Yue 概述
* 包含21800小时的大规模粤语语音语料库，附带丰富的多维度标注，是目前开源领域规模最大的粤语语音研究资源。
* 所有元数据存储于单个JSON文件中，涵盖音频路径、时长、文本置信度、说话人身份、信噪比（Signal-to-Noise Ratio, SNR）、DNSMOS评分、年龄、性别以及字符级时间戳等信息，未来或将新增更多元数据标签。
* 覆盖十大应用领域：故事讲述、娱乐、戏剧、文化、Vlog、评论、教育、播客、新闻及其他类别。
<div align="center">
<img width="800px" src="https://github.com/ASLP-lab/WenetSpeech-Yue/raw/main/figs/data_distribution.png" />
</div>

### 元数据格式
我们采用标准化JSON格式存储所有音频元数据，核心字段包括：`utt_id`（单条音频片段的唯一标识符）、`rover_result`（三种自动语音识别（Automatic Speech Recognition, ASR）结果的ROVER融合结果）、`confidence`（文本转录置信度评分）、`jyutping_confidence`（粤语拼音（Jyutping）转录置信度评分）以及`duration`（音频时长）；说话人属性包含`speaker_id`、`gender`及`age`；音频质量评估指标涵盖`sample_rate`（采样率）、`DNSMOS`及`SNR`；时间戳信息包含`timestamp`（精确记录片段起止边界，含`start`与`end`字段）；`meta_info`字段下的扩展元数据包括`program`（节目名称）、`region`（地域信息）、`link`（原始内容链接）及`domain`（领域分类）。

JSON示例：
json
{
"key": "xg0054364_9798410_9801030",
"rover_result": "人多一齐食咁样先至知味",
"confidence": 0.879,
"jyutping_confidence": 0.909,
"duration": 2.816,
"meta_info": {
"region": "Hong Kong",
"program": "Cantonese radio drama "I'll Send You Flowers Next Year" featuring Kathy Chow, Jacob Tsui, and Law Wai-kit. A 2002 production by Radio Television Hong Kong (RTHK).",
"time_stamp": "9798.410_9801.030",
"link": "<link>",
"domain": "Drama"
},
"speaker_attributes": {
"spk_id": "xg0054364_SPEAKER_08",
"gender": "Male",
"age": "YOUTH"
},
"speech_quality": {
"sampling_rate": 16000,
"DNSMOS": 3.2549686431884766,
"SNR": 25.29012680053711
},
"timestamps": [
[["<eps>", [0.0, 0.26]], ["人", [0.26, 0.48]], ["多", [0.48, 0.64]], ["一", [0.64, 0.74]], ["齐", [0.74, 0.92]]],
[["食", [0.93, 1.15]], ["<eps>", [1.15, 1.39]], ["咁", [1.39, 1.53]], ["样", [1.52, 1.6]], ["先", [1.6, 1.75]]],
[["至", [1.75, 1.83]], ["知", [1.83, 2.04]], ["味", [2.04, 2.4]], ["<eps>", [2.4, 2.78]]]
]
}

### WenetSpeech-Yue 使用方式
您可通过元数据文件（`wenetspeech_yue_meta.json`）中的`link`字段获取原始视频源，并根据`cut_point`字段对音频进行分段，提取对应语音片段。如需获取预处理后的音频数据，请通过下方联系方式与我们取得联系。

## 联系方式
若您有任何疑问或合作意向，可通过以下邮箱联系我们的研究团队：[e-mail] 或 [e-mail]

您也可加入我们的微信技术交流群，获取最新动态与技术讨论——如前文所述，该群亦可提供预处理音频数据的获取渠道。

<p align="center">
<img src="https://github.com/ASLP-lab/WenetSpeech-Yue/raw/main/figs/wechat.jpg" width="300" alt="微信讨论群二维码"/>
<em>扫码加入我们的微信讨论群</em>
</p>

<p align="center">
<img src="https://github.com/ASLP-lab/WenetSpeech-Yue/raw/main/figs/[e-mail]" width="300" alt="官方公众号二维码"/>
</p>

搜集汇总

背景与挑战

背景概述

WenetSpeech-Yue是一个包含21,800小时粤语语音的大规模语料库，提供多维度标注信息并覆盖10个不同领域，是目前最大的开源粤语语音研究资源。