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

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

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

📑 Paper    |    🐙 GitHub    |    🤗 HuggingFace
🖥️ HuggingFace Space    |    🎤 Demo Page    |    💬 Contact Us

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.
  

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: [email protected] or [email protected]

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

Scan to join our WeChat discussion group