A sample Cantonese (Yue Chinese) audio corpus with transcriptions, romanization, and scenario labels. Each audio file is paired with metadata in CSV format.
| File | Description |
|---|---|
*.wav |
Audio recordings named by ID (e.g. 00020.wav) |
examples.csv |
Sample entries with text, pinyin, and scenario labels |
scenarioes.csv |
Scenario taxonomy extracted from the full corpus index |
LICENSE |
Apache License 2.0 |
- Format: WAV (RIFF PCM)
- Channels: mono
- Sample rate: 22050 Hz
- Bit depth: 16-bit
| Column | Description |
|---|---|
| 序号 | Entry ID; matches the numeric prefix of the corresponding .wav file |
| 粤语文本 | Cantonese text |
| 普通话文本 | Mandarin translation |
| 拼音 | Jyutping romanization (Hong Kong Linguistic Society scheme) |
| 场景 | Scenario label |
| Column | Description |
|---|---|
| 场景 | Scenario name |
| 父场景 | Parent scenario (for sub-scenarios) |
| 条目数 | Number of entries in the full corpus for this scenario |
Audio files are referenced by ID. For example, the row with 序号 20 in examples.csv corresponds to 00020.wav.
00020.wav ←→ examples.csv row where 序号 = 20
The corpus covers everyday Cantonese dialogue across many situations, including greetings, dining, shopping, travel, healthcare, banking, and more. See scenarioes.csv for the full list of scenarios and sub-scenarios.
Licensed under the Apache License 2.0.