Dataset Card for GT Mozilla Foundation Common Voice Corpus 18.0
Languages: Cantonese (yue), English (eng), Mandarin (cmn)
Source: https://commonvoice.mozilla.org/en/datasets
Data processing:
- Download the Cantonese (
yue,zh-HK), English (en), and Mandarin (zh-CN) datasets from the source. - Consolidate the
validated,dev,train,test, andothertags. - Remove files with duplicate names.
- Clean the datasets:
- For English: Remove entries with incorrect data structures.
- For Cantonese and Mandarin: Remove entries with
??in sentences.
| Languages | Numbers of data |
|---|---|
| Cantonese | 292012 |
| English | 2143694 |
| Mandarin | 915103 |
| Languages | Sizes of dataset |
|---|---|
| Cantonese | 8.43 GB |
| English | 89.91 GB |
| Mandarin | 38.96 GB |
| Languages | Durations of dataset |
|---|---|
| Cantonese | 324.23 Hrs |
| English | 3057.80 Hrs |
| Mandarin | 1128.64 Hrs |
| Languages | Unique tokens in dataset |
|---|---|
| Cantonese | 4329 |
| English | 305731 |
| Mandarin | 6512 |
For more details, refer to the token JSON file for each language in the dataset.
| Languages | Total tokens in dataset |
|---|---|
| Cantonese | 3007664 |
| English | 24066512 |
| Mandarin | 13049033 |
- Downloads last month
- 1,607