---
dataset_info:
- config_name: cantonese
  features:
  - name: id
    dtype: int32
  - name: audio
    dtype: audio
  - name: raw_transcription
    dtype: string
  - name: transcription
    dtype: string
  - name: num_samples
    dtype: int32
  - name: gender
    dtype: int64
  - name: lang_id
    dtype: int64
  - name: language
    dtype: string
  - name: lang_group_id
    dtype: int64
  splits:
  - name: data
    num_bytes: 2680103452.96
    num_examples: 3120
  download_size: 2542018371
  dataset_size: 2680103452.96
- config_name: english
  features:
  - name: id
    dtype: int32
  - name: audio
    dtype: audio
  - name: raw_transcription
    dtype: string
  - name: transcription
    dtype: string
  - name: num_samples
    dtype: int32
  - name: gender
    dtype: int64
  - name: lang_id
    dtype: int64
  - name: language
    dtype: string
  - name: lang_group_id
    dtype: int64
  splits:
  - name: data
    num_bytes: 2379988073.948
    num_examples: 3643
  download_size: 2360973517
  dataset_size: 2379988073.948
- config_name: mandarin
  features:
  - name: id
    dtype: int32
  - name: audio
    dtype: audio
  - name: raw_transcription
    dtype: string
  - name: transcription
    dtype: string
  - name: num_samples
    dtype: int32
  - name: gender
    dtype: int64
  - name: lang_id
    dtype: int64
  - name: language
    dtype: string
  - name: lang_group_id
    dtype: int64
  splits:
  - name: data
    num_bytes: 3193694434.6
    num_examples: 4600
  download_size: 3198917645
  dataset_size: 3193694434.6
configs:
- config_name: cantonese
  data_files:
  - split: data
    path: cantonese/data-*
- config_name: english
  data_files:
  - split: data
    path: english/data-*
- config_name: mandarin
  data_files:
  - split: data
    path: mandarin/data-*
license: other
task_categories:
- automatic-speech-recognition
- text-to-speech
language:
- yue
- eng
- cmn
pretty_name: GT Google FLEURS
size_categories:
- 10K<n<100K
---
# Dataset Card for GT Google FLEURS

**Languages**: Cantonese (`yue`), English (`eng`), Mandarin (`cmn`)

**Source**: https://huggingface.co./datasets/google/fleurs

**Data processing**:

1. Download the Cantonese (`yue_hant_hk`), English (`en_us`), and Mandarin (`cmn_hans_cn`) datasets from the source.
2. Consolidate the `train`, `validation`, and `test` tags.
3. Remove files with duplicate names.

| Languages | Numbers of data |
| :-------: | :-------------: |
| Cantonese |      3120       |
|  English  |      3643       |
| Mandarin  |      4600       |

| Languages | Sizes of dataset |
| :-------: | :--------------: |
| Cantonese |     2.68 GB      |
|  English  |     2.38 GB      |
| Mandarin  |     3.19 GB      |

| Languages | Durations of dataset |
| :-------: | :------------------: |
| Cantonese |      11.05 Hrs       |
|  English  |      10.31 Hrs       |
| Mandarin  |      14.07 Hrs       |

| Languages | Unique tokens in dataset |
| :-------: | :----------------------: |
| Cantonese |           2535           |
|  English  |           9123           |
| Mandarin  |           2559           |

> For more details, refer to the token JSON file for each language in the [dataset](https://huggingface.co./datasets/gtomato/google_fleurs/tree/main).

| Languages | Total tokens in dataset |
| :-------: | :---------------------: |
| Cantonese |         118157          |
|  English  |          89738          |
| Mandarin  |         192631          |