Hugging Face's logo

Datasets:
opedromartins
/
ASR-datasets-ptbr

Tasks:
Automatic Speech Recognition
Modalities:
Audio
Text
Formats:
parquet
Languages:
Portuguese
Size:
1M - 10M
Tags:
ASR
Speech
Portuguese
Libraries:
Datasets
Dask
Dataset card Data Studio Files
xet
 Community
1
ASR-datasets-ptbr / README.md
opedromartins's picture
opedromartins
Update README.md
b4c83f0 verified
preview code
|
Raw
History Blame Contribute Delete
14.4 kB
---
language:
- pt
pretty_name: Datasets de Áudios em Português
task_categories:
- automatic-speech-recognition
tags:
- ASR
- Speech
- Portuguese
size_categories:
- 1M<n<10M
dataset_info:
- config_name: cetuc
 features:
 - name: audio_filename
 dtype: string
 - name: speaker
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 18625301836.532
 num_examples: 100998
 download_size: 16203229760
 dataset_size: 18625301836.532
- config_name: cml
 features:
 - name: audio_filename
 dtype: string
 - name: speaker
 dtype: int64
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 20718889028.07
 num_examples: 34265
 - name: validation
 num_bytes: 622662501.704
 num_examples: 1134
 - name: test
 num_bytes: 672966343.66
 num_examples: 1297
 download_size: 14416811016
 dataset_size: 22014517873.434
- config_name: codigo-de-defesa-do-consumidor
 features:
 - name: audio_filename
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 163568316
 num_examples: 253
 download_size: 163344293
 dataset_size: 163568316
- config_name: common-voice-22
 features:
 - name: audio_filename
 dtype: string
 - name: speaker
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 663017029.222
 num_examples: 22923
 - name: validation
 num_bytes: 297281875.8
 num_examples: 9640
 - name: test
 num_bytes: 310753188.028
 num_examples: 9641
 download_size: 1212797949
 dataset_size: 1271052093.0500002
- config_name: constituicao-federal
 features:
 - name: audio_filename
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 1046116434.215
 num_examples: 1255
 download_size: 1031691001
 dataset_size: 1046116434.215
- config_name: coraa-mupe
 features:
 - name: audio_filename
 dtype: string
 - name: speaker
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 43077595985.954
 num_examples: 276881
 - name: validation
 num_bytes: 899185092.872
 num_examples: 9894
 - name: test
 num_bytes: 3488584728.856
 num_examples: 30968
 download_size: 41797963037
 dataset_size: 47465365807.68201
- config_name: coraa-nurc-sp
 features:
 - name: audio_filename
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 34969174532.398
 num_examples: 166971
 - name: validation
 num_bytes: 670050973.104
 num_examples: 3142
 - name: test
 num_bytes: 1445635984.73
 num_examples: 7110
 download_size: 27433293493
 dataset_size: 37084861490.232
- config_name: coraa-v1.1
 features:
 - name: audio_filename
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 62195592221.578
 num_examples: 382258
 - name: validation
 num_bytes: 1038744634.576
 num_examples: 7522
 - name: test
 num_bytes: 2393506411.512
 num_examples: 12676
 download_size: 66894810728
 dataset_size: 65627843267.666
- config_name: fleurs
 features:
 - name: audio_filename
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 2337501626.501
 num_examples: 2793
 - name: validation
 num_bytes: 297534890
 num_examples: 386
 - name: test
 num_bytes: 746473185
 num_examples: 919
 download_size: 3355807403
 dataset_size: 3381509701.501
- config_name: lapsbm
 features:
 - name: audio_filename
 dtype: string
 - name: speaker
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 103794148
 num_examples: 700
 download_size: 103780948
 dataset_size: 103794148
- config_name: mls
 features:
 - name: audio_filename
 dtype: string
 - name: speaker
 dtype: int64
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 9043539824.764
 num_examples: 37533
 - name: validation
 num_bytes: 224395981
 num_examples: 826
 - name: test
 num_bytes: 222289068
 num_examples: 871
 download_size: 10051399800
 dataset_size: 9490224873.764
- config_name: sidney
 features:
 - name: audio_filename
 dtype: string
 - name: speaker
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 1126762543.78
 num_examples: 5093
 download_size: 1145725002
 dataset_size: 1126762543.78
- config_name: tedx
 features:
 - name: audio_filename
 dtype: string
 - name: speaker
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 33928748416.572
 num_examples: 90244
 - name: validation
 num_bytes: 241633133.147
 num_examples: 1013
 - name: test
 num_bytes: 337367535.98
 num_examples: 1020
 download_size: 26230244292
 dataset_size: 34507749085.699
- config_name: tts-portuguese-corpus
 features:
 - name: audio_filename
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 3206466857.192
 num_examples: 3624
 download_size: 3298225717
 dataset_size: 3206466857.192
- config_name: voxforge
 features:
 - name: audio_filename
 dtype: string
 - name: speaker
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 71093523
 num_examples: 760
 download_size: 68849090
 dataset_size: 71093523
- config_name: yodas-granary
 features:
 - name: audio_filename
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 276696522207.66003
 num_examples: 6627902
 download_size: 1533787826166
 dataset_size: 276696522207.66003
- config_name: youtube-commons-granary
 features:
 - name: audio_filename
 dtype: string
 - name: audio
 dtype: audio
 - name: transcription
 dtype: string
 - name: duration
 dtype: float64
 splits:
 - name: train
 num_bytes: 26120801861.175
 num_examples: 214537
 download_size: 91287267946
 dataset_size: 26120801861.175
configs:
- config_name: cetuc
 data_files:
 - split: train
 path: cetuc/train-*
- config_name: cml
 data_files:
 - split: train
 path: cml/train-*
 - split: validation
 path: cml/validation-*
 - split: test
 path: cml/test-*
- config_name: codigo-de-defesa-do-consumidor
 data_files:
 - split: train
 path: codigo-de-defesa-do-consumidor/train-*
- config_name: common-voice-22
 data_files:
 - split: train
 path: common-voice-22/train-*
 - split: validation
 path: common-voice-22/validation-*
 - split: test
 path: common-voice-22/test-*
- config_name: constituicao-federal
 data_files:
 - split: train
 path: constituicao-federal/train-*
- config_name: coraa-mupe
 data_files:
 - split: train
 path: coraa-mupe/train-*
 - split: validation
 path: coraa-mupe/validation-*
 - split: test
 path: coraa-mupe/test-*
- config_name: coraa-nurc-sp
 data_files:
 - split: train
 path: coraa-nurc-sp/train-*
 - split: validation
 path: coraa-nurc-sp/validation-*
 - split: test
 path: coraa-nurc-sp/test-*
- config_name: coraa-v1.1
 data_files:
 - split: train
 path: coraa-v1.1/train-*
 - split: validation
 path: coraa-v1.1/validation-*
 - split: test
 path: coraa-v1.1/test-*
- config_name: fleurs
 data_files:
 - split: train
 path: fleurs/train-*
 - split: validation
 path: fleurs/validation-*
 - split: test
 path: fleurs/test-*
- config_name: lapsbm
 data_files:
 - split: train
 path: lapsbm/train-*
- config_name: mls
 data_files:
 - split: train
 path: mls/train-*
 - split: validation
 path: mls/validation-*
 - split: test
 path: mls/test-*
- config_name: sidney
 data_files:
 - split: train
 path: sidney/train-*
- config_name: tedx
 data_files:
 - split: train
 path: tedx/train-*
 - split: validation
 path: tedx/validation-*
 - split: test
 path: tedx/test-*
- config_name: tts-portuguese-corpus
 data_files:
 - split: train
 path: tts-portuguese-corpus/train-*
- config_name: voxforge
 data_files:
 - split: train
 path: voxforge/train-*
- config_name: yodas-granary
 data_files:
 - split: train
 path: yodas-granary/train-*
- config_name: youtube-commons-granary
 data_files:
 - split: train
 path: youtube-commons-granary/train-*
---
# 📚 Datasets de Áudio em Português (PT-BR)
Este repositório reúne **diversos corpora públicos de fala em português do Brasil**, combinados em um único dataset para facilitar treinamentos e pesquisas em **ASR (Automatic Speech Recognition)**.
O objetivo é fornecer um recurso amplo, padronizado e de fácil acesso para a comunidade.
---
## 📂 Datasets Integrados
A tabela abaixo lista todos os datasets incluídos, com suas informações:
| Dataset | Config Name | **TOTAL** | train | test | validation | Speaker | Fonte |
|---|---|:---:|:---:|:---:|:---:|:---:| :---: |
| Todos os Datasets | - | **15774.98 h** | 15650.38 h | 79.75 h | 44.85 h | - | - |
| YODAS-Granary | yodas-granary | **13423.68 h** | 13423.68 h | - | - | ❌ | [Fonte](https://huggingface.co/datasets/nvidia/Granary) |
| YouTube-Commons-Granary | youtube-commons-granary | **806.54 h** | 806.54 h | - | - | ❌ | [Fonte](https://huggingface.co/datasets/nvidia/Granary) |
| CORAA MuPe | coraa-mupe | **365.04 h** | 319.90 h | 32.90 h | 12.25 h | ✅ | [Fonte](https://huggingface.co/datasets/nilc-nlp/CORAA-MUPE-ASR) |
| CORAA v1.1 | coraa-v1.1 | **290.53 h** | 273.38 h | 11.24 h | 5.91 h | ❌ | [Fonte](https://github.com/nilc-nlp/CORAA) |
| CORAA NURC-SP | coraa-nurc-sp | **239.23 h** | 224.40 h | 10.23 h | 4.60 h | ❌ | [Fonte](https://huggingface.co/datasets/nilc-nlp/CORAA-NURC-SP-Audio-Corpus) |
| MLS | mls | **168.35 h** | 160.96 h | 3.74 h | 3.64 h | ✅ | [Fonte](https://www.openslr.org/94) |
| TEDx | tedx | **155.91 h** | 152.55 h | 1.82 h | 1.54 h | ✅ | [Fonte](https://www.openslr.org/100) |
| CETUC | cetuc | **144.66 h** | 144.66 h | - | - | ✅ | [Fonte](https://gitlab.com/fb-audio-corpora/alcaim16k-DVD1de4) |
| CML | cml | **85.36 h** | 78.09 h | 3.68 h | 3.59 h | ✅ | [Fonte](https://www.openslr.org/146) |
| Common Voice 22 | common-voice-22 | **51.32 h** | 26.38 h | 12.90 h | 12.04 h | ✅ | [Fonte](https://commonvoice.mozilla.org/pt/datasets) |
| Fleurs | fleurs | **14.71 h** | 10.18 h | 3.24 h | 1.29 h | ❌ | [Fonte](https://huggingface.co/datasets/google/fleurs) |
| TTS Portuguese Corpus | tts-portuguese-corpus | **10.48 h** | 10.48 h | - | - | ❌ | [Fonte](https://github.com/Edresson/TTS-Portuguese-Corpus) |
| Constituição Federal | constituicao-federal | **8.97 h** | 8.97 h | - | - | ❌ | [Fonte](https://gitlab.com/fb-audio-corpora/constituicao16k) |
| Sidney | sidney | **7.28 h** | 7.28 h | - | - | ✅ | [Fonte](https://igormq.github.io/datasets) |
| Código de Defesa do Consumidor | codigo-de-defesa-do-consumidor | **1.42 h** | 1.42 h | - | - | ❌ | [Fonte](https://gitlab.com/fb-audio-corpora/codigodefesaconsumidor16k) |
| LapsBM | lapsbm | **0.90 h** | 0.90 h | - | - | ✅ | [Fonte](https://igormq.github.io/datasets) |
| VoxForge | voxforge | **0.62 h** | 0.62 h | - | - | ✅ | [Fonte](https://igormq.github.io/datasets) |
---
## 🗂️ Estrutura do Dataset
O dataset está organizado em configs, onde cada config corresponde a um dataset ou corpora específico. Dentro de cada config, os dados estão divididos nos splits:
* `train`
* `test`
* `validation`
⚠️ **Importante**: Apenas os splits originais dos datasets foram mantidos, nem todos os datasets possuem splits de `test` e `validation`.
Cada amostra contém os seguintes campos:
* `audio_filename`: nome original do arquivo
* `speaker`: identificador do falante, presente apenas quando a informação está disponível no dataset original
* `audio`: conteúdo do áudio
* `transcription`: transcrição do áudio
* `duration`: duração em segundos
---
## 🚀 Como Utilizar
O dataset pode ser carregado pelo [🤗 Hugging Face Datasets](https://huggingface.co/docs/datasets):
### Listar todas as configurações disponíveis
```python
from datasets import get_dataset_config_names
configs = get_dataset_config_names("opedromartins/ASR-datasets-ptbr")
print(configs)
```
### Carregar uma configuração específica
```python
from datasets import load_dataset
# Exemplo: carregar todos os splits do CORAA v1.1
ds = load_dataset("opedromartins/ASR-datasets-ptbr", name="coraa-v1.1")
# Apenas o split de treino
ds_train = load_dataset("opedromartins/ASR-datasets-ptbr", name="coraa-v1.1", split="train")
```
### Uso em modo [streaming](https://huggingface.co/docs/datasets/stream)
```python
from datasets import load_dataset
ds_stream = load_dataset(
 "opedromartins/ASR-datasets-ptbr",
 name="coraa-v1.1",
 split="train",
 streaming=True
)
print(next(iter(ds_stream)))
```
---
## ⚖️ Licenciamento
Cada dataset mantém **sua licença original**.
É responsabilidade do usuário verificar e respeitar os termos de uso de cada corpus.
Consulte os links das fontes listadas na [tabela](#📂-datasets-integrados) para mais detalhes.