covost2

Riferimenti:

en_de

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_de')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

it_tr

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_tr')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

en_fa

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_fa')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

en_sv-SE

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_sv-SE')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

it_mn

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_mn')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

en_zh-CN

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_zh-CN')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

en_cy

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_cy')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

en_ca

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_ca')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

en_sl

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_sl')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

en_et

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_et')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

it_id

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_id')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

en_ar

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_ar')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

en_ta

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_ta')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

it_lv

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_lv')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

en_ja

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/en_ja')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 15531
'train' 289430
'validation' 15531
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

fr_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/fr_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 14760
'train' 207374
'validation' 14760
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

de_it

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/de_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 13511
'train' 127834
'validation' 13511
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

es_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/es_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 13221
'train' 79015
'validation' 13221
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ca_it

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/ca_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 12730
'train' 95854
'validation' 12730
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

it_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/it_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 8951
'train' 31698
'validation' 8940
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ru_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/ru_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 6300
'train' 12112
'validation' 6110
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

zh-CN_it

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/zh-CN_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 4898
'train' 7085
'validation' 4843
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

pt_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/pt_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 4023
'train' 9158
'validation' 3318
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

fa_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/fa_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 3445
'train' 53949
'validation' 3445
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

et_it

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/et_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 1571
'train' 1782
'validation' 1576
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

mn_it

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/mn_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 1759
'train' 2067
'validation' 1761
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

nl_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/nl_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 1699
'train' 7108
'validation' 1699
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

tr_it

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/tr_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 1629
'train' 3966
'validation' 1624
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ar_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/ar_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 1695
'train' 2283
'validation' 1758
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

sv-SE_it

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/sv-SE_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 1595
'train' 2160
'validation' 1349
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

lv_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/lv_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 1629
'train' 2337
'validation' 1125
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

sl_it

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/sl_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 360
'train' 1843
'validation' 509
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ta_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/ta_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 786
'train' 1358
'validation' 384
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

ja_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/ja_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 684
'train' 1119
'validation' 635
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

id_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/id_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 844
'train' 1243
'validation' 792
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}

cy_en

Utilizzare il comando seguente per caricare questo set di dati in TFDS:

ds = tfds.load('huggingface:covost2/cy_en')
  • Descrizione :
CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. The dataset is created using Mozilla’s open source Common Voice database of crowdsourced voice recordings.

Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .mp3 format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:


python
import torchaudio

def map_to_array(batch):
    speech_array, _ = torchaudio.load(batch["file"])
    batch["speech"] = speech_array.numpy()
    return batch

dataset = dataset.map(map_to_array, remove_columns=["file"])
  • Licenza : nessuna licenza conosciuta
  • Versione : 1.0.0
  • Divide :
Diviso Esempi
'test' 690
'train' 1241
'validation' 690
  • Caratteristiche :
{
    "client_id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "file": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "sentence": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "translation": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    },
    "id": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    }
}