README.md

metadata

dataset_info:
  features:
    - name: id
      dtype: string
    - name: audio
      dtype:
        audio:
          sampling_rate: 16000
    - name: text_indo
      dtype: string
    - name: text_en
      dtype: string
  splits:
    - name: train
      num_bytes: 69142263.41007572
      num_examples: 2303
    - name: validation
      num_bytes: 17165592.74192428
      num_examples: 576
  download_size: 85971677
  dataset_size: 86307856.15200001
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*

Dataset Details

This is the Indonesia-to-English dataset for Speech Translation task. This dataset is acquired from CoVoST. CoVoST2 is end-to-end speech-to-text translation. The dataset is based on using Mozillas open-source Common Voice database of crowdsourced voice recordings. CoVoST2 is covering several languages, one of which is Indonesia. The Indonesian data has 2879 utterances and approximately 2 hours and 58 minutes of audio data.

Processing Steps

Before the Fleurs dataset is extracted, there are some preprocessing steps to the data:

1. Remove some unused columns.
2. Switch the id column position into the first column.
3. Rename the sentence column to text_indo and translation column to text_en.
4. Cast the audio column into Audio object.
5. Split into Train and Validation.

Dataset Structure

DatasetDict({
    train: Dataset({
        features: ['id', 'audio', 'text_indo', 'text_en'],
        num_rows: 2892
    }),
    validation: Dataset({
      features: ['id', 'audio', 'text_indo', 'text_en'],
      num_rows: 724
    }),
})

Citation

@article

@misc{wang2020covost,
    title={CoVoST 2: A Massively Multilingual Speech-to-Text Translation Corpus},
    author={Changhan Wang and Anne Wu and Juan Pino},
    year={2020},
    eprint={2007.10310},
    archivePrefix={arXiv},
    primaryClass={cs.CL}

Credits:

Huge thanks to Yasmin Moslem for mentoring me.