README.md

metadata

language:
  - ga
  - en
size_categories:
  - 1K<n<10K
task_categories:
  - automatic-speech-recognition
  - text-to-speech
  - translation
dataset_info:
  features:
    - name: audio
      dtype:
        audio:
          sampling_rate: 44100
    - name: text_ga
      dtype: string
    - name: text_en
      dtype: string
    - name: text_en_raw
      dtype: string
    - name: pronunciation
      dtype: string
    - name: url
      dtype: string
  splits:
    - name: train
      num_bytes: 155452186.824
      num_examples: 6149
  download_size: 153741856
  dataset_size: 155452186.824
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Dataset Details

This dataset was originally created by Bitesize Irish as part of their “In Irish” series.

Dataset Structure

The dataset duration is approx. 5 hours, precisely 5:14:58, and its structure is as follows:

Dataset({
    features: ['audio', 'text_ga', 'text_en', 'text_en_raw', 'pronunciation', 'url'],
    num_rows: 6159
})

License

All rights reserved to Bitesize Irish Ltd. The dataset is made available for educational and research purposes only. For licensing queries, contact [email protected]

Citation

@inproceedings{moslem2024leveraging,
  title={Leveraging Synthetic Audio Data for End-to-End Low-Resource Speech Translation},
  author={Moslem, Yasmin},
  booktitle={Proceedings of the 2024 International Conference on Spoken Language Translation (IWSLT 2024)},
  year={2024},
  month={April},
  url={https://arxiv.org/abs/2406.17363},
  address={Bangkok, Thailand}
}