Datasets:

mosesdaudu
/

translation_dataset

dataset_info:
  features:
    - name: english_text
      dtype: string
    - name: language
      dtype: string
    - name: translated_text
      dtype: string
    - name: split
      dtype: string
  splits:
    - name: train
      num_bytes: 36890855
      num_examples: 198084
    - name: test
      num_bytes: 4071501
      num_examples: 22009
  download_size: 21823665
  dataset_size: 40962356
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*

Loading and Splitting Dataset To Various Languages

In this example, I will show you how to load the dataset and split by language for your downstream task.

>>> from datasets import load_dataset

>>> # load dataset
>>> dataset = load_dataset("mosesdaudu/translation_dataset")

>>> dataset
DatasetDict({
    train: Dataset({
        features: ['english_text', 'language', 'translated_text', 'split'],
        num_rows: 198084
    })
    test: Dataset({
        features: ['english_text', 'language', 'translated_text', 'split'],
        num_rows: 22009
    })
})

>>> # Filter Dataset To Pidgin Language Only
>>> pidgin_dataset = dataset.filter(lambda example: example['language'] == 'pidgin')

>>> pidgin_dataset
DatasetDict({
    train: Dataset({
        features: ['english_text', 'language', 'translated_text', 'split'],
        num_rows: 22476
    })
    test: Dataset({
        features: ['english_text', 'language', 'translated_text', 'split'],
        num_rows: 2497
    })
})