metadata
dataset_info:
config_name: eng-spa
features:
- name: source_text
dtype: string
- name: target_text
dtype: string
- name: source_lang
dtype: string
- name: target_lang
dtype: string
splits:
- name: validation
num_bytes: 18576444
num_examples: 197299
- name: test
num_bytes: 2577290
num_examples: 24514
download_size: 9208845
dataset_size: 21153734
configs:
- config_name: eng-spa
data_files:
- split: validation
path: eng-spa/validation-*
- split: test
path: eng-spa/test-*
This is a subset of the Tatoeba Challenge dataset. It contains only the English-to-Spanish data.
The full Tatoeba Challenge dataset is available at Helsinki-NLP/tatoeba_mt, but it does not work with the latest Datasets library because it contains custom code. A newer version which works with the latest Datasets library is available at Helsinki-NLP/tatoeba_mt_train, however it does not currently contain all language pairs (in particular, the Spanish-to-English subset that I needed is not there). This issue may be fixed by the time you read this.