README.md · laelhalawani/opus_and_europarl_en

metadata

license: mit
task_categories:
  - translation
language:
  - en
  - ro
size_categories:
  - 1M<n<10M

This dataset is a combination and conversion of two En-Ro datasets published by University of Helsinki, specifically Opus-100 and Europarl.

1,404,356 En-Ro text-pair samples were extracted and combined into a single dataset. Each sample is a dict with two keys corresponding to the text-pair languages names: en and ro and values of those keys represent the matching text in the given languages.

NOTE: Some initial analysis reveals many of these samples face formatting issues, and might be facing semantic translation issues.

To use

import datasets
ds = datasets.load_dataset("laelhalawani/opus_and_europarl_en_ro")

print(ds) # preview dataset
print(ds["train"]) # preview default split (all samples are here)
print(ds["train"][0]) # preview first sample
print(["train"][0]["en"]) # preview EN part of the first sample
print(["train"][0]["ro"]) # preview RO part of the first sample