thevox/en-nb-10k

Update

See the larger dataset: en-nb-15k

Methadology

Used GPT-3.5 with a translation prompt to give two versions of translation in Norwegian from English - normal and "more natural". Context is also generated.

Data

Input sample format:

Oversett til Norsk:
{text}

Each output sample is formatted like this:

Kontekst: {a, b, c}
Oversettelse:
{translation}
Mer naturlig:
{improved_translation}

Future work

Dataset will be used to train LLM-based translation models based on LLama 2 and similar to try to rival DeepL and ChatGPT machine translation.

Dataset used

Original english text from: nampdn-ai/tiny-textbooks

Author Contaxt

jonaslsa@uio.no