UN Document Translator Dataset Overview

Description

• Dataset Name: UN Parallel Corpora
• Purpose: Fine‑tuning MarianMT models to support multilingual translation of United Nations documents.
• Features: Contains high‑quality multilingual parallel texts covering the United Nations' official languages.

Characteristics

• Multilingual Support: Supports translation among the six official UN languages.
• High‑Quality Parallel Text: Offers precise parallel translations suitable for formal, technical, and nuanced language.
• UN‑Specific Terminology: Accurately handles terminology unique to the United Nations.

Applications

• Model Fine‑tuning: Improves MarianMT models for higher accuracy and contextual awareness in UN document translation.
• Translation Services: Hosted on the Hugging Face platform for developers, linguists, and international organizations.

Performance

• Translation Accuracy: Cosine similarity scores above 93 % between model translations and native‑speaker references on unseen data.
• Advantage: Outperforms human translations in blind tests.

Source

• Origin: United Nations Parallel Corpus provided by the UN.
• Citation:
  • Ziemski, M., Junczys‑Dowmunt, M., and Pouliquen, B., (2016), The United Nations Parallel Corpus, Language Resources and Evaluation (LREC’16), Portorož, Slovenia, May 2016.