Word Translation Without Parallel Data

ICLR 2018 Alexis ConneauGuillaume LampleMarc'Aurelio RanzatoLudovic DenoyerHervé Jégou

State-of-the-art methods for learning cross-lingual word embeddings have relied on bilingual dictionaries or parallel corpora. Recent studies showed that the need for parallel data supervision can be alleviated with character-level information... (read more)

PDF Abstract

Evaluation results from the paper

Task Dataset Model Metric name Metric value Global rank Compare
Word Alignment en-es Adv - Refine - CSLS [email protected] 81.7 # 1
Word Alignment en-fr Adv - Refine - CSLS [email protected] 82.3 # 1
Word Alignment es-en Adv - Refine - CSLS [email protected] 83.3 # 1
Word Alignment fr-en Adv - Refine - CSLS [email protected] 82.1 # 1