Scalable Cross Lingual Pivots to Model Pronoun Gender for Translation
This addresses pronoun gender errors in translation for languages with gendered pronouns, offering an incremental improvement over existing methods.
The paper tackled the problem of machine translation errors when translating dropped or neutral pronouns into languages with gendered pronouns by proposing a cross-lingual pivoting technique to generate gender labels, achieving 92% F1 for Spanish dropped feminine pronouns compared to 30-71% for baseline models.
Machine translation systems with inadequate document understanding can make errors when translating dropped or neutral pronouns into languages with gendered pronouns (e.g., English). Predicting the underlying gender of these pronouns is difficult since it is not marked textually and must instead be inferred from coreferent mentions in the context. We propose a novel cross-lingual pivoting technique for automatically producing high-quality gender labels, and show that this data can be used to fine-tune a BERT classifier with 92% F1 for Spanish dropped feminine pronouns, compared with 30-51% for neural machine translation models and 54-71% for a non-fine-tuned BERT model. We augment a neural machine translation model with labels from our classifier to improve pronoun translation, while still having parallelizable translation models that translate a sentence at a time.