Scalable Cross Lingual Pivots to Model Pronoun Gender for Translation

TL;DR

This paper introduces a scalable cross-lingual pivoting method to accurately predict pronoun gender in translation, improving gender accuracy in machine translation systems by fine-tuning classifiers with high-quality gender labels.

Contribution

It presents a novel cross-lingual pivoting technique for automatic gender labeling, enhancing pronoun translation accuracy in neural machine translation models.

Findings

BERT classifier achieves 92% F1 on Spanish feminine pronouns.

Improved pronoun translation accuracy in NMT models.

Cross-lingual pivoting enables high-quality gender label generation.

Abstract

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.