Online Meetup: Translating Gendered Pronouns

Please register with Zoom at
https://us02web.zoom.us/webinar/register/WN_MS_6r1goTEGXjpqvmSvKVA

The meeting password will be emailed on the day of the event. Please log on from 11:00 AEDT for an 11:05 start.

Hi everyone!

We hope you're keeping well. We're thrilled to announce our year-end extravaganza - on Translating Gendered Pronouns with Kellie Webster. Kellie will be calling in from New York and talking about some really cool work she is doing to improve Google Translate's ability to handle gendered pronouns. If you're interested in asking a question, there will be a forum available during the event for that in the Zoom call - we'll see any questions you post. Look forward to seeing you there!

The Sydney NLP Team

Kellie Webster: Scalable Cross Lingual Pivots to Model Pronoun Gender for Translation

SUMMARY
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.

BIO
Kellie Webster has worked at Google since she was awarded her PhD from the University of Sydney in 2016. She is interested in a range of topics around making trustworthy NLP including the representation of gender, algorithmic bias and fairness, and interpretable models.