Gender Bias in English-to-Portuguese Translation

This repository contains the data and code used in my master's thesis "Diving into Gender Translation Bias for the Portuguese Language - A Comparative Analysis of Commercial Machine Translation Systems, General-Purpose LLMs, and Non-Commercial Translation-Specific Models", supervised by Helena Moniz and Luísa Coheur.

Contents:

• Original WinoMT dataset, adapted to include Portuguese (with additional experiments and metrics)
• New extensions of the dataset:
  • inter-2: Two-sentence context
  • inter-3: Three-sentence context
• Fine-tuned model: configuration, scripts, and fine-tuning dataset

Acknowledgments

This work was inspired by and builds on datasets and methods from:

• Evaluating Gender Bias in Machine Translation (Stanovsky et al., 2019)
• Reducing Gender Bias in Neural Machine Translation as a Domain Adaptation Problem (Saunders and Byrne, 2020)