Fine-tuning a large speech translation model on GPT-4-reformulated gender-balanced training data raises MuST-SHE feminine-form accuracy from about 10% to over 84% without BLEU loss.
dev” in the MuSTC-v1.0-SET column. The evaluation phase is limited to the remaining utterances not labeled as “dev
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Addressing speaker gender bias in large scale speech translation systems
Fine-tuning a large speech translation model on GPT-4-reformulated gender-balanced training data raises MuST-SHE feminine-form accuracy from about 10% to over 84% without BLEU loss.