Dataset helps evaluate gender bias in machine translation models
1,150 text segments. Nine languages. Amazon just released the dataset that exposes gender bias in AI translation.

Why it matters
This standardized dataset provides AI leaders with concrete tools to measure and address gender bias in translation models, making bias evaluation more systematic and comparable across different AI systems.
The key facts
4 to know1,150 text segments in test set
Nine languages covered
Focus on gender bias evaluation
Machine translation model testing
Go to the source
Amazon Scienceamazon.science
Publisher excerpt: Test set includes 1,150 text segments, each in nine languages.

