Building Bridges: A Dataset for Evaluating Gender-Fair Machine Translation into German

本研究介绍了名为‘Building Bridges’的数据集，旨在评估性别公平的英语到德语机器翻译。该数据集由格拉茨大学等机构创建，包含115个名词，涵盖职业和普通名词，用于测试性别中性词汇的翻译。数据集通过增强社区创建的性别公平语言词典，并从百科全书文本和议会演讲中抽样多句测试实例构建。此数据集的应用领域主要集中在解决机器翻译中的性别偏见问题，推动性别公平语言的使用。

This study introduces a dataset named 'Building Bridges' for evaluating gender-fair English-to-German machine translation. Developed by institutions including the University of Graz, this dataset consists of 115 nouns covering both occupational and general terms, designed to test the translation of gender-neutral vocabulary. It is constructed by augmenting community-created gender-fair language lexicons and sampling multi-sentence test instances from encyclopedia texts and parliamentary speeches. The primary applications of this dataset lie in addressing gender bias in machine translation and advancing the adoption of gender-fair language.