FRMT

FRMT是一个针对少样本区域感知机器翻译的新数据集和评估基准，包含从英语到葡萄牙语和普通话的两种区域变体的专业翻译。数据集选择源文档以支持对感兴趣现象的详细分析，包括词汇上不同的术语和干扰术语。该数据集旨在评估少样本区域感知翻译的质量，覆盖了葡萄牙语的巴西和葡萄牙两个区域，以及普通话的大陆和台湾两个区域。通过专业的人工翻译和质量验证，FRMT旨在捕捉区域特定的语言差异，并提供一个测试平台，以探索少样本属性控制。

FRMT is a novel dataset and evaluation benchmark for few-shot region-aware machine translation, containing professional translations of two regional variants from English to Portuguese and Mandarin. The dataset selects source documents to support detailed analysis of phenomena of interest, including lexically distinct terms and distractor terms. It aims to evaluate the quality of few-shot region-aware translation, covering two regional variants of Portuguese: Brazilian Portuguese and European Portuguese, as well as two regional variants of Mandarin: Mainland Mandarin and Taiwanese Mandarin. Through professional human translation and quality validation, FRMT seeks to capture region-specific linguistic differences and provide a testbed for exploring few-shot attribute control.