CoCoA-MT
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/amazon-science/contrastive-controlled-mt/tree/main/IWSLT2023
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为CoCoA-MT，旨在训练翻译任务中的属性控制器，以实现对正式程度和语法性别的控制。该数据集包含了多种语言的翻译，这些语言具有不同的属性，用于评估多语种对迁移学习的影响。它特别关注低资源语言，这些语言的训练数据相对有限。该数据集的主要任务是实现对机器翻译中正式程度和语法性别的属性控制。
This dataset, named CoCoA-MT, is intended for training attribute controllers in machine translation tasks to achieve control over formality level and grammatical gender. It includes translations across multiple languages with distinct attributes, which are used to evaluate the impact of multilingualism on transfer learning. It specifically focuses on low-resource languages that have relatively limited training data. The core task of this dataset is to realize attribute control over formality level and grammatical gender in machine translation.
提供机构：
Nadejde et al.
