APE

该数据集包含了训练自动后编辑（APE）系统所需的三元组：源句子、机器翻译输出以及人工后编辑内容，同时还附有键盘敲击记录特征。该数据集允许与之前在WMT 2016-18 APE共享任务中提交的系统进行比较，使用相同的开发集和测试集。其规模包含23,000个训练样本（其中16,068个与WMT数据集相交），任务是对翻译进行自动后编辑。

This dataset provides the triplets necessary for training automatic post-editing (APE) systems: source sentences, machine translation outputs, and human post-edited content, paired with keyboard keystroke logging features. It allows for direct comparison with systems submitted to the WMT 2016–18 APE Shared Tasks, as it adopts the same development and test sets used in those tasks. The dataset contains 23,000 training samples, 16,068 of which overlap with the WMT dataset, and the core task is automatic post-editing of machine translations.