MTME multi-candidate dataset
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/google-research/mt-metrics-eval
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为MTME，包含了源句子及其由多个系统生成的候选翻译。这些数据被用于构建质量预测和草案细化任务的训练数据。此外，为了形成质量预测的训练实例，从不同质量水平的候选翻译中进行了抽样。该数据集所涉及的任务包括质量预测和草案细化。
The dataset named MTME contains source sentences and candidate translations generated by multiple machine translation systems. These data are employed to construct training data for the tasks of quality prediction and draft refinement. Additionally, to create training instances for quality prediction, sampling was performed on candidate translations across different quality tiers. The tasks covered by this dataset include quality prediction and draft refinement.
提供机构：
Google Research
