naist-nlp/XQ-MEval
收藏资源简介:
XQ-MEval是一个用于评估自动指标在跨语言评分偏见方面的基准数据集,基于[CC BY-S 4.0](https://creativecommons.org/licenses/by-sa/4.0/)许可发布。该数据集通过在高质量翻译中注入不同数量的多维质量度量(MQM)定义的错误,实现了跨语言的可控和可比较的翻译质量。数据集基于Flores+构建,包含多种语言对和错误类型(如添加、遗漏、误译和未翻译)。数据集的组织结构包括两个主要文件夹:results(包含单错误注入的输出)和merged_result(包含多错误合并的输出)。每个实例包含源句子、参考翻译和带有错误标记的机器翻译。数据集还提供了详细的统计数据和使用说明。
XQ-MEval is a benchmark released under [CC BY-S 4.0](https://creativecommons.org/licenses/by-sa/4.0/) for evaluating automatic metrics with respect to cross-lingual scoring bias. This dataset is constructed by injecting varying numbers of Multidimensional Quality Metric (MQM)-defined errors into high-quality translations, enabling controlled and comparable translation quality across languages. The dataset is based on Flores+ and includes multiple language pairs and error types (e.g., Addition, Omission, Mistranslation, Untranslated). The dataset is organized into two main folders: results (containing outputs with single injected errors) and merged_result (containing outputs with multiple merged errors). Each instance includes the source sentence, reference translation, and machine translation with error annotations. The dataset also provides detailed statistics and usage instructions.




