ParsiNLU

该数据集是首个包含多种高级任务的波斯语基准测试，涵盖了超过14.5千个新实例，分布在6个不同的自然语言理解（NLU）任务中。数据集还包括了由母语人士进行的手动标注，并通过与黄金标签对比评估来估计人类的表现。这些任务涵盖了阅读理解、文本蕴含、情感分析、问题改写、多项选择题问答以及机器翻译，共计14.5千个实例。

This dataset is the first Persian benchmark covering multiple advanced tasks, containing over 14,500 new instances across 6 distinct natural language understanding (NLU) tasks. It includes manual annotations conducted by native Persian speakers, and human performance is estimated via comparison against gold standard labels. The covered tasks include reading comprehension, textual entailment, sentiment analysis, question paraphrasing, multiple-choice question answering, and machine translation, with a total of 14,500 instances.