NYTK/HuCommitmentBank

HuCommitmentBank是一个包含自然发生的 discourse 的语料库，其最后一个句子包含一个在取消蕴含算子下的从句嵌入谓词。该数据集也是匈牙利语言理解评估基准工具包[HuLU](https://hulu.nytud.hu/)的一部分。该数据集基于CommitmentBank语料库（de Marneffe等人，2019年）设计。数据收集和注释过程是在Péter Hatvani的硕士论文中进行的（_A Corpus to Investigate Projection Methods: The Hungarian Commitment Bank_）。9名母语为匈牙利语的注释者从MNSZ2中收集了1100个有效的文本片段，并对这些示例进行了7点李克特量表的标注。数据集分为训练集（250）、开发集（103）和测试集（250），并只在标准差小于1的实例中包含。测试集不包含标签。

HuCommitmentBank is a corpus of naturally occurring discourses whose final sentence contains a clause-embedding predicate under an entailment canceling operator. This dataset is part of the Hungarian Language Understanding Evaluation Benchmark Kit [HuLU](https://hulu.nytud.hu/). It was designed based on the CommitmentBank Corpus (de Marneffe et al., 2019). The data collection and annotation were conducted in Péter Hatvanis master thesis. Nine native Hungarian annotators collected 1100 valid text fragments from MNSZ2 and labeled them on a 7-point Likert scale. The dataset is split into a training set (250), a development set (103), and a test set (250), including only instances with a standard deviation less than 1. The test set is distributed without labels.