GPT-NL/DuidelijkeTaal-v1.0-split

该数据集是Duidelijk Taal数据集的副本，添加了训练和测试分割，由GPT-NL的基准测试团队创建，以便于重用和记录数据分割。原始描述：语言材料“自动文本简化的人类评估：众包结果”是在“Duidelijke taal”项目期间创建的，包含电子表格CrowdsourcingResults.cvs。该电子表格包含来自SoNaR语料库的句子/文本，这些句子/文本经过GPT-4简化，并包含人类对这些简化在复杂性、准确性和流畅性方面的评分。This dataset is a copy of the Duidelijk Taal dataset with training and test splits, created by the benchmarking team of GPT-NL to allow for re-use with documented data splits. Original description: The language material "Human evaluation of automated text simplification: crowdsourcing results" was created during the project "Duidelijke taal" and consists of the spreadsheet CrowdsourcingResults.cvs. That spreadsheet contains sentences/text from the SoNaR corpus, a version of those sentences / that text simplified by GPT-4 and the human ratings of those simplifications with respect to complexity, accuracy and fluency.