Touché23-ValueEval

Touché23-ValueEval数据集由包豪斯大学创建，旨在识别论点背后的人类价值观。该数据集包含9324条来自多种来源的论点，如宗教文本、政治讨论、报纸社论等，每条论点由3名众包工作者标注54种价值观。数据集扩展自Webis-ArgValues-22，增加了数据量和文化多样性，支持跨文化价值检测模型的开发。数据集适用于大规模数据集支持的社会科学研究、论证评估、目标受众的论证生成或选择，以及识别争议话题双方的对立和共享价值观。

The Touché23-ValueEval dataset, developed by Bauhaus University, is designed to identify human values underlying arguments. It comprises 9,324 arguments sourced from diverse materials including religious texts, political discussions, newspaper editorials and more. Each argument is annotated with 54 distinct value categories by three crowdworkers. As an extension of the Webis-ArgValues-22 dataset, Touché23-ValueEval has expanded data scale and enhanced cultural diversity, supporting the development of cross-cultural value detection models. This dataset is applicable to large-scale dataset-driven social science research, argumentation evaluation, target-audience-oriented argument generation or selection, as well as the identification of opposing and shared values held by both sides of controversial topics.