HeadlineCause

HeadlineCause是一个用于检测新闻标题间隐含因果关系的数据集，由莫斯科物理技术学院和Yandex创建。该数据集包含超过5000对英文新闻标题和9000对俄文新闻标题，均通过众包方式标注。数据集内容涵盖从完全无关到包含因果和反驳关系的标题对。创建过程涉及使用多种过滤器提取候选标注对，并通过众包平台进行标注。数据集主要应用于自然语言理解领域，旨在解决文本中隐含因果关系的自动推理问题，对新闻聚合和事件预测具有重要意义。HeadlineCause is a dataset designed to detect implicit causal relationships between news headlines, developed by Moscow Institute of Physics and Technology and Yandex. This dataset includes over 5,000 pairs of English news headlines and 9,000 pairs of Russian news headlines, all annotated via crowdsourcing. The dataset covers headline pairs ranging from completely unrelated samples to those containing causal or refuting relationships. Its creation process involved extracting candidate annotation pairs through multiple filtering steps, followed by annotation via crowdsourcing platforms. Primarily applied in the field of natural language understanding, this dataset aims to solve the automatic reasoning problem of implicit causal relationships in text, and holds significant value for news aggregation and event prediction.