SemEval-2022 Task 8: Multilingual news article similarity dataset

该数据集包含了根据多个相似性维度评估的新闻文章对，这些维度包括地理焦点、叙事模式、时间段和语气等。它特别对新闻文章对是否涵盖了相同的实质性新闻故事进行评分。为了增强数据集的多样性和规模，通过增加合成写作风格的调整、实体替换以及翻译成其他语言的方式进行扩充。经过增强后，该数据集包含了410万篇文章对，其任务是对新闻文章进行聚类和相似性评估。

This dataset consists of news article pairs evaluated across multiple similarity dimensions, including geographic focus, narrative patterns, time periods, and tone. It specifically scores whether pairs of news articles cover the same substantive news story. To enhance the dataset's diversity and scale, it was augmented via synthetic writing style adjustments, entity replacement, and translation into other languages. After augmentation, the dataset contains 4.1 million article pairs, and is designed for tasks of news article clustering and similarity evaluation.