ACLSUM

ACLSUM是一个由曼海姆大学专家精心制作和评估的新型摘要数据集，专注于科学出版物的多方面摘要。该数据集包含250篇文档，每篇文档都附有手动制作和验证的摘要，涵盖挑战、方法和成果等深度内容。ACLSUM的创建过程依赖于领域专家的标注，采用两阶段摘要标注方法，确保每篇源文档都有两种类型的黄金标准标注：与每个方面相关的句子集和抽象参考摘要。该数据集主要应用于自动多方面摘要的科学论文领域，旨在解决高质摘要生成的问题。ACLSUM is a novel summarization dataset meticulously curated and evaluated by experts from the University of Mannheim, focusing on multi-faceted summarization of scientific publications. The dataset consists of 250 documents, each paired with manually created and validated summaries covering in-depth content such as research challenges, methodologies, and findings. The creation of ACLSUM relies on annotations from domain experts, adopting a two-stage summarization annotation approach to ensure that each source document has two types of gold-standard annotations: a sentence set associated with each aspect, and an abstractive reference summary. This dataset is primarily applied in the field of automatic multi-faceted summarization for scientific papers, aiming to address the challenge of generating high-quality summaries.