遇见数据集

wikiHow-image

收藏
arXiv2025-09-30 收录
官方服务:

资源简介:

该数据集旨在通过将WikiHow文章分解为步骤级摘要、文档以及相应的图片,为多模态摘要中事实性的主观评估构建而成。它不仅支持未来研究在步骤级和文章级摘要任务上进行实验,而且其规模包括6000篇随机选出的文章,用于验证集和测试集,其中验证集和测试集分别包含52个和193个步骤级示例。该数据集的任务是多模态摘要评估。

This dataset is constructed to facilitate factual subjective evaluation in multimodal summarization by decomposing WikiHow articles into step-level summaries, their complete source documents, and corresponding images. It not only supports experimental research on step-level and article-level summarization tasks in future studies, but also encompasses 6000 randomly selected articles for the validation and test sets, which contain 52 and 193 step-level examples respectively. The core task of this dataset is multimodal summarization evaluation.

提供机构:
Authors of the paper
二维码
社区交流群
二维码
科研交流群
商业服务