WikiDiverse Dataset

WikiDiverse is a high-quality human-annotated MEL dataset with diversified contextual topics and entity types from Wikinews. It has 8K image-caption pairs and uses Wikipedia as the corresponding knowledge base. WikiDiverse is provided by Alibaba Damo NLP Group.

搜集汇总

背景与挑战

背景概述

WikiDiverse是一个高质量的人工标注多模态实体链接数据集，包含8000个图像-标题对，源自Wikinews并利用Wikipedia作为知识库。该数据集由阿里巴巴达摩院NLP组提供，专注于多样化的上下文主题和实体类型。