V-MIND

该数据集是一个大规模的多模态新闻推荐数据集，包含了相应的图片和文本的新闻条目。相较于原有的MIND数据集，V-MIND数据集确保了每条新闻都有一个对应的图片，实现了100%的视觉覆盖率，并且对原始封面的恢复率也较高。该数据集包含了130,379条新闻和876,956位用户，旨在支持多模态新闻推荐任务。

This is a large-scale multimodal news recommendation dataset that includes news entries paired with corresponding images and textual content. Compared with the original MIND dataset, the V-MIND dataset ensures that every news entry has a matching image, achieving 100% visual coverage and boasting a high recovery rate for original news cover images. It encompasses 130,379 news entries and 876,956 users, and is designed to support multimodal news recommendation tasks.

搜集汇总

背景与挑战

背景概述

V-MIND是基于MIND数据集增强的版本，增加了新闻图片，文件名与MIND新闻ID对应。数据集提供了下载链接、示例数据格式和相关研究论文引用。