MediaEval

该数据集来源于推特，专为消息级别的验证任务设计，其中包含了与谣言相关事件的推文。此外，该数据集还包含了411张独特的图片，分别与6,225条真实推文和7,558条虚假推文相关联。在规模上，开发集包含了约9,000条谣言推文和6,000条非谣言推文；而测试集则包含了约2,000条推文。该数据集的任务目标是谣言检测。

This dataset is sourced from Twitter and specifically designed for message-level verification tasks, containing tweets related to rumor-associated events. Additionally, it includes 411 unique images, each linked to 6,225 genuine tweets and 7,558 fake tweets. In terms of scale, the development set contains approximately 9,000 rumor tweets and 6,000 non-rumor tweets, while the test set comprises around 2,000 tweets. The task objective of this dataset is rumor detection.

搜集汇总

背景与挑战

背景概述

MediaEval数据集专注于视频记忆性预测，提供包含10,000个三秒视频的Memento10k数据集和10,000个七秒视频的VideoMem数据集，以及EEG记录。该数据集旨在通过多任务设置（视频和EEG基础预测）推动多媒体内容语义理解，适用于广告、教育等领域的研究。