MuMiN

MuMiN是一个大规模的多语言多模态事实核查错误信息社交网络数据集，由布里斯托大学工程数学系创建。该数据集包含2100万条推文，属于2.6万个Twitter线程，每个线程都与1.3万个经过事实核查的声明语义链接，涵盖数十个主题、事件和领域，跨越41种不同语言，时间跨度超过十年。数据集通过Python包mumin提供，旨在帮助研究人员开发和评估自动错误信息检测模型，解决社交媒体上的错误信息问题。MuMiN is a large-scale multilingual multimodal fact-checking misinformation social network dataset, created by the Department of Engineering Mathematics, University of Bristol. This dataset contains 21 million Tweets belonging to 26,000 Twitter threads, each semantically linked to 13,000 fact-checked claims. It covers dozens of topics, events and domains, spans 41 different languages, and has a time span of over ten years. The dataset is provided via the Python package `mumin`, aiming to help researchers develop and evaluate automatic misinformation detection models to address the misinformation problem on social media.

搜集汇总

背景与挑战

背景概述

MuMiN是一个大规模多语言多模态事实核查的虚假信息社交网络数据集，结构为异构图，包含超过2100万条推文和190万用户，覆盖41种语言和10年时间，用于训练自动虚假信息检测模型。数据集提供两个图分类任务：声明分类和推文分类，旨在预测声明或推文是否涉及虚假信息，支持多语言和多模态分析。