FACTIFY

FACTIFY是一个关于多模式事实验证的数据集。它包含图像，文本声明，参考文本文档和图像。任务是借助支持数据将索赔分为支持，证据不足和反驳类别。我们的目标是通过提供这种多模式数据集来打击社交媒体时代的假新闻。Factify包含50,000个带有100,000图像的声明，分为培训，验证和测试集。

FACTIFY is a multimodal fact verification dataset. It contains images, textual claims, reference text documents, and images. The task is to classify claims into three categories—supported, insufficient evidence, and refuted—using supporting data. Our goal is to combat fake news in the social media era by releasing this multimodal dataset. Factify includes 50,000 claims paired with 100,000 images, which are split into training, validation, and test sets.

搜集汇总

背景与挑战

背景概述

FACTIFY是一个多模式事实验证数据集，包含50,000个声明和100,000张图像，用于分类声明为支持、证据不足或反驳类别，旨在打击社交媒体时代的假新闻。数据集由国立阳明交通大学于2021年发布，分为训练、验证和测试集。