Mocheg
收藏资源简介:
Mocheg数据集是由弗吉尼亚理工大学的研究人员创建的一个大规模数据集,包含15,601个声明,每个声明都标注了真实性标签和裁决声明。该数据集旨在支持端到端多模态事实检查和解释生成研究,通过检索相关证据(包括文章、图像、视频和推文)并预测真实性标签(支持、反驳或信息不足)来评估声明的真实性。此外,数据集还包括33,880个文本段落和12,112张图像作为证据,用于生成总结和解释推理及裁决过程的声明。Mocheg数据集的应用领域主要集中在自动化事实检查,旨在解决通过多模态信息源自动验证声明真实性的问题。
The Mocheg dataset is a large-scale dataset created by researchers at Virginia Tech, containing 15,601 claims, each annotated with a veracity label and a ruling statement. This dataset aims to support end-to-end multimodal fact-checking and explanation generation research, where the authenticity of claims is evaluated by retrieving relevant evidence including articles, images, videos, and tweets, and predicting veracity labels such as supporting, refuting, or insufficient information. Additionally, the dataset includes 33,880 text passages and 12,112 images as evidence for generating summaries and explanations that elaborate on the reasoning and ruling processes for the claims. The main application area of the Mocheg dataset is automated fact-checking, which targets the problem of automatically verifying the authenticity of claims through multimodal information sources.

- 1End-to-End Multimodal Fact-Checking and Explanation Generation: A Challenging Dataset and Models弗吉尼亚理工大学 · 2023年



