Hateful meme with Reasons Dataset (HatReD)

该数据集是一个全新的多模态仇恨表情包数据集，其中标注了潜在的仇恨语境原因。研究人员在下载这些表情包时必须同意Facebook的数据访问协议。该数据集涵盖了不同的社会目标群体，且平均解释长度为13.62个字符。规模上，它包含了3,304个针对3,228个仇恨表情包的标注原因。该数据集的任务是对仇恨表情包进行解释和分类。

This dataset is a novel multimodal hateful meme dataset where potential hateful contextual causes are annotated. Researchers must agree to Facebook's Data Access Agreement when downloading these memes. This dataset covers diverse social target groups, with an average explanation length of 13.62 characters. In terms of scale, it contains 3,304 annotated reasons for 3,228 hateful memes. The tasks of this dataset are explanation and classification of hateful memes.

搜集汇总

背景与挑战

背景概述

HatReD是一个仇恨性迷因数据集，包含对迷因的社会目标和推理注释，用于支持多模态仇恨内容分析。该数据集基于Facebook AI的Hateful Memes数据集构建，旨在解码仇恨性迷因的潜在含义，并提供可解释的标注信息。