Implicit Hate

仇恨言论在社交媒体上显着增加，对所有人口统计的受害者造成严重后果。尽管对歧视性言论的特征和检测给予了很多关注，但大多数工作都集中在明确或公开的仇恨言论上，未能解决基于编码或间接语言的更普遍的形式。为了填补这一空白，这项工作引入了一种理论上合理的隐含仇恨言论分类法和一个基准语料库，其中每条消息及其含义都有细粒度的标签。我们使用当代基线对我们的数据集进行系统分析，以检测和解释隐含的仇恨言论，并讨论挑战现有模型的关键特征。该数据集将继续作为理解这一多方面问题的有用基准。

Hateful speech has increased dramatically on social media, causing severe consequences for victims across all demographic groups. While much attention has been paid to the characteristics and detection of discriminatory speech, most work has focused on explicit or overt hateful speech, failing to address the more prevalent forms rooted in coded or indirect language. To fill this gap, this work introduces a theoretically grounded taxonomy of implicit hateful speech and a benchmark corpus, where each message and its underlying meaning are labeled with fine-grained tags. We conduct a systematic analysis of our dataset using contemporary baselines to detect and explain implicit hateful speech, and discuss key features that challenge existing models. This dataset will continue to serve as a valuable benchmark for understanding this multifaceted problem.

搜集汇总

背景与挑战

背景概述

Implicit Hate数据集是一个用于检测和解释社交媒体上隐含仇恨言论的基准语料库，每条消息及其含义都有细粒度的标签。该数据集由佐治亚理工学院和加州大学圣地亚哥分校于2021年发布，旨在填补现有研究在隐含仇恨言论检测方面的空白。