SBIC (Social Bias Inference Corpus)

社会偏见框架是一种表示语言中隐含的偏见和冒犯性的新方式。例如，这些框架旨在提炼“我们不应该降低我们的标准以雇用更多女性”这一声明背后的“女性（候选人）资格较低”的含义。我们收集了社会偏见推理语料库 (SBIC)，其中包含 15 万条社交媒体帖子的结构化注释，涵盖了大约 1000 个人口群体的超过 3.4 万个含义。

Social bias framing is a novel paradigm for representing implicit biases and offensiveness in language. For example, these frameworks aim to extract the underlying meaning of the statement "We should not lower our standards to hire more women", which is that "women (candidates) are less qualified". We collected the Social Bias Inference Corpus (SBIC), which contains 150,000 structurally annotated social media posts, covering over 34,000 meanings across approximately 1,000 demographic groups.

搜集汇总

背景与挑战

背景概述

SBIC (Social Bias Inference Corpus) 是一个专注于社会偏见分析的数据集，包含15万条社交媒体帖子的结构化注释，覆盖约1000个人口群体和超过3.4万个含义，旨在通过社会偏见框架推理语言中的隐含偏见和冒犯性。该数据集由艾伦人工智能研究所于2020年发布，适用于自然语言处理和语言模型预训练等领域，帮助研究社会偏见在文本中的表现。