TalkDown
收藏OpenXLab2026-04-18 收录
下载链接：
https://openxlab.org.cn/datasets/OpenDataLab/TalkDown
下载链接
链接失效反馈官方服务：
资源简介：
TalkDown 是一个标记数据集，用于在上下文中进行居高临下的检测。该数据集来自 Reddit，这是一组内容和语气各不相同的在线社区。数据集由 COMMENT 和 REPLY 对构建，其中 REPLY 将 COMMENT 中的特定引用跨度 (QUOTED) 定位为居高临下。该数据集包含 3,255 个正（居高临下）样本和 3,255 个负样本。
TalkDown is a labeled dataset for condescension detection in conversational contexts. This dataset is sourced from Reddit, an online discussion platform hosting diverse communities with varied content and tonal styles. It is constructed from COMMENT-REPLY pairs, where the REPLY identifies a specific quoted span (QUOTED) within the COMMENT as condescending. The dataset contains 3,255 positive (condescending) samples and 3,255 negative samples.
提供机构：
OpenDataLab
创建时间：
2022-05-23
