CONAN (COunter NArratives through Nichesourcing)

COunter NArratives through Nichesourcing (CONAN) 是一个数据集，由 3 种语言的 4,078 对组成。此外，还提供了 3 种类型的元数据：专家人口统计、仇恨言论子主题和反叙述类型。该数据集通过翻译（从意大利语/法语到英语）和释义进行了扩充，使对的总数达到 14.988。

The Counter Narratives through Nichesourcing (CONAN) dataset consists of 4,078 pairs across three languages. Additionally, three types of metadata are provided: expert demographics, hate speech subtopics, and counter-narrative types. This dataset was augmented via translation (from Italian/French to English) and paraphrasing, bringing the total number of pairs to 14,988.

搜集汇总

背景与挑战

背景概述

CONAN 是一个多语言数据集，包含4,078对数据，涉及三种语言的仇恨言论反叙述内容，并提供了专家人口统计、仇恨言论子主题和反叙述类型三种元数据。通过翻译和释义扩充后，总对数达到14,988，由特伦托大学和凯斯勒布鲁诺基金会于2019年发布。