Media Bias Identification Benchmark (MBIB)
收藏资源简介:
Media Bias Identification Benchmark (MBIB) 是由康斯坦茨大学的Martin Wessel等人创建的第一个媒体偏见识别基准任务和数据集集合。该数据集包含22个子数据集,涵盖了语言、认知、政治等多种类型的媒体偏见。MBIB旨在通过一个统一的框架测试潜在的偏见检测技术如何泛化。数据集通过广泛的文献搜索,从115个相关数据集中精选而出。MBIB的应用领域广泛,旨在解决媒体偏见检测中的多任务问题,推动开发更强大的系统,并促进媒体偏见检测评估的范式转变,使其能够同时处理多种媒体偏见类型。
Media Bias Identification Benchmark (MBIB) is the first media bias identification benchmark task and dataset collection developed by Martin Wessel et al. from the University of Konstanz. This dataset comprises 22 sub-datasets covering diverse types of media bias, including linguistic, cognitive, political and other categories. MBIB aims to evaluate the generalization performance of potential bias detection technologies via a unified framework. The dataset was curated from 115 relevant datasets through extensive literature searches. Boasting a wide range of application scenarios, MBIB is designed to address multi-task problems in media bias detection, promote the development of more robust detection systems, and facilitate a paradigm shift in media bias detection evaluation, enabling concurrent handling of multiple media bias types.

- 1Introducing MBIB -- the first Media Bias Identification Benchmark Task and Dataset Collection康斯坦茨大学 · 2023年



