iNews
收藏资源简介:
iNews是一个大规模的多模态数据集,由剑桥大学创建,专门捕捉个人对新闻标题的主观情感反应。该数据集包含来自291名英国不同地理区域的参与者在Facebook上对2899个主要英国媒体新闻帖的注释,注释内容涵盖情感维度评分、离散情绪分类、内容相关性判断、分享可能性以及模态重要性评分等。此外,还收集了全面的标注者特征信息,这些信息能够解释15.2%的注释方差,超过了现有的NLP数据集。iNews数据集将有助于推动大型语言模型个性化、主观性、情感计算和个人行为模拟等领域的研究。
iNews is a large-scale multimodal dataset created by the University of Cambridge, specifically designed to capture individuals' subjective emotional responses to news headlines. This dataset contains annotations from 291 participants across different geographical regions in the UK on 2899 news posts from major UK media outlets on Facebook. The annotations cover emotional dimension scoring, discrete emotion classification, content relevance judgment, sharing likelihood, and modality importance scoring, among others. In addition, comprehensive annotator characteristic information is collected, which can explain 15.2% of the annotation variance, outperforming existing NLP datasets. The iNews dataset will facilitate research in fields such as large language model (LLM) personalization, subjectivity, affective computing, and individual behavior simulation.

- 1iNews: A Multimodal Dataset for Modeling Personalized Affective Responses to News剑桥大学 · 2025年



