Multimodal Hyperbole Detection Dataset

本数据集名为‘Multimodal Hyperbole Detection Dataset’，由北京大学王选计算机技术研究所创建，旨在支持多模态夸张检测的研究。数据集包含2160条来自微博的帖子，每条帖子均包含文本和图像两种模态信息。创建过程中，研究团队从微博自动抓取约10000条帖子，并经过精细筛选和标注，确保数据质量。该数据集主要用于研究文本和图像如何共同表达夸张，以及评估不同模型在跨领域情况下的表现。应用领域包括社交媒体分析、情感计算等，旨在解决多模态信息处理中的夸张表达识别问题。

This dataset is named 'Multimodal Hyperbole Detection Dataset', developed by the Wangxuan Institute of Computer Technology at Peking University, with the core purpose of supporting research on multimodal hyperbole detection. It comprises 2160 Weibo posts, each integrating both textual and visual modal data. During the dataset construction phase, the research team first automatically crawled approximately 10,000 posts from Weibo, then performed rigorous screening and manual annotation to guarantee high data quality. This dataset is primarily employed to investigate how textual and visual modalities jointly convey hyperbolic expressions, as well as to evaluate the performance of various models across diverse domains. Its applicable scenarios cover social media analysis, affective computing and other related fields, aiming to address the challenge of hyperbole recognition in multimodal information processing.