Chimera
收藏资源简介:
Chimera是一个基于多模态情感分类任务的数据集,由华东师范大学、新加坡南洋理工大学和新加坡国立大学合作创建。该数据集旨在通过结合文本和图像信息,捕捉特定目标或方面的情感极性,支持多模态内容分析。数据集整合了细粒度的视觉特征和文本描述,利用大型语言模型生成的语义和情感理由来增强模型对情感线索的理解。目前,该数据集已公开,可通过GitHub访问,但具体的数据集条数未在文中提及。
Chimera is a multimodal sentiment classification dataset co-developed by East China Normal University, Nanyang Technological University (Singapore), and National University of Singapore. This dataset is designed to capture the sentiment polarity of specific targets or aspects by integrating textual and visual information, thereby supporting multimodal content analysis. It integrates fine-grained visual features and textual descriptions, and utilizes semantic and affective rationales generated by large language models to enhance models' understanding of emotional cues. Currently, this dataset has been publicly released and is accessible via GitHub; however, the exact number of samples included in the dataset is not specified in the available literature.

- 1Exploring Cognitive and Aesthetic Causality for Multimodal Aspect-Based Sentiment Analysis华东师范大学计算机科学与技术学院, 新加坡南洋理工大学计算机学院, 新加坡国立大学公共卫生学院 · 2025年



