OV-MERD
收藏资源简介:
OV-MERD数据集是由中国科学院自动化研究所等机构合作构建的开放词汇多模态情感识别数据集。该数据集包含248种情感类别,每个样本通常有2到4个标签,远超现有数据集的情感类别数量。数据集的创建过程结合了人类和大型语言模型(LLM)的协作标注策略,确保了标签的丰富性和准确性。OV-MERD数据集旨在解决传统情感识别方法中标签空间有限的问题,通过捕捉更广泛的情感表达,推动情感AI的发展,特别是在人机交互等应用领域。
The OV-MERD dataset is an open-vocabulary multimodal emotion recognition dataset jointly constructed by the Institute of Automation of the Chinese Academy of Sciences and other institutions. This dataset encompasses 248 emotion categories, with each sample typically carrying 2 to 4 labels, which far exceeds the number of emotion categories in existing datasets. The dataset's construction adopts a collaborative annotation strategy combining humans and large language models (LLMs), which ensures the richness and accuracy of the labels. The OV-MERD dataset aims to address the limited label space issue in traditional emotion recognition methods, and promote the development of affective AI by capturing a wider range of emotional expressions, especially in application scenarios such as human-computer interaction.




