MUSE
收藏资源简介:
MUSE是由东北大学计算机科学与工程学院开发的首个多模态对话推荐数据集,专注于服装领域。该数据集包含7000个对话,共计83,148条话语,涵盖了丰富的多模态交互和自然对话。数据通过多代理框架自动合成,结合了多模态大语言模型(MLLMs)的力量,创新地从真实场景中生成用户画像,而非依赖人工设计或历史数据。数据集的应用领域主要集中在多模态对话推荐系统,旨在解决传统文本推荐系统在模拟真实购物行为中的不足,特别是在视觉驱动的领域如服装推荐中,提供了更全面的多感官决策支持。
MUSE is the first multimodal conversational recommendation dataset developed by the School of Computer Science and Engineering, Northeastern University, focusing on the fashion domain. This dataset includes 7,000 conversations with a total of 83,148 utterances, covering rich multimodal interactions and natural dialogues. The data is automatically synthesized through a multi-agent framework, leveraging the power of multimodal large language models (MLLMs) to innovatively generate user personas from real-world scenarios, rather than relying on manually designed resources or historical data. The dataset is primarily applied in multimodal conversational recommendation systems, aiming to address the limitations of traditional text-based recommendation systems in simulating real-world shopping behaviors, especially in vision-driven fields such as clothing recommendation, by providing more comprehensive multi-sensory decision-making support.




