MMChat中文社交媒体多模态聊天多轮对话数据集

<p>MMChat是一个大规模社交媒体多模态多轮对话数据集，包含与图片相关联的中文对话，涵盖丰富的情感与社交场景。数据集通过多种策略保障对话质量，并提供图片与对话相关性的人工标注。其规模达数十万对话对，具有多模态、多轮交互和高质量语义关联等特点，适用于多模态对话系统、图文联合理解、情感分析等任务，推动社交场景下人机交互技术的研究与发展。</p>

MMChat is a large-scale multi-modal multi-turn dialogue dataset collected from social media, which contains Chinese dialogues paired with images, covering a wide range of emotional and social scenarios. The dataset adopts multiple strategies to guarantee the quality of dialogues, and provides manual annotations for the relevance between images and their associated dialogues. Boasting a scale of hundreds of thousands of dialogue pairs, it features multi-modality, multi-turn interaction and high-quality semantic association, and is applicable to tasks including multi-modal dialogue systems, joint image-text understanding, sentiment analysis and so on, thus promoting the research and development of human-computer interaction technologies in social scenarios.