SMILE
收藏资源简介:
SMILE数据集是由韩国科学技术院创建的,专注于理解视频中笑声的多模态数据集。该数据集包含887个视频片段,每个片段都配有语言描述,解释人们为何发笑。数据集主要关注观众笑声,因为这种笑声通常信号清晰且代表性强。创建过程中,数据集使用了TED演讲和情景喜剧两种不同来源的视频,确保了数据多样性。SMILE数据集的应用领域广泛,旨在通过机器理解人类社交互动中的非语言信号,如笑声,从而推动人工智能在社交智能方面的进步。
The SMILE Dataset is a multimodal dataset developed by the Korea Advanced Institute of Science and Technology (KAIST) that focuses on understanding laughter in videos. It comprises 887 video clips, each paired with textual explanations for why people are laughing. The dataset primarily centers on audience laughter, as this type of laughter typically features clear signals and strong representativeness. During its development, videos from two distinct sources—TED Talks and situation comedies (sitcoms)—were utilized to ensure data diversity. The SMILE Dataset has a wide range of application scenarios, aiming to advance artificial intelligence in the field of social intelligence by enabling machines to comprehend non-verbal signals (such as laughter) in human social interactions.




