WITS
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/lcs2-iiitd/maf
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为WITS，包含了来自印度电视剧的多模态、多方参与、语码混合的讽刺性对话。每个讽刺实例都附有相应的自然语言语码混合解释。文本输入采用罗马字编码混合格式，包含了讽刺、幽默和情绪的标注。此外，该数据集还经过了增强，创建了如sWITS、hWITS和eWITS等变体，以适应不同的任务需求，如讽刺解释、讽刺检测、幽默识别和情绪识别。
The dataset named WITS comprises multimodal, multi-party, code-mixed sarcastic dialogues derived from Indian television dramas. Each sarcastic instance is paired with corresponding natural language code-mixed explanations. The text inputs adopt a mixed romanized encoding format, with annotations covering sarcasm, humor and emotion. Additionally, this dataset has been augmented, and variants including sWITS, hWITS and eWITS have been developed to suit diverse task demands, such as sarcasm explanation, sarcasm detection, humor recognition and emotion recognition.
