TEIDAN多模态多方对话语料库
收藏资源简介:
TEIDAN多模态多方对话语料库由京都大学信息学研究科创建,旨在支持多方对话系统的研究。该语料库包含30个三人自由讨论的会话,每个会话持续约5-10分钟,讨论主题包括日本替代首都、荒岛生存必需品以及周末旅行目的地。语料库的标注部分包含受话者信息,揭示了约20%的对话轮次中明确指定了受话者。该语料库的应用领域主要集中在多方对话系统的开发,特别是受话者识别任务,旨在解决多方对话中复杂的受话者识别问题。
The TEIDAN Multimodal Multi-Party Dialogue Corpus was developed by the Graduate School of Informatics at Kyoto University to support research on multi-party dialogue systems. This corpus comprises 30 free-form three-party discussion sessions, each lasting approximately 5 to 10 minutes, with topics covering alternative capitals for Japan, essential supplies for desert island survival, and weekend travel destinations. The annotated subset of the corpus includes addressee information, revealing that addressees are explicitly specified in roughly 20% of dialogue turns. The primary application scope of this corpus is the development of multi-party dialogue systems, particularly the addressee recognition task, aiming to address the complex addressee recognition challenges in multi-party dialogues.




