VivesDebate-Speech

VivesDebate-Speech是由邓迪大学和瓦伦西亚理工大学合作创建的一个口语辩论数据集，旨在利用音频特征进行论据挖掘。该数据集包含29场专业辩论的音频和文本数据，每场辩论都进行了全面的标注，以便捕捉自然语言论据单元（ADUs）之间的长距离依赖关系。数据集的创建过程涉及将原有的VivesDebate数据集扩展，增加了音频格式的辩论内容，并创建了BIO标签文件以支持自动识别ADUs。VivesDebate-Speech数据集的应用领域主要集中在自然语言处理中的论据挖掘，特别是通过音频特征改善论据分析的准确性和效率。

VivesDebate-Speech is a spoken debate dataset jointly created by the University of Dundee and Universitat Politècnica de València, aiming to leverage audio features for argument mining. This dataset contains audio and text data from 29 professional debates, with comprehensive annotations for each debate to capture long-distance dependencies between Argumentative Discourse Units (ADUs) in natural language. The development of VivesDebate-Speech involved expanding the original VivesDebate dataset by adding audio-format debate content, as well as creating BIO tag files to support automatic ADU recognition. The primary application fields of VivesDebate-Speech focus on argument mining in natural language processing, particularly improving the accuracy and efficiency of argument analysis via audio features.