Discussion Tracker Corpus

Discussion Tracker Corpus是由匹兹堡大学创建的一个注释数据集，专注于美国高中英语课堂中的口语多党讨论转录。该数据集包含29个讨论，总计985分钟的音频转录，并针对协作论证的三个维度（论证动作、特异性和协作）进行了注释。数据集的创建旨在解决缺乏包含多个同时注释的多党同步论证数据集的问题。该数据集不仅为NLP研究提供了丰富的数据资源，还通过提供性能基准和相关代码，支持了多任务学习等研究方法的发展，为NLP领域的进一步研究开辟了新的可能性。

Discussion Tracker Corpus is an annotated dataset developed by the University of Pittsburgh, focusing on transcribed multiparty spoken discussions in American high school English classrooms. The dataset comprises 29 discussions totaling 985 minutes of audio transcripts, annotated along three core dimensions of collaborative argumentation: argumentative moves, specificity, and collaboration. This dataset was created to address the gap in multiparty synchronous argumentation datasets with multiple concurrent annotations. It not only provides a rich data resource for NLP research, but also supports the advancement of research methodologies such as multi-task learning by offering performance benchmarks and associated code, opening up new possibilities for further research in the NLP domain.