Spot the Difference Corpus

Spot the Difference Corpus是由瑞典皇家理工学院创建的一个多模态数据集，包含54组任务导向的自发口语交互记录。数据集内容丰富，包括音频、视频、转录和多种注释，旨在研究人类沟通的多维度，如轮流发言和指称表达。创建过程中，参与者通过描述和比较两幅相似场景图来找出差异，从而产生数据。该数据集适用于研究语言、非语言行为及其在时间上的演变，特别是在任务导向对话中的应用，有助于开发更人性化的口语对话系统。

The Spot the Difference Corpus is a multimodal dataset developed by the Royal Institute of Technology, Sweden, consisting of 54 recordings of task-oriented spontaneous spoken interactions. The dataset includes diverse modalities such as audio, video, transcripts, and multiple annotations, with the goal of exploring multi-dimensional facets of human communication, including turn-taking and referential expressions. Data in this corpus was generated by having participants describe and compare two similar scene images to identify differences. This dataset is applicable to research on language, non-verbal behaviors, and their temporal evolution, particularly in task-oriented dialogues, and contributes to the development of more human-centric spoken dialogue systems.