Spontal: A Swedish Spontaneous Dialogue Corpus of Audio, Video and Motion Capture

Abstract

Speech was conceived in a face-to-face interaction setting, and spoken dialogue is the cradle in which it evolved. As of this day, every-day face-to-face communicative interaction is the context in which most of our language use occurs. We present the Spontal database of spontaneous Swedish dialogues, a new resource for studies of everyday face-to-face communicative interaction. 120 dialogues of at least 30 minutes each have been richly captured on four channels of high-quality audio, two channels of high-resolution video and on a motion capture system tracking torso, arm, hand and head movements. Within the project, additional recordings have been made using a similar setup, for example several recordings in other languages, such as English, Italian, Norwegian and Arabic, as well as several recordings with slightly different setups. Currently, before it is completed and released, already this massively multimodal corpus constitutes the foundation for several national research projects. The corpus is undergoing processing, validation and annotation, and will be made available for research at the end of the project. 1.