As a result from the German SmartWeb project three speech corpora, one of them multimodal, have been published by the Bavarian Archive for Speech Signals (BAS). They contain speech and video signals from human-machine in-teractions in real indoor and outdoor environments. The scenarios for these cor-pora are a typicial handheld PDA interaction (SHC), an interaction on a running motorcycle (SMC) and another handheld PDA interaction in a human-human-machine setting with face video capture (SVC). The speech was transcribed using the Verbmobil annotation scheme extended by off-talk taggings while the video was labelled for on/off-focus and face boundaries. For all three scenarios the newly developed situational prompting technique SitPro was applied by which spontaneous interactions in different domains can be elicitated effectively. This chapter describes the prompting scheme SitPro, the recording technique used to obtain real situation data as well as the properties of the resulting corpora.
展开▼