A New Data-Set for Research on Audio Detection and Modeling of Social Micro-Contexts

We introduce the concept of social micro-contexts and their relation to modeling social situations and discuss the problem of detecting and modeling social micro-contexts. With respect to the validity of the problem, we argue that an interesting class of applications for such models exist and thus the problem is indeed a valid research problem. We introduce a novel data-set intended to allow for the development and validation of distributed collaborative approaches for the detection and modeling of social micro-contexts and social situations on the basis of audio-signals. For both concepts we present suitable definitions. We describe the social experiment that the data was collected in as well as the data-collection methods. Furthermore the annotation application, annotation process and annotation policies that were used are described. We conclude with a roadmap of planned further development activities.