The Interaction Ontology: Low-Level Cue Processing in Real-Time Group Conversations

The Interaction Ontology is a software component that receives a stream of low-level information from audiovisual analysis in group-to-group videoconferencing situations and detects higher-level events on the real-time stream. Based on this higher-level information, the Orchestrator acts as a virtual director that decides which video streams are displayed for individual participants. The Interaction Ontology component comprises an OWL model defining simple and complex events as well as states of the system and its participants. Based on reasoning and Jena rules, state changes are determined and used to inform the Orchestrator. Regarding the performance and scalability evaluation of the OWL-based approach, we conclude that it performs good enough for our limited test scenario, but won't scale for the larger setups we envision.