Recognizing team context during simulated missions.

ABSTRACTWe investigated ways to automatically analyze movement and verbal behavior of teams of soldiers engaged in simulated military missions. Analysis of location data revealed that soldiers' locations were consistent with 3 stationary patterns and 2 movement patterns. Analysis of their dialogue detected statistical regularities. An automated text classifier was developed that employed these regularities to code dialogue utterances. These analyses demonstrate the feasibility of interpreting collaborative activity automatically and constructing models of team context for teams of soldiers engaged in trained activities.