Probabilistic Latent Semantic Analysis of Verbal and Behavioral Data in Dynamic Mission Operations

Assessment of communications in simulation based training suffers from (1) slow pace, (2) a reliance on communications data to the exclusion of behavioral and contextual data, (3) use of a single language model to handle multiple forms of language, and (4) failure to systematically address Mission Essential Competencies. We propose to build on our insights and success from Phase I to (1) develop a communications assessment system that generates assessments immediately, for use in After Action Review. We will (2) combine linguistic and behavioral data as input to an integrated system, and communications context as a cue to (3) choose between several communications models the one with the greatest accuracy. Finally, (4) we will apply this system to assess well-validated, Mission Essential Competencies in Distributed Mission Operations. In particular, we propose to develop a model of knowledge and skills associated with Air-to-Air and AWACS training based on validated MECs, evaluate it, and develop an assessment module that can be integrated into a training system such as the DMO. We call this system DECA2DE: Dynamic Embedded Communication & Action Assessment in Distributed Environments (see Figure 1).