Exploiting Latent Semantics in Adaptive Information Warfare Training Systems

We propose a multifaceted approach to improving Information Warfare training that exploits recent advances in intelligent tutoring technology and natural language processing. The focus of our efforts will be to develop technology that can produce highlyaccurate assessments of individual student knowledge/skills within the context of team training. By exploiting diverse evidence, the resulting tools will allow student progress tracking, course content/structure evaluation, instructional interventionplanning, and real-time coaching. The proposed techniques will combine evidence gathered by monitoring student task performance with that extracted from an automated evaluation of the information products (e.g., team communications, RFIs, threatassessments) developed in the course of an exercise. This will be made possible by the types of quantitative models found in voice recognition and latent semantic analysis (LSA). In Phase I we verified the benefits (and necessity) of exploiting thisadditional evidence to reduce the ambiguity in student models. We further demonstrated how new technologies can enable the fielding of a completely instructor/SME supportable scenario-based training system that does not require programming. These Phase Isuccesses set the stage for the complete implementation/deployment of our training technology in Phase II, which will in turn open a multitude of Phase III opportunities.