Multiple studies have demonstrated significant learning gains from intelligent tutoring systems. At the same time, training organizations are heavily invested in eLearning and schoolhouse materials. Moving from these more traditional and less effective instructional methods to personalized, interactive and more effective methods is desirable, but it is also impractical unless it can be done efficiently and cost effectively.

The research presented in this paper describes a process that can accomplish this goal by transforming existing digital learning content into cloud instances of interactive, dialogue-based intelligent tutoring systems. The process in question starts with a collection of instructional material and applies data mining techniques to identify key concepts. It then applies a different set of techniques to break the content into nuggets associated with those concepts and find an optimal learning path. In the resulting system, the nuggets are presented to learners as text and speech. Learners then answer questions about the content, and the system uses techniques from computational linguistics to analyze and adapt to learner responses. This paper outlines the process, describes the underlying technology and methods, and discusses lessons learned from prototyping the intelligent tutoring process with combat medic training.

Although the chief value of this paper to the community is in the transformational process it represents, the paper also reports several results concerning automated processing of text-based learning content that are of interest in their own right and that fit in with the modern trend of applying methods from "big data" to problems in other areas, including education and training.