Carnegie Hall: An Intelligent Tutor for Command-Reasoning Practice Based on Latent Semantic Analysis.

摘要

Report developed under a Small Business Innovation Research Program 99.2 contract for topic OSDOO-CRO2. Scenario-based training techniques, e.g., U. S. Army Research Institute for the Behavioral and Social Sciences' (ARI's) 'Think Like a Commander,' exercise command cognitive readiness skills. These techniques currently depend on discussion with live mentors. Phase I demonstrated that such scenarios could be taught using a web-based interactive facilitator/mentor. The web-based facilitator asks questions relevant to one scenario, and students write short text responses. Using Latent Semantic Analysis' (LSA) understanding of natural language, the intelligent mentor/ facilitator analyzes the essay's content and determines the student's weak areas for further questioning. The LSA-based prototype was constructed rapidly and greatly benefited from automatically training the system on a large amount of military text. It did not require the handcrafted knowledge models and rule- bases of conventional intelligent tutors.