Hybrid Dialog Tutor

We propose a - g a very large breadth of material and mixed-initiative. In-depth dialog about specific complex concepts. The hybrid approach was driven by the desire to provide the Army with a method to rapidly convert large volumes of instructional material supported by tutorial dialog. This method uses Latent Semantic Analysis (LSA) which statistically reduces material into a very compact knowledge representation. LSA will support normal tutorial dialog, such as. evaluating student responses and answering factbased questions. We provide the mixed-initiative in-depth dialog through an NLP approach. We have an easy to use template-driven authoring tool for the conceptual knowledge base. Continuous speech recognition is provided, as well as a talking head tutor with realistic facial animation that is synchronized with a text-to-speech synthesis system. Every component is already operational and the concept for integration is described in this proposal. The final proposed tutor is available to the government for free distribution and requires no separate purchase agreements for any of the components. The proposed hybrid tutor can be delivered over the Internet and will have a reasonable footprint (memory, storage) and speed on a Pentium in Windows NT. BENEFITS: The benefit to the Army is that large volumes of instructional material can be rapidly converted into a realistic, simulated learning environment delivered on a PC or over the Internet for self-development and just-in-time training. All of the components are easily extensible for foreign language training.Offering a realistic task environment in which the trainee can interact naturally through speech with a simulated tutor along with an authoring capability provides this product an edge over existing commercial computer-based training and assessment products.