Training Services
The training and tutoring domain is an excellent application area for the Knowledge Systems focus on effective use of knowledge. Our training work features the themes of reasoning from a pedagogical model and student model to individualized instruction, delivery of reasoning services within Web applications, and supplying the student the data they need at the time it will have the most impact.
Applications we have investigated include:
- Component-based training solutions
- Intelligent tutoring systems and adaptive, Wseb-based instruction
- On-the-job support, refresher and just-in-time training
- Simulated mentors and coaches, help desks, and product support
- Knowledge-based web guides and advisors
Representative projects
V-CTC - Virtual Combat Training Center
Sponsored by: DARPA DARWARS
The goal of this project was to provide technology
for creating real-time intelligent tutoring for
soldiers during training simulations, and a robust,
individualized after-action review when the simulated
mission is over. V-CTC utilizes an innovative hybrid
architecture so that the intelligent tutoring can
be integrated with existing simulations. This cost-effective
approach leverages the considerable existing investments
in training simulations.
This approach was demonstrated through integration with an existing commercial, high-fidelity tactical simulation of combined-arms warfare at the battalion and company called Armored Task Force.
PETA: Personal Education and Training Assistant
Sponsored by: Naval Air Warfare Training Systems
Division (NAWCTSD)
The goal of this project was to build a Web-based
intelligent agent system that can personalize interaction
with individual users and be extended to meet a
variety of distributed learning needs. PETA is knowledge-based,
can reason and employ strategies. We demonstrated
the foundational agent architecture and the extension
capabilities through a selected application and
domain.
Key to PETA's success was the development of a Web-based knowledge server that supports inferencing over the Web for large numbers of simultaneous users. It is this inference capability that enables agent functions such as planning, creating a user model, and reasoning for customized interaction with individual users.
DAS: Dialog Agent System
Sponsored by: Naval Air Warfare Training Systems
Division (NAWCTSD)
This project made advances in natural language (NL)
understanding capabilities for agents in distributed
learning applications. Ir demonstrated the NL capability
within a tutoring application. The tutor evaluated
trainees' NL answers to tutor questions. Trainees
are able to ask NL questions about the domain at
any time, and the tutor formulates appropriate responses
from its knowledge base.
Our NL approach used a controlled English grammar and limited vocabulary to increase reliability and accuracy of parsing user input. For trainee queries, the input is translated into a logic form that can query the knowledge base. Knowledge is identified, inferences are made, and a relevant response is formed into a grammatical NL sentence. Trainee answers to tutor questions are evaluated for equivalence of meaning to exemplar answers.
Intelligent Dialog Agent System
Sponsored by the U.S. Army Research Institute
The goal of this work was to develop authorable,
dialog-enabled intelligent agents for tutoring and
performance support systems. Users interact with
intelligent agents who carry out strategies and
goals and can engage in mixed-initiative dialog
via a natural language understanding and generation
system. Non-programmers can author new domains and
scenarios and create new conversational agents.