SBIR-STTR Award

The high operations tempo, frequent deployments and other fact-of-life events limit the availability for team training. To mitigate this training availability problem, training simulations need to incorporate constructive entities as teammates in a way that enhances training for the human participants. In modern operational environments text messaging communications are increasingly common. To be effective, synthetic teammates will need to be capable of text messaging with their human teammates. To address the limitations of the language understanding technologies, the Air Force has conducted research and development in natural language understanding capabilities in its synthetic teammate program. Our team proposes to design and develop a natural language interface toolkit, SynChat, that will leverage prior Air Force investments and will develop and integrate other needed technologies. We see a significant opportunity to advance the state of the art in the integration and application of text-based communications for command and control training environments. Much of the challenge of integrating language capabilities into new domains in training systems is both the development of domain specific knowledge that is usable by the constructive entities and extending domain general reasoning capabilities for the new domain. Our solution focuses on these key challenges.

Benefit:
With the SynChat natural language toolkit, the U.S. Air Force will have the capability to develop and integrate natural language capabilities into training systems in a way that will be more affordable than today""s development methods. More affordable development will permit the Air Force to apply these capabilities across a broader set of operational domains and allow more of our warfighters to benefit from these new training capabilities. The broader application of language capable constructive models will help address the training availability problem and will improve mission and combat readiness. The SynChat natural language toolkit will provide similar efficiency improvements for Army, Navy, Marine, and other non-Department of Defense government agencies.

The ability to access high-quality training anywhere and anytime continues to grow in importance due to the complex nature of operational team environments. Training simulations that use virtual agents that respond and behave like live teammates enhance training and help achieve the “train as you fight” objective without requiring a complete team of human participants. However, the domain-specific language and reasoning ability needed to make a virtual agent effective for training is challenging to construct. A solution is needed to simplify and accelerate the construction of natural language capable agents for training systems. Such a solution would help satisfy the overarching goal to improve the cost effectiveness of achieving the “train as you fight” objective when some team members are not available. TiER1 Performance Solutions is leading the SynChat: Synthetic Language Interface Toolkit for Chat research and development project. The goal is to develop a web-based suite of tools to support rapid development of agent models, and to define and implement a flexible communications interface built upon pervasive open protocols and deployed as a set of web services. This toolkit will simplify and accelerate the work of training system developers in integrating advanced language capabilities into training systems.

Benefit:
With the SynChat natural language toolkit, the U.S. Air Force will have the capability to develop and integrate natural language capabilities into training systems in a way that will be more affordable than today’s development methods. More affordable development will permit the Air Force to apply these capabilities across a broader set of operational domains and allow more warfighters to benefit from these new training capabilities. The broader application of language capable constructive models will help address the training availability problem and will improve mission and combat readiness. The SynChat natural language toolkit will provide similar efficiency improvements for Army, Navy, Marine, and other non-Department of Defense government agencies.

Keywords:
Natural language, knowledge engineering, text messaging, text chat, language capable constructive entity, synthetic teammate, rapid knowledge elicitation, team training