Social Language for Synthetic Teammates (SLST)

ABSTRACT: The Air Force is increasingly using text-chat based communications in operational environments, such as operations using Remotely Piloted Aircraft (RPA) and in Air Operations Centers (AOC). Training Airmen for these operations using simulations requires synthetic teammates that understand and use text-chat in a realistic, human-like manner, exhibiting both the strengths and weaknesses of real humans. To address this need, we propose to develop Social Language for Synthetic Teammates (SLST) (pronounced"Celeste"), a set of technologies and tools that enable the rapid and affordable creation of synthetic teammates that understand and generate socio-linguistically realistic text-based language. BENEFIT: We expect the full-scope Social Language for Synthetic Teammates (SLST) to have immediate and tangible benefit to military training programs using simulation and synthetic teammates in domains where text-chat is used. The improved ability to generate realistic chat-text based on social context will both make training more realistic, but also enable new members of the community to efficiently learn the jargon and communications conventions of their new domain. With the SLST technology, Airmen will be better prepared to meet real-world operational challenges. We will enhance Persona, our commercial product for agent authoring, to use more socio-linguistically realistic language, broadening Persona"s commercial appeal and enabling us to serve a wider customer base in the simulation and virtual training markets.