SOAR TECHNOLOGY, LLC — Department of Defense SBIR Phase I: N202-126

Description

Over the last half decade, deliberate misuse and manipulation of social networks to promote and to amplify false narratives has become a national security issue. Information Operations teams, established to counter such threats, lack tools to allow them to explore, develop, and train Tactics, Techniques, and Procedures (TTPs). A specific and acute gap is the ability to create realistic and valid examples with which to develop and to test TTPs. We propose to develop tools and methods that will allow generation of synthetic but realistic text content at scale in support of specific experimentation and training usage goals. Transformer models can be adapted for content generation to support large-scale synthesis and form the foundation of the proposed work. A  user will be able to “seed” topics and to review, edit, and tag generated text. Accepted text will be packaged to allow use (deployment) in a synthetic environment. A Phase I proof-of-concept will demonstrate the feasibility of realistic text generation and placement into the context of scenarios and relevant use cases.