SentiMetrix has established a world-class, cross-disciplinary team to research and develop new capabilities in COSMIC: Cloud-Oriented Social Media Inference & Counteraction. COSMIC will combine novel algorithms from sentiment analysis, probabilistic temporal learning of diffusion of messages/opinion in social networks, forecasting and prediction of the reach of messages in social media, and game-theoretic methods to counteract diffusion of messages, into a highly modular, loosely-coupled framework and software platform to address the objectives of DARPA’s Social Media in Strategic Communication (SMISC) Program. Specifically, COSMIC will:
- Track and predict information flow within and across social media sources to understand what is being said, who is saying it, and what the impacts may be from the presence or absence of information in this structure. COSMIC will enhance the exploitation of any single data source, and maximize the combined value of multiple sources for situation awareness in a highly dynamic and rapidly evolving social communication environment.
- Detect emergence of patterns, behaviors, and events of interest, providing broad situation awareness without requiring pre-specification of group behaviors or topics of interest. COSMIC will use topic-independent behavior discovery to recognize activities and behaviors of interest, and then studies how context affects opinion on selected entities.
- Enable analysts to use social media both strategically and tactically to identify social influencers and potentially damaging campaigns, and develop effective, directed messaging or counter messaging. COSMIC will identify the best places to interact with the population via social media, calibrate the impact of doing so, and empower analysts to make deeply informed decisions about how to use social media most effectively even in the presence of an intelligent adversary who is capable of putting himself into our shoes and performing similar reasoning about what we might do.
RPI shall execute the following technical tasks during the contracting period:
- RPI shall define, in collaboration with Sentimetrix and other subcontactors designated by them as term members, appropriate methods, techniques and representations to support the Objectives of the proposed effort, focused in the areas of:
- Development of a Parallel RDF store with scalable semantic inferencing to provide cloud services for social media analysis
- Collection and semantic categorization of real-time social media metadata
- Visualization and inferencing technologies for use by the COSMIC team
- RPI will provide algorithms and prototype software to show functionality of the appropriate methods, techniques and representations listed above. RPI will continue to devise and implement new features with guidance from Sentimetrix. RPI will also provide development guidance to Sentimetrix and other team developers as needed so that they can make the COSMIC platform more robust as needed for experimentation, evaluation, and possibly transition to government prototype systems.