Leveraging social media to generate narrative for virtual patient simulations

This work describes the research and development of semi-automated, user-supervised narrative generation for virtual patient (VP) simulators. We outline the system architecture required of such a system, and propose leveraging data from the health-related content of social networking websites (specifically, Facebook, PatientsLikeMe, and Inspire), in addition to electronic medical record (EMR) datasets. Our research focuses on four key areas as we work toward finalizing our system design: 1) Exploring the utilization of the Open Biomedical Ontologies and other natural language processing tools to facilitate concept identification, synonym generation, and knowledge base construction; 2) Designing templates that structure the presentation of narrative content according to author-selected parameters that serve as queries into the knowledge base; 3) Comparing various user interfaces to best support the author’s interaction with the plot graph and the logical design of narrative cases; 4) Piloting protocol for evaluating the quality of simulation narratives and its influence on simulation fidelity