R3: Reading, Reasoning, and Reporting
R3 uses deep natural language understanding to read academic journal papers and extract information, and then cross-references this information against large scientific databases to identify contributions and update its models. At present, R3 reads academic journal articles on Ras-based cancers, and it identifies corroborating, conflicting, and novel information across articles in less than five seconds per article. R3 advances the state-of-the-art in automated systems that utilize and extend scientific models with deep natural language understanding.
SIFT leads the R3 project for DARPA's Big Mechanism program, teaming with the University of Colorado Denver Medical School, Brandeis University, and Tufts University.
A fact sheet is available for download: [pdf]