The rate of scientific discovery continues to accelerate. For now, this knowledge is most effectively presented in academic papers, which provide a detailed overview of the methods used to make the discovery, a contextualization of the discovery in terms of known theories and previous results, and a discussion of the potential impact of the discovery. However, it is impossible for any individual to read the thousands or even tens of thousands of scientific articles that are published yearly even for a single research domain. This talk presents a series of recent visualization projects, developed as part of DARPA’s Big Mechanism program, that bring together fragmented domain knowledge about the complex functionality of biological pathways, normally scattered across journal publications and conference proceedings. Rather than organizing knowledge around keywords or authors, the Big Mechanism project extracts "knowledge fragments" and assembles them into an interactive knowledge network. While the visualization projects— Dynamic Influence Networks, ReactionFlow, and BranchingSets, and others— each emphasize different analysis tasks, the underlying approach makes makes it possible: to see how the understanding of a particular concept evolves over time; to identify which concepts are confirmed by subsequent experiments or to make sense of why new investigations contradict our earlier understanding; and to effectively reason about the contexts in which specific knowledge is applicable.
Posted by: Nathan Galli