DOD wants a ‘Big Mechanism’ for mining research data

DARPA announces a $45 million program that aims to revolutionize the way research on complex subjects is understood.

The Pentagon wants to revolutionize the way research data is mined, with a $45 million program it calls Big Mechanism.

The Defense Advanced Research Projects Agency has issued a solicitation calling for innovative, automated methods of essentially reading multiple research papers on complex subjects and finding causal connections and overarching patterns. And the solicitation emphasizes that DARPA is looking for a breakthrough — evolutionary improvements to existing practices are “specifically excluded.”

The notion of reading research papers might sound fairly simple, but DARPA notes that some of the areas thee Defense Department studies, such as ecosystems, economic systems or the human brain, are very complex. Studies of those systems are scattered and inconsistently shared, and as a result a proper understanding of them can be lacking. Hence the need for a Big Mechanism.

DARPA describes a Big Mechanism as an explanatory model “of complicated systems in which interactions have causal effects.” It will likely be built on a combination of technologies, including machine reading and natural language processing, ontology, various analytics and reasoning processes, visualization, simulation and statistical models, and some of those technologies will have to be improved. Machine reading, for example, will require “deeper semantics to represent the causal and often kinetic models described in research papers,” DARPA said.

The initial focus of the program is to study is cancer biology and signal pathways for cancer cells, but DARPA said the overall goal is “a new kind of science in which research is integrated more or less immediately – automatically or semi-automatically – into causal, explanatory models of unprecedented completeness and consistency.”  

The program is designed to last 42 months, broken up into three phases. Each phase will focus on the four aspects on a Big Mechanism—reading, assembly of fragments into a Big Mechanism, explanation and integration. Responses to the solicitation are due by March. 18.