Dawn Lim 3:32 PM ET
The Pentagon is launching a four-and-a-half-year research funding effort to build computer software that can analyze intelligence data and “see through language” to tease out hidden meanings in texts, cutting the time for analysts to file intel reports.
The military’s venture capital arm, the Defense Advanced Research Projects Agency, issued a request for proposals May 22 enlisting scientists for developmental work. The goal is to create technology that picks up on nuances that aren’t easily inferred from the mounds of intelligence data that analysts are fed each day. The experimental program, called Deep Exploration and Filtering of Text, was first reported by Nextgov earlier in May.
“Automated, deep natural-language understanding technology may hold a solution for efficient processing of text information and for understanding connections in text that might not be readily apparent to humans,” the solicitation reads.
DARPA envisions creating “an end-to-end system that will assist the warfighter and intelligence analyst at every stage -- from searching through data to identifying threats and then formulating analyses.” Because of the range of systems used in varied intel missions, DARPA wants an open-source platform that can be easily tailored for users, the broad agency announcement notes.
The funding program will be broken up into three 18-month phases. In the first, performers will build algorithms for the software. In the second, researchers will integrate the formulas into a computer program.
In the final phase, the scientists will have to demonstrate how the software works in deciphering and annotating unclassified and classified data. They also will have to show that their technology can be applied to at least one foreign language from a list that includes dialectal Arabic, Cebuano, Chinese, Pashto, Dari, Farsi, Ilocano, Spanish, Tagalog and Urdu.
Funding for individual awards has not been determined and will depend on the quality of the proposals received, the RFP notes.
(Image via Cleo /Shutterstock.com)