Semantic Parsing and role Labeling In Combination Effort (SPLICE)

Period of Performance: 09/21/2015 - 04/20/2016

$150K

Phase 1 STTR

Recipient Firm

Decisive Analytics Corp.
1400 Crystal Drive Array
Arlington, VA 22202
Firm POC
Principal Investigator

Research Institution

ICSI
1947 Center Street, Suite 600
Berkeley, CA 94704
Institution POC

Abstract

With missions surrounding counterproliferation, nonproliferation and WMD reduction, the stakes for DTRA are incredibly high. Effectively conducting this work involves cooperation and coordination across a wide range of military and government organizations throughout the United States and the world. The decentralization of threats has led to an explosion in the sources and types of data needed to identify, track, and mitigate these threats. Managing highly diverse information drawn from a wide range of agencies and open sources is exceptionally challenging. Even when this problem is restricted to unstructured text, there is simply too much data coming from too many sources for any analyst to appropriately handle all available information and effectively exploit the knowledge it contains with current NLP technologies. Under the Semantic Parsing and role Labeling In Combination Effort (SPLICE), DAC has partnered with the ICSI to build a Deep NLP capability that combines the strengths of Semantic Parsing with those of Semantic Role Labeling. The resulting Deep NLP capability will: 1) find links between entities, events and concepts that previously could only be found through manual analysis; and 2) support population of semantically structured knowledge bases with the confidence that all relevant available information can be fully exploited.