A Fast Track Approach Towards Automatic Detection of Gray Zone Activities from Text

Abstract

Gray Zone Operations (GZ Ops) are increasingly being used by countries to achieve national security objectives that have historically only been achieved through armed conflict. GZ Ops employ actions that are diverse, covert, and adversarial but fall below the threshold required to trigger an aggressive response from major powers. We aim to help analysts with early identification of Gray Zone Operations by automatically detecting taxonomy-backed Gray Zone activities in news stories and text. We map our proven geo-political event extraction and data processing capabilities previously developed under the World-Wide Integrated Crisis Early Warning System (W-ICEWS) program to Gray Zone activities and apply these mappings to text from a corpus of Russian news stories from 2015 to 2016 to generate a set of sentences labeled with specific Gray Zone activities. We evaluate the validity of these matches to determine if the sentences and stories accurately reflect their label.