Vocabulary Intelligence, Surveillance and Reconnaissance (V-ISR)

The Army has recognized that its intelligence analysts are being swamped by the tremendous amounts of unstructured textual data they must try to manually process. While a number of systems have been developed to automatically extract information from text, these systems have suffered from the problem of requiring analysts to manually define and maintain lists of terms and phrases. What is needed is a tool that can automatically identify new terms and categorize them (person, terrorist, location, weapon, etc.) so that analysts can focus on the job at hand. The Vocabulary – Intelligence, Surveillance and Reconnaissance (V-ISR) Phase II project directly solves this problem using advanced machine learning techniques to recognize new vocabulary terms and categorize them within a semantic model called an ontology. The results of our Phase I project clearly demonstrated that new ontology terms and the contexts that denote them could be ‘harvested’ from realistic military reports in a timely manner. Modus Operandi proposes to build on this success by constructing a robust vocabulary mining capability that will be transitioned to Army programs of record including the Distributed Common Ground System – Army (DCGS-A).