Significant information encapsulation and valence exploitation (SIEVE) for discovery

In intelligence analysis environments, content such as entities, events and relationships appear in different source documents and contexts, and relating them is a challenging and intensive task. This paper presents an approach to reducing the volume and variety of the content by automatically associating them. The SIEVE architecture is built on the following technologies: (1) Backus-Naur Form (BNF) grammar structures to capture the possible relationships between people, places and organizations, (2) parsing structures to transform the relationships into numerical values, (3) relating these values to the analyst's model of interest or “initial shoebox” and the creation of information graphs, and (4) parsing the graphs and using semantic algorithms to link these graphs to external information in larger data repositories. A graph analytic approach for associating entities is presented in this paper.