Enhancing Real Time Situational Awareness with Latent Relationship Discovery

ABSTRACT: The proposed work, URIEL (Unified Representation and Inference for Exploitation of Links), aims to simplify the work of intelligence analysts in discovering and exploring latent relationships from multiple sources of massive amounts of information. URIEL will support 1) automated extraction, transformation, and load at scale from multiple data sources, including document archives, social media, still imagery, and video, 2) integration of the resulting graph-structured content into a scalable graph database, and 3) invocation of latent relationship discovery algorithms on user-specified subsets of this content and visualization of the results. Key strengths of our approach include:1) Tractable information integration by alignment to a simple common schema that specifies only a small number of entity types (e.g., people, places, things, groups); 2) Incorporation of probabilistic methods for discovery of latent relations specifically designed to accommodate noisy inputs from error-prone ETL engines and not necessarily conforming to a curated relation schema; and 3) A focus on scaling to high streaming data rates (e.g., thousands of social media posts per second) and large overall data volumes (tens of terabytes of relational data, potential scalability to petabytes for imagery and video).; BENEFIT: We believe that URIEL has the potential to offer intelligence analysts in many organizations new, useful capabilities for finding connections in diverse multi-source data, and is also has private sector applications to business intelligence.Our primary target for initial commercialization is transition to the National Air and Space Intelligence Center (NASIC), however we are also pursuing multiple other commercialization avenues, including transition via AFRLs E2AT program to PCPAD-X.