Generalized Ontology Discovery Enabling Semantic Search (GODESS)

A central challenge in the intelligence community is managing and effectively integrating large amounts of disparate information sources for concise presentation of knowledge to an analyst. Currently, the high volume of near-constant incoming intelligence imposes a substantial burden on the analyst to review and digest the raw intelligence causing analyst overload and fatigue, which could lead to missed intelligence. Generalized Ontology Discovery Enabling Semantic Search (GODESS) addresses this urgent need by analyzing the data"s structure and the analyst"s information needs to produce an attributed semantic network extracted from disparate, multi-source document stores containing incomplete, inconsistent and sparse data. At the core of GODESS is the Infinite Relational Model (IRM; Kemp, 2010), a Machine Learning Algorithm that automatically discovers systems of related concepts and entities structured in the data to learn ontologies that are optimal for representing the knowledge encapsulated in the database. GODESS relearns the ontologies as new data sources become available, making the approach tractable for environments with continuous, high-volume information flow. The emergent ontology effectively"connects the dots"for analysts, providing them with a complete picture of ambiguous entities and hidden relationships appearing in the datastore, thereby increasing the relevance of semantic searches performed on the database.