Disambiguation of Entity Association Statements Navy SBIR FY2010.2 Sol No.: Navy SBIR FY2010.2 Topic No.: N102-176 Topic Title: Disambiguation of Entity Association Statements Proposal No.: N102-176-1050 Firm: SEMANDEX NETWORKS, Inc 5 Independence Way Suite 309 Princeton, New Jersey 08540 Contact: Dave Ihrie Phone: (301) 233-4780 Web Site: www.semandex.net Abstract: Existing techniques for level 1 fusion of association statements in large RDF data stores have proven inadequate to the challenge of entity disambiguation from multiple intelligence sources. The resulting entity and association uncertainty causes problems of missed associations, redundant RDF statements that limit scalability, and significant limitations on higher level reasoning algorithms. Numerous researchers have established the need for higher level context in addressing this problem, and while necessary, context alone is not sufficient. We propose to leverage an existing software-based characteristic matcher operating in a context-aware framework by using a simulated annealing algorithm to support level 1 fusion of a large RDF data store. We will develop a capability whose goal is to generate a single connected graph from large RDF that contains no redundant entities and no missed connections. Specifically, the algorithm we propose will address (i) entity uncertainty, (ii) entity information from different knowledge bases that results in a contradiction, (iii) creation of statements regarding an entity in a knowledge base or common feature space that do not contradict existing statements on that entity, and (iv) allows for the deletion of an entity or entity statements without breaking other associations that may refer to that entity. The result will be an algorithm that integrates and connects new RDF-expressed sources while preserving the original data and semantics, allowing the expression of a large data corpus as a single connected graph. The Feasibility Criterion for Phase I will be to measure and show clear progress in RDF statement disambiguation against a data store containing tens of thousands of statements. Benefits: One of the most challenging problems in intelligence gathering and processing is resolving the issues of ambiguous entity references. Approaches to resolving ambiguous entity references dramatically increase the chances of finding otherwise hidden connections and decrease the possibilities for erroneous analysis. On the commercial side the applications are widespread to reduce errors for any application that maintains client lists (financial services, marketing companies, retail loyalty programs and so forth), as well as firms involved in security or investigations where identity is critical. These use cases provide a range of opportunities in DoD, Other Government groups and the commercial marketplace for effective approaches to resolving ambiguous entity references. Return