SBIR-STTR Award

To keep abreast of groups and individuals operating in a particular area, US forces gather vast amounts of intelligence. Extracted from this sea of information are the entities that exist in the data. We have made great leaps in entity extraction technology. The next challenge is to overcome entity ambiguity that remains at the end of this processing pipeline. These entities are typically stored as connected graphs in the Resource Description Framework (RDF). Because of ambiguities associated with entities, these data stores become filled with redundant statements, preventing Warfighters from finding everything known about specific entities rapidly and accurately. To address this shortfall, DAC is partnering with Cobham (formerly SPARTA) to bring together best-of-breed technologies for entity disambiguation. Cobham has developed the SPEAR NLP package that performs entity extraction. SPEAR is an end-to-end capability for automated Level 1 fusion of text-based multi-INT data. We build on SPEAR to provide the RDF Entity and Association Disambiguation (READ) capability. READ will provide Entity and Relationship Co-Reference Resolution, RDF Conflict Identification and Resolution, and improved RDF Schema for Context Storage. These capabilities will provide intelligence analysts the necessary tools to rapidly and accurately obtain all of the necessary information about specific entities.

Benefit:
The key benefits of the proposed solution are to maximize the accuracy and depth of knowledge stored within RDF data stores. Through the enhanced RDF data stores, analysts will be able to obtain all of the necessary information about specific entities without worrying about missing information due to naming conventions or getting incorrect information due to conflicting or redundant statements within the RDF.

Keywords:
Markov Logic Networks, Markov Logic Networks, entity extraction, conflict resolution, entity disambiguation, Semantic Web Resources, Resource Description Framework, Context Storage

To keep abreast of groups and individuals operating in a particular area, US forces gather vast amounts of intelligence. Extracted from this sea of information are the entities that exist in the data. We have made great leaps in entity extraction technology. The next challenge is to overcome entity ambiguity that remains at the end of this processing pipeline. These entities are typically stored as connected graphs in the Resource Description Framework (RDF). Because of ambiguities associated with entities, these data stores become filled with redundant statements, preventing Warfighters from finding everything known about specific entities rapidly and accurately. To address this requirement, DAC is partnering with Cobham (formerly SPARTA) to bring together best-of-breed technologies for entity disambiguation and conflict resolution. READ provides Entity and Relationship Co-Reference Resolution, RDF Conflict Identification and Resolution, and improved RDF Schema for Context Storage. These capabilities will provide intelligence analysts the necessary tools to rapidly and accurately obtain all of the necessary information about specific entities. In particular, the READ system will utilize logic-based probabilistic models to allow for the analysis and exploitation of large scale data stores consisting of static, dynamic, continuous, and discrete attributes.

Benefit:
The key benefits of the proposed solution are to maximize the accuracy and depth of knowledge stored within RDF data stores. Through the enhanced RDF data stores, analysts will be able to obtain all of the necessary information about specific entities without worrying about missing information due to naming conventions or getting incorrect information due to conflicting or redundant statements within the RDF.

Keywords:
Resource Description Framework, entity disambiguation, Context Storage, entity extraction, Logic-based Probabilistic Models, conflict resolution, Semantic Web Resources