Description:
OBJECTIVE: Provide improved real-time situational awareness through discovery of unknown relationships across multiple structured and unstructured textual data sources. DESCRIPTION: The number of textual data sources, formats, and types available to information analysts has exploded in recent years. Often, the relevant data about an entity or event of interest is scattered across multiple data sources and is incomplete within a single data source. There is a need to discover previously unknown relationships pertaining to entities and events of interest across these multiple data sources in order to help analysts find more complete information about those entities and events. Such a capability would support tasks such as maintaining situational awareness, patterns of life analysis, and providing real time alert notifications to the analyst from multiple text-based intelligence data sources. Current approaches to relationship extraction suffer from limitations of using contextual and syntactical cues, which are not scalable and have limited statistical reliability in identifying previously unknown relationships. By leveraging new developments in statistical relational learning, novel relationship discovery algorithms are emerging. The challenge in an operational environment is to provide situational awareness to the analyst in real time. This requires the development of incremental learning models that reduce the time complexity currently required. Basic algorithm work is needed to not only leverage the advances in statistical relational learning for relationship discovery, but to do so in an operational setting enabling online updates of models across multiple unstructured and structured data sources. This analytic operation would significantly increase the accuracy of reports or intelligence summaries that could be used to provide real-time alert notifications to the user. The goal of this topic is to research and develop algorithms for discovering previously unknown relationships, pertaining to entities and events of interest, across multiple unstructured semi-structured and structured text-based intelligence data sources, mIRC, and various databases. The algorithms should be scalable and capable of operating in real time, in order to support both forensic and real-time use cases. Such a capability would enable intelligence analysts to efficiently leverage more sources of data for their tasks and enhance their situational awareness by providing more complete information than can be found in each data source alone or by manually identifying relationships across sources in a time-limited window of analysis. Military Application: The higher order fusion of information is fundamental building capability to enhancing the output from today"s intelligence analysts. The techniques and algorithms developed can be applied to information centric systems to simplify the discovery of not just information but to detect broader relationships than possible today. Relationships can be identified based on holistic analysis of all available data to enable support of warfighters in an operational setting where improved situational awareness is crucial. The results of this research would also be useful to Department of Homeland Security's many databases, like that used to check passengers on various forms of traffic (such as airplanes, trains and ships). Commercial Application: The development and products from this research endeavor can be applied to information centric systems that require the discovery of relations that can lead to unknown intelligence or knowledge. Examples of systems that would benefit would be FBI database scheme, and state and local police databases. The ability to discover relations in data would also be beneficial to banking and state motor vehicle agencies. PHASE I: The goal of Phase I is to investigate scalable algorithms for discovering unknown relationships, pertaining to entities and events of interest, across multiple unstructured and structured textual data sources in real time. The investigation should produce a prototype design for relationship discovery across these sources. Developed capabilities will be demonstrated by use of open source data. PHASE II: The goal of Phase II is to implement the Phase I design into a prototype system that can be demonstrated across multiple unstructured and structured textual data sources, in forensic and real-time scenarios. Unclassified and classified data will be provided for evaluation of developed research. The prototype system should be SOA-compliant to facilitate integration with other systems. PHASE III DUAL USE APPLICATIONS: A capability for discovering unknown relationships across multiple data sources would benefit military intelligence analysts. It would also benefit customers, such as FBI, state/local police, and Department of Homeland Security, who leverage data found in multiple databases and unstructured sources.
