REDI: Resolving Difference via Inference

ABSTRACT: Intelligence analysts have access to expanding collections of rapidly evolving information stores, but not all information is equally credible or trustworthy. What is needed is the ability to examine inconsistencies in contextusing an understanding of temporal evolution, to consider how different conflicts may be related, and how the underlying sources factor into the credibility of the information. In response, we will develop inference algorithms for detecting and resolving conflicting information in multi-source text data, and will design algorithms that place the potentially conflicting information in context to suggest resolutions and to understand the social connections behind the differing information. We will design an extensible framework that integrates data preprocessing, merging, conflict detection and assessment techniques, and methods for performing the necessary inferences to resolve conflicting data, when appropriate. Our work will ensure that the system is able to rapidly adapt to new and evolving data sets, and integrates with analysts"information systems. Further, by leveraging our ongoing work with Joint Inter Agency Task Force South (JIATFS) in Key West, FL, we will design our system to be readily integrated into a real multi-source intelligence analysis system, and can design experiments on real, operational data. BENEFIT: Phase I will build the foundation to automatically detect and resolve conflicts for intelligence applications. This capability enables analysts to quickly and reliably integrate growing and diverse data stores with the much-needed support for identifying and managing the potentially erroneous, misleading, or deceptive content that may appear within these collections. By reviewing the sources"social connections, shared information, and potential biases, we will place conflicting information in its proper context to make informed inferences about its resolution. Our extensible framework will allow our system to be integrated as a component in multiple government intelligence systems, and has commercial applications for news reporting, marketing, and public relations firms that must rapidly aggregate and resolve conflicting data to determine its correctness and potential biases that may be present.