Reference: Fikes, R.; Marwick, A.; & Thurman, D. Knowledge Associates for Novel Intelligence (KANI). Knowledge Systems Laboratory, October, 2003.
Abstract: We describe here a project being conducted in the Novel Intelligence from Massive Data (NIMD) program of the intelligence community's Advanced Research and Development Activity (ARDA). This project is developing a system of automated associates called KANI that is intended to actively support and participate in the analysis task by helping an analyst identify, structure, aggregate and visualize task-relevant information; by enabling an analyst to construct explicit models of alternative hypotheses (scenarios, relationships, causality, etc.); and by actively assisting an analyst in analytical reasoning such as hypothesis refinement, contradiction recognition, and assumption testing. The primary enabler of these capabilities is the production and use of computer interpretable knowledge expressed in formal knowledge representation languages. Our focus is on developing technologies for knowledge representation in support of hypothesis modeling, automated reasoning in support of evidence analysis, knowledge extraction, knowledge-enhanced search, explanation generation, and knowledge and process visualization. Our objective is to create technologies that can enable acceleration and deepening of analytical processes as they are being orchestrated and directed by the analyst. We are developing the following KANI associates: an "Hypothesis Generation and Tracking Associate" that employs formal knowledge representation and reasoning techniques to model, guide, accelerate, and deepen analytical processes as they are being orchestrated and directed by the analyst; a "Massive Data Extraction and Structuring Associate" that will be the main component for ingesting text documents, identifying and extracting relevant knowledge, structuring that knowledge as document annotations and ontologies, and presenting that knowledge to the other KANI associates; a "Background Knowledge Identification and Assembly Associate" that an analyst can use to identify and assemble relevant structured, semi-structured, and unstructured background information for a given set of documents and a given task; and an "Information Interaction Associate" that will facilitate analyst interaction with each of the other KANI associates and provide an interactive representation of the analytic process that can be inspected, revised, shared, and analyzed for patterns, biases, and deficiencies. In this paper, we summarize the work being done on each associate and provide pointers to technical reports that provide more detailed descriptions of that work.
Full paper available as doc.