Current state-of-the-art systems for event extraction lack the ability to yield high accuracy, concept-based results. We propose to design and build an enabling technology for concept-based event extraction that includes 1) a knowledge-based approach for disambiguation, normalization, and consolidation of event information, 2) use of automatically identified linguistic structures to collect as much relevant event information as possible even across sentence boundaries, 3) use of a linguistic database approach to support future applications to aid analysts, and 4) development of a prototype to demonstrate effectiveness of the overall approach. Phase I deliverables will include the assessment of the feasibility of our approach and design of a software prototype for high accuracy concept-based event extraction.
