Concept-Based Event Extraction Utilizing Rich Semantics (CONVERSE)

This effort involves a novel approach to event extraction that relies on rich semantic knowledge. Semantic information ranging from semantic word senses, semantic frames that describe conventional situations, predicate-argument structures and named entity classes are used to produce conceptual representations of events and to identify all essential information. In addition, in this effort we propose to enhance the semantic information with discourse-level information for capturing event-related knowledge which spans across multiple sentences. The concept-based event-extraction methodology proposed herein will enable information analysts to effectively monitor and analyze event-relevant information from textual data sources.