Evaluating Dependency Representations for Event Extraction

Abstract

The detailed analyses of sentence structure provided by parsers have been applied to address several information extraction tasks. In a recent bio-molecular event extraction task, state-of-the-art performance was achieved by systems building specifically on dependency representations of parser output. While intrinsic evaluations have shown significant advances in both general and domain-specific parsing, the question of how these translate into practical advantage is seldom considered. In this paper, we analyze how event extraction performance is affected by parser and dependency representation, further considering the relation between intrinsic evaluation and performance at the extraction task. We find that good intrinsic evaluation results do not always imply good extraction performance, and that the types and structures of different dependency representations have specific advantages and disadvantages for the event extraction task.