|Designing Test-beds for General Anaphora Resolution
|Year of Publication
|Postolache O, Cristea D
|Discourse Anaphora and Anaphor Resolution Colloquium-DAARC
|St. Miguel, Portugal
This paper proposes a framework for evaluating coreference resolution systems, taking into account the contribution of subcomponents. Our goal is to have a way to quickly identify bottlenecks so that the development effort can focus on the weakest part of the processing chain. We describe experiments on the contribution of the module for searching potential referential expressions using two types of input, plain text and Penn Treebank-style syntactically annotated data. We propose a metric for evaluating a coreference resolution system when the set of potential referential expressions in the input is not the same as the set of potential referential expressions in the gold.