Joint Entity and Event Coreference Resolution across Documents

Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 489–500, Jeju Island, Korea, 12–14 July 2012. c⃝2012 Association for Computational Linguistics Joint Entity and Event Coreference Resolution across Documents Heeyoung Lee, Marta Recasens, Angel Chang, Mihai Surdeanu, Dan Jurafsky Stanford University, Stanford, CA 94305 {heeyoung,recasens,angelx,mihais,jurafsky}@stanford.edu Abstract We introduce a novel coreference resolution system that models entities and events jointly. Our iterative method cautiously constructs clusters of entity and event mentions using lin- ear regression to model cluster merge opera- tions. As clusters are built, information flows between entity and event clusters through fea- tures that model semantic role dependencies. Our system handles nominal and verbal events as well as entities, and our joint formulation allows information from event coreference to help entity coreference, and vice versa. In a cross-document domain with comparable doc- uments, joint coreference resolution performs significantly better (over 3 CoNLL F1 points) than two strong baselines that resolve entities and events separately. 1 Introduction Most coreference resolution systems focus on enti- ties and tacitly assume a correspondence between entities and noun phrases (NPs). Focusing on NPs is a way to restrict the challenging problem of coref- erence resolution, but misses coreference relations like the one between hanged and his suicide in (1), and between placed and put in (2). 1. (a) One of the key suspected Mafia bosses ar- rested yesterday has hanged himself. (b) Police said Lo Presti had hanged himself. (c) His suicide appeared to be related to clan feuds. 2. (a) The New Orleans Saints placed Reggie Bush on the injured list on Wednesday. (b) Saints put Bush on I.R. As (1c) shows, NPs can also refer to events, and so corefer with phrases other than NPs (Webber, 1988). By being anchored in spatio-temporal dimen- sions, events represent the most frequent referent of verbal elements. In addition to time and location, events are characterized by their participants or ar- guments, which often correspond with discourse en- tities. This two-way feedback between events and their arguments (or entities) is the core of our ap- proach. Since arguments play a key role in describ- ing an event, knowing that two arguments corefer is useful for finding coreference relations between events, and knowing that two events corefer is use- ful for finding coreference relations between enti- ties. In (1), the coreference relation between One of the key suspected Mafia bosses arrested yesterday and Lo Presti can be found by knowing that their predicates (i.e., has hanged and had hanged) core- fer. On the other hand, the coreference relations be- tween the arguments Saints and Bush in (2) helps to determine the coreference relation between their predicates placed and put. In this paper, we take a holistic approach to coref- erence. We annotate a corpus with cross-document coreference relations for nominal and verbal men- tions. We focus on both intra and inter-document coreference because this scenario is at the same time more challenging and more relevant to real-world …

Content truncated to 500 words