Corry: A System for Coreference Resolution

Proceedings of the 5th International Workshop on Semantic Evaluation, ACL 2010, pages 100–103, Uppsala, Sweden, 15-16 July 2010. c⃝2010 Association for Computational Linguistics Corry: A System for Coreference Resolution Olga Uryupina CiMeC, University of Trento uryupina@gmail.com Abstract Corry is a system for coreference resolution in English. It supports both local (Soon et al. (2001)-style) and global (Integer Linear Programming, Denis and Baldridge (2007)- style) models of coreference. Corry relies on a rich linguistically motivated feature set, which has, however, been manually reduced to 64 features for efficiency reasons. Three runs have been submitted for the SemEval task 1 on Coreference Resolution (Recasens et al., 2010), optimizing Corry’s performance for BLANC (Recasens and Hovy, in prep), MUC (Vilain et al., 1995) and CEAF (Luo, 2005). Corry runs have shown the best performance level among all the systems in their track for the corresponding metric. 1 Introduction Corry is a system for coreference resolution in En- glish. It supports both local (Soon et al. (2001)-style) and global (ILP, Denis and Baldridge (2007)-style) models of coreference. The backbone of the system is a family of SVM classifiers for pairs of mentions: each mention type receives its own classifier. A sep- arate anaphoricity classifier is learned for the ILP setting. Corry relies on a rich linguistically moti- vated feature set, which has, however, been manu- ally reduced to 64 features for efficiency reasons. Corry has only participated in the “open” setting, as it has already a number of preprocessing mod- ules integrated into the system: the Stanford NLP toolkit for parsing (Klein and Manning, 2003) and NE-tagging (Finkel et al., 2005), Wordnet for se- mantic classes and the U.S. census data for assigning gender values to person names. Three runs have been submitted for the Se- mEval task 1 on Coreference Resolution, optimizing Corry’s performance for BLANC, MUC and CEAF. The runs differ with respect to the model (local for BLANC, global for MUC and CEAF) and the defi- nition of mention types. 2 Preprocessing and Mention Extraction In our previous study (Uryupina, 2008) we have shown that up to 35% recall and 20% precision er- rors in coreference resolution for MUC corpora are due to inaccurate mention detection. We have there- fore invested substantial efforts into our mention de- tection module. Most state-of-the-art coreference resolution sys- tems operate either on gold markables or on the output of an ACE-style mention detection module. We are not aware of extensive studies on mention extraction algorithms for such datasets as SemEval (OntoNotes) where mentions are complex NPs not constrained with respect to their semantic types. We rely on the Stanford NLP toolkit for extract- ing named entities (Finkel et al., 2005) and parse trees for each sentence (Klein and Manning, 2003). We then merge the output of the NE-tagger and the parser to create a list of mentions in the following way: 1. Named entities are considered mentions if they correspond to a sequence of parsing con- straints. 2. Pronouns are considered mentions if they are …

Content truncated to 500 words