This paper describes the participation of RelaxCor in the CoNLL-2011 shared task: "Modeling Unrestricted Coreference in Ontonotes". RELAXCOR is a constraint-based graph partitioning approach to coreference resolution solved by relaxation labeling. The approach combines the strengths of groupwise classifiers and chain formation methods in one global method.
Sapena, E.; Padró, L.; Turmo, J. RelaxCor participation in CoNLL shared task on coreference resolution. A: Conference on Computational Natural Language Learning. "CONLL Shared Task '11 Proceedings of the Fifteenth Conference on Computational Natural Language Learning: Shared Task". Portland, Oregon: 2011, p. 35-39.