RelaxCor: A Global Relaxation Labeling Approach to Coreference Resolution
Emili Sapena, Lluís Padró, Jordi Turmo · 2010
This paper presents a constraint-based graph partitioning approach to corefer-ence resolution solved by relaxation label-ing. The approach combines the strengths of groupwise classifiers and chain forma-tion methods in one global method. Ex-periments show that our approach signifi-cantly outperforms systems based on sep-arate classification and chain formation steps, and that it achieves the best results in the state of the art for the same dataset and metrics. 1