... partitionsthan the MUC scorer, for the following reasons.First, our candidate partitions in general agree more on singleton clusters than on non-singleton clusters.Second, by employing a non-uniform ... possiblycontradictory pairwise coreference classification de-cisions and constructs a partition on the given set ofNPs, with one cluster for each set of coreferent NPs.Though reasonably successful, ... First, while the MUC scorer only rewardscorrect identification of coreferent links, B-CUBEDadditionally rewards successful recognition of non-coreference relationships. Second, the MUC scorerapplies...