Unsupervised Event Coreference Resolution with Rich Linguistic Features

Abstract

This paper examines how a new class of nonparametric Bayesianmodelscanbeeffectively applied to an open-domain event coreference task. Designed with the purpose of clustering complex linguistic objects, these models consider a potentially infinitenumber of features andcategorical outcomes. The evaluation performed for solving both within- and cross-document event coreference shows significant improvements of themodelswhencompared against twobaselines for this task. 1