Chinese zero pronoun resolution: An unsupervised probabilistic model rivaling supervised resolvers

Abstract

State-of-the-art Chinese zero pronoun resolution systems are supervised, thus relying on training data containing manually resolved zero pronouns. To eliminate the reliance on annotated data, we present a generative model for unsupervised Chinese zero pronoun resolution. At the core of our model is a novel hypothesis: A probabilistic pronoun resolver trained on overt pronouns in an unsupervised manner can be used to resolve zero pronouns. Experiments demonstrate that our unsupervised model rivals its state-ofthe- Art supervised counterparts in performance when resolving the Chinese zero pronouns in the OntoNotes corpus.