Chinese Zero Pronoun Resolution: An Unsupervised Probabilistic Model Rivaling Supervised Resolvers

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 763–774, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Chinese Zero Pronoun Resolution: An Unsupervised Probabilistic Model Rivaling Supervised Resolvers Chen Chen and Vincent Ng Human Language Technology Research Institute University of Texas at Dallas Richardson, TX 75083-0688 {yzcchen,vince}@hlt.utdallas.edu Abstract State-of-the-art Chinese zero pronoun res- olution systems are supervised, thus re- lying on training data containing manu- ally resolved zero pronouns. To elimi- nate the reliance on annotated data, we present a generative model for unsuper- vised Chinese zero pronoun resolution. At the core of our model is a novel hy- pothesis: a probabilistic pronoun resolver trained on overt pronouns in an unsuper- vised manner can be used to resolve zero pronouns. Experiments demonstrate that our unsupervised model rivals its state-of- the-art supervised counterparts in perfor- mance when resolving the Chinese zero pronouns in the OntoNotes corpus. 1 Introduction A zero pronoun (ZP) is a gap in a sentence that is found when a phonetically null form is used to refer to a real-world entity. An anaphoric zero pronoun (AZP) is a ZP that corefers with one or more preceding noun phrases (NPs) in the asso- ciated text. Below is an example taken from the Chinese TreeBank (CTB), where the ZP (denoted as *pro*) refers to 俄罗斯 (Russia). [俄罗斯] 作为米洛舍夫维奇一贯的支持者， *pro* 曾经提出调停这场政治危机。 ([Russia] is a consistent supporter of Milošević, *pro* has proposed to mediate the political crisis.) As we can see, ZPs lack grammatical attributes that are useful for overt pronoun resolution such as number and gender. This makes ZP resolution more challenging than overt pronoun resolution. Automatic ZP resolution is typically composed of two steps. The first step, AZP identification, in- volves extracting ZPs that are anaphoric. The sec- ond step, AZP resolution, aims to identify an an- tecedent of an AZP. State-of-the-art ZP resolvers have tackled both of these steps in a supervised manner, training a classifier for AZP identification and another one for AZP resolution (e.g., Zhao and Ng (2007), Chen and Ng (2013)). In this paper, we focus on the second task, AZP resolution, designing a model that assumes as in- put the AZPs in a document and resolves each of them. Note that the task of AZP resolution alone is by no means easy: even when gold-standard AZPs are given, state-of-the-art supervised resolvers can only achieve an F-score of 47.7% for resolving Chinese AZPs (Chen and Ng, 2013). For the sake of completeness, we will evaluate our AZP resolu- tion model using both gold-standard AZPs as well as AZPs automatically identified by a rule-based approach that we propose in this paper. Our contribution lies in the proposal of the first unsupervised probabilistic model for AZP resolu- tion that rivals its supervised counterparts in per- formance when evaluated on the Chinese portion of the OntoNotes 5.0 corpus. Its main advan- tage is that it does not require training data with manually resolved AZPs. This, together with the fact that its …

Content truncated to 500 words