Japanese Zero Reference Resolution Considering Exophora and Author/Reader Mentions

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 924–934, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Japanese Zero Reference Resolution Considering Exophora and Author/Reader Mentions Masatsugu Hangyo Daisuke Kawahara Sadao Kurohashi Graduate School of Informatics, Kyoto University Yoshida-honmachi, Sakyo-ku Kyoto, 606-8501, Japan {hangyo,dk,kuro}@nlp.ist.i.kyoto-u.ac.jp Abstract In Japanese, zero references often occur and many of them are categorized into zero ex- ophora, in which a referent is not mentioned in the document. However, previous studies have focused on only zero endophora, in which a referent explicitly appears. We present a zero reference resolution model considering zero exophora and author/reader of a docu- ment. To deal with zero exophora, our model adds pseudo entities corresponding to zero exophora to candidate referents of zero pro- nouns. In addition, we automatically detect mentions that refer to the author and reader of a document by using lexico-syntactic patterns. We represent their particular behavior in a dis- course as a feature vector of a machine learn- ing model. The experimental results demon- strate the effectiveness of our model for not only zero exophora but also zero endophora. 1 Introduction Zero reference resolution is the task of detecting and identifying omitted arguments of a predicate. Since the arguments are often omitted in Japanese, zero reference resolution is essential in a wide range of Japanese natural language processing (NLP) appli- cations such as information retrieval and machine translation. (1) パスタが pasta-NOM 好きで like 毎日 everyday (φ ガ) (φ-NOM) (φ ヲ) (φ-ACC) 食べます。 eat (Liking pasta, (φ) eats (φ) every day) For example, in example (1) , the accusative argu- ment of the predicate “食べます” (eat) is omitted .1 The omitted argument is called a zero pronoun. In this example, the zero pronoun refers to “パスタ” (pasta). Zero reference resolution is divided into two sub- tasks: zero pronoun detection and referent identifi- cation. Zero pronoun detection is the task that de- tects omitted zero pronouns from a document. In example (1), this task detects that there are the zero pronouns in the accusative and nominative cases of “食べます” (eat) and there is no zero pronoun in the dative case of “食べます”. Referent identifica- tion is the task that identifies the referent of a zero pronoun. In example (1), this task identifies that the referent of the zero pronoun in the accusative case of “食べます” is “パスタ” (pasta). These two subtasks are often resolved simultaneously and our proposed model is a unified model. Many previous studies (Imamura et al., 2009; Sasano et al., 2008; Sasano and Kurohashi, 2011) have treated only zero endophora, which is a phe- nomenon that a referent is mentioned in a document, such as “パスタ” (pasta) in example (1). However, zero exophora, which is a phenomenon that a ref- erent does not appear in a document, often occurs in Japanese when a referent is an author or reader of a document or an indefinite pronoun. For example, in example (1), the referent of the zero pronoun of the …

Content truncated to 500 words