Multiple argument ellipses resolution in Japanese

Multiple argument ellipses resolution in Japanese Shigeko Nariyama MIALS, The University of Melbourne Parkville Victoria Australia 3010 shigeko@unimelb.edu.au Abstract Some Japanese clauses contain more than one argument ellipsis, and yet this fact has not adequately been accounted for in the study of ellipsis resolution in the current literature, which predominantly focus resolving one ellipsis per sentence. This paper proposes a method using a "salient referent list", which identifies the referents of such multiple argument ellipses as well as offers ellipsis resolution as a whole by considering contextual information. Key words Ellipsis resolution, multiple argument ellipses, salient referent list 1 Introduction It has been widely recognised and has been a challenge in Japanese-to-English machine translation that Japanese frequently unexpresses nominal arguments, such as the subject and the object, which must be identified and made explicit in order to be translated into grammatical English. So far most attempts have been made on resolving only one ellipsis in a clause (not a sentence), predominantly of the subject ellipsis. However, some clauses contain multiple ellipses within the clause, e.g. the subject as well as the object. This paper proposes a method using a "salient referent list" for resolving such multiple argument ellipses by considering contextual information. This single method is not ad hoc, but offers a unified account that resolves not only multiple argument ellipses, but also ellipsis resolution as a whole. 2 Earlier studies Studies on ellipsis resolution in Japanese have focused to resolve one ellipsis per simplex sentence. For example, Centering Theory (Kameyama 1985, Walker et al. 1994) is confined to assign only one Cb (backward looking center) per sentence, consequently resolving at best one ellipsis. Kameyama (1986) proposed the "property sharing constraint" which is meant to account for this problem. However, as is shown later in this section, some examples show the inadequacy of her proposal. In addition, Centering Theory is particularly problematic, in that it mostly deals with simplex sentences, despite the fact that complex sentences are reported to comprise 87.5% of Japanese narrative texts (Nariyama 2000). In more recent work, Kameyama (1998) proposed an account for intrasentential Centering by breaking a complex sentence into a hierarchy of center-updating units, i.e. clauses in more general terms. My assumption on her account is to utilise conjunctive particles, by which a hierarchy of center-updating units for each complex sentence is determined. However, this account requires additional convoluted hierarchies and yet its results were shown to be still inadequate by Strube (1998). Even with the potential increase in the accuracy, this method still retains an unsolved problem – resolving non-subject ellipsis. Moreover, it is designed for English complex sentences, and implications for Japanese sentences are not addressed in her work. A similar account is taken by Nakaiwa who has produced a series of works on Japanese ellipsis resolution (1995, 1998, inter alia), which are said to account for intrasentential ellipsis by means of conjunctive particles. However, conjunctive particles per se can retrieve at best the identity of only a subject ellipsis, in terms …

Content truncated to 500 words