A Novel Japanese Anaphora Resolution Method Using Deep Cases

Anaphora resolution resolves a reference of a pronoun to a noun in a phrase. It is an active area of research, especially in text mining, because any result can be missed due to a mean of sentences including abbreviations or pronouns. An existing system, called zero anaphora analysis, cannot guarantee accurate results of detecting and complementing abbreviated pronouns as for Japanese sentence. Previous studies of anaphora resolution used surface case, which represents the semantic relationships between verbs and nouns. However, they analyzed only at the surface information of the emerging particle; therefore, it is difficult to obtain an appropriate result. Obviously, not only the analysis of superficial sentences but also the deep case should be used to address the problem. In this paper, we propose a new anaphora resolution method by using deep case.