语言、逻辑与计算互动视角下汉语直接被动句的MMCCG处理

Processing of Chinese Direct Passive Sentences in Multi-modal Combinatory Categorial Grammar (MMCCG) Under the Interaction Perspective of Language, Logic and Computation

汉语被动句有直接和间接之分,也有长短之别,如何处理汉语被动句历来是计算语言学家需要攻克的难题。汉语直接被动句的显著特征是主谓语的受事指向主句主语,传统分析方法使用移位、删除等句法操作分析汉语被动句,否定了被动句的原生成性,也出现了许多难以解释的现象。多模态组合范畴语法MMCCG抛弃了移位和删除等句法操作,以表层组合的方式生成各种被动句的句法-语义推演树,避免了传统方法分析被动句时出现的所有问题,并从计算语言学视角证明了被动句的原生成性。

Chinese passive sentences can be divided into direct and indirect sentences, as well as their length. How to deal with Chinese passive sentences has always been a difficult problem for computational linguists. The remarkable feature of Chinese direct passive sentences is that the patient of the predicate in subject sentence points to the subject in it. In the traditional analysis, syntactic operations, such as movement and deletion are often used to analyze Chinese passive sentences, which denies the original generation of passive sentences, and there are many phenomena that are difficult to explain. Multi-modal Combinatory Categorial Grammar(MMCCG)discards syntactic operations, such as shift and deletion, and generates syntactic semantic deduction trees of various passive sentences in the way of surface combination, so it can avoid all the problems in the traditional analysis of passive sentences, and prove the original generation of passive sentences from the perspective of computational linguistics.