Chinese Metaphor Computation: the State of Art
ZHOU Chang-lei
DOI: https://doi.org/10.3785/j.issn.1008-942x.2007.05.008
2007-01-01
Abstract:Metaphor, an ordinary everyday phenomenon of language use, has become the focus of mind and language mechanism. The comprehension of metaphor by machine will be a bottleneck problem in natural language understanding and machine translation. Therefore, great attention has been paid to computational mechanisms of metaphor within the last few decades, with respect to cognitive mechanism, machine understanding, machine recognition, application, etc. Although metaphor computation may encounter many problems for the rich expressive power of natural language and cognitive nature of human being, scientists still have achieved a lot on constructing computational models of metaphor. In fact, most of the current achievements are limited to the English language, whereas research on Chinese metaphor computation has just got started. In this paper, the author attempts to make a summary of the research results concerning Chinese metaphor computation that have been achieved in recent years, and suggests reconsidering the existing problems and limitations revealed in the study, and makes proposals for further research on computational mechanisms of metaphor in general.Specifically, achievements in the study of Chinese metaphor computation have been scored from three main perspectives, i.e., categorization architecture, machine recognition and classification, logic representation and interpretation of metaphor. Firstly, based on the view of cognitive structures on the relations among tenor, vehicle and the similarity between them in a metaphor, we have put forward a system of Chinese metaphor categorization architecture, according to which there are altogether nine types of Chinese metaphor. Statistical verification and linguistic explanation are also proposed to support this categorization. Secondly, we have constructed a computational model of metaphor representation, called metaphorical semantic network (MSN), according to the characteristics of metaphor. This model offers methods to formalize Chinese sentences, and has applications on metaphorical classification. Thirdly, we have also created a metaphor logic system by giving definition, constructing and analyzing the properties of the logic, which has good capability to analyze the metaphor sentences' meaning, and gives instructional method for computer to perform metaphor comprehension. Furthermore, we have attempted to introduce subjective factor into the logical system and cognitive computation. Although this study is just on an elementary phase, it is no doubt a rewarding and beneficial attempt.Still, the present research on Chinese metaphor computation has also revealed many problems and limitations. In the final part of this paper, the author bears in mind all these challenges, and proposes several suggestions for future research. Especially, metaphor classification based on human understanding should be constructed; the metaphor properties knowledge base should be built up; the logical system of metaphor representation and analogical reasoning should be set up; and in application, the Chinese-to-English machine translation system and search engine should get rebuilt based on metaphor understanding.