【作者】 许雅缘；

【导师】 陈维振；

【作者基本信息】 福建师范大学 ， 英语语言文学， 2008， 硕士

【摘要】 作为自然语言理解最棘手的问题之一,隐喻逐渐引起了学者们的关注。当前的篇章处理已经取得了较大的成就,但在准确性上仍然无法令人满意。句子是篇章的重要构成元素,对句子的处理是篇章处理的关键。对于常规的句型处理,目前的篇章处理技术已经基本上可以满足需求。然而,对篇章中的一些较为特殊的句子的理解,机器处理还存在较大问题,隐喻句的处理就属于这一类。隐喻问题若不能得到较好的解决,篇章理解的准确度便难以突破,一系列依赖于篇章理解准确度的技术比如自动文摘、机器翻译、信息检索等也难以取得令人满意的进展。本文致力于隐喻的自动处理研究。我们从认知角度对隐喻的定义、分类和识别进行探讨,指出隐喻识别的主要标准——冲突。接着,以WordNet这个在线词典数据库系统里的各种语义知识和语义关系为基础,开发出一个处理“A is B”形式的英文隐喻自动处理系统。该系统主要包括对语句的预处理和隐喻语句的自动判别。系统对语句先进行预处理,包括分词和句法分析;并对主、表语进行知识搜索和语义计算(语义相似度),进而判别是否为隐喻。我们采用数据库管理系统的相关技术,使得隐喻理解处理时需要的知识可以很方便的在其中搜索,提高系统对隐喻处理的效率。最后,我们分别从Goatly创建的英语隐喻语料库Metalude和桂诗春创建的中国学习者英语语料库中选取大量隐喻和非隐喻的例子,对系统进行测试和评估,证实了该系统的合理性和可行性。更多还原

【Abstract】 Metaphor has always been the focus of research. Since Lakoff and Johnson put forward the Conceptual Metaphor theory, more and more people have devoted themselves to metaphor studies from, the perspective of cognitive science. Metaphor is now no longer taken as a deviation from the regular language, which is merely a device of the figure of speech for decoration, but as a basic human thinking way or cognitive process. The scope of metaphor studies has been greatly extended, covering cognition, psychology, philosophy, pragmatics, semiotics, and Artificial Intelligence, etc. Natural Language Processing, which is a branch of Artificial Intelligence, plays an important role nowadays. Great achievements have been made on regular sentence processing, but other sentences including metaphorical sentences remain a bottleneck for Natural Language Processing. Automatic processing of metaphor, will pave the way for Automatic digest, Machine translation and Information Retrieval.We endeavor to carry out a research in automatic processing of metaphor. Firstly, we explore the metaphor theories, including definition, classification and identification of metaphor. Based on WordNet, an on-line lexical reference system, we then design a metaphor automatic processing system which deals with the A is B pattern. The system consists of Sentence Parsing Module, Word Searching Module, Database Administration Module and Similarity Calculating Module. It analyzes the sentence, extracts the subject and predicate, calculates their similarity and judges whether the sentence is a metaphor. We adopt Java Database Connectivity to administrate database so that the system will work more efficiently. Both metaphorical and literal evidence is offered in this thesis to evaluate the system. A large number of metaphorical sentences are collected from the English metaphor corpus—Metalude. Literal sentences are mainly collected from the Chinese Learner English Corpus.更多还原