【作者】 黎斌；

【导师】 唐跃勤；

【作者基本信息】 西南交通大学 ， 外国语言学及应用语言学， 2004， 硕士

【摘要】 机器翻译是指计算机通过指令和程序将一种自然语言转换为另一种自然语言的过程。机器翻译是一个跨学科的研究领域，涉及计算机科学、语言学、人类学和心理学等多方面的知识。本文首先回顾了机器翻译在世界和中国的发展。经过几十年的努力，机器翻译已取得了令人瞩目的成绩。特别是作为人们的辅助翻译工具，机器翻译得到大多数人的认可，近年来国内外翻译软件的蓬勃发展充分证明了这一点。尤其是诸如译星、雅信2.5、IBM翻译家、东方快车XP、金山快译2003和译典通2003之类的全文翻译软件在市场上广受欢迎。 然而，目前大多数全文翻译软件的操作程度、翻译准确性、可读性等都不很理想。这些翻译软件存在的问题主要有：歧义消解不当、代词指代不明、省略补充不当以及词序(即句法结构)处理不当。本文通过英汉对比研究，主要讨论机器翻译软件在翻译英语特殊句型时，如何恰当地调整词序以符合汉语语法特点。 本文以翻译等值、转换生成语法和句法关系为理论基础，对英语中的几种特殊句型进行结构分析，通过英汉对比研究找出其异同。在不考虑个别单词词义的情况下，将There be句型、It句型和被动语态等特殊句型在上述六种全文翻译软件中进行测试。测试结果表明，在绝大多数情况下，全文翻译软件大都只是在实现查词功能后，将单个的字或词按原文顺序排列出来，即字对字的硬译，有时译文令人费解，甚至荒谬可笑。这些软件主要存在两个方面的问题：一是对虚词的处理，如There be句型中的There，It句型中的It，以及对被动结构“be+v.-ed”处理；二是译成的中文仍然按照原来的英语词序排列，没有调整为符合汉语特征的语序。 在分析以上两点不足的基础上，本文为克服全文翻译软件的这些缺陷提出了种种假设。并以Delphi 7设计了一个试验性程序。对此程序的验证说明，文中提出的假设是可行的，这给机器翻译研究者解决其它机器翻译问题提供了语言学方面的参考，尤其是对机器翻译软件的发展提供了一定的帮助。更多还原

【Abstract】 Machine Translation (MT), automatic translation of human languages by computers, has brought about great popularity as well as controversy throughout the world. This thesis starts with a brief review of MT development, its birth and growth in China and worldwide. After great efforts devoted in the past decades, considerable success has been achieved in MT research, which could be strongly proved by the prosperity of MT software, specifically software for article translation, such as Transtar, Yacin CAT 2.5, IBM Manager 2000, Dongfangkuaiche XP, Jinshankuaiyi 2003 and Dr.eye (Yidiantong 2003).However, there is still a long way to go in MT for the reason that some crucial problems remain unsolved, such as ambiguity, pronoun reference, restoration of ellipsis and adjustment of word orders in translating typical patterns. Some typical factors contribute to these difficulties. For instance, words with multiple meanings or sentences with multiple grammatical structures may lead to ambiguity; the pronoun reference depends heavily on the context recognition and determination, on the coordination of related words or the acting range of conjunction.As a result, the paper, from the perspective of comparative study between English and Chinese, just proceeds to focus on the adjustment of word orders in translating some typical patterns from English into Chinese. Based on three fundamental theories available for MT: translation equivalence, Generative-Transformational grammar and syntactic relations, diagnostic tests are carried out to examine the efficiency of the MT software, in terms of the adjustment of word orders.A conclusion can be drawn that, by means of contrastive study, there is a big distinction in some typical patterns between English and Chinese: There be pattern, It pattern, passive voice and the like. Translation skills or techniques in particular are required to adjust versions to the unique characteristics of the patterns in both languages. This makes it difficult for MT software to render such patterns from English into Chinese: one is the dealing with such empty anticipatory as There in There be pattern and It in It pattern, and the structure "be + verb-ed" in passive voice; the other is the adjustments of word orders adapting to Chinese styles.Some hypotheses are presented in the paper to cope with both the disadvantages in MT software on the basis of the diagnostic tests and analyses in detail. In the end, a tentative program for There belli pattern is designed based on those hypotheses, which can be practically attested by the samples that originally the MT software has problems in translating. The success of the tentative program is bound to benefit some MT researchers from the perspective of linguistics, specifically those who are involved in the study of MT software.更多还原