【作者】 霍跃红；

【导师】 汪榕培；

【作者基本信息】 大连理工大学 ， 计算机应用技术， 2010， 博士

【摘要】 译者研究是翻译史研究的重要组成部分。目前国内外翻译研究关注的重点仍然是翻译的标准和翻译的技巧,而对翻译的主体——译者,则缺乏系统的、有深度的研究。本文把译者研究与计算机技术有效地结合起来,运用语料库和机器学习的方法研究译者和译者文体,语料库为机器学习提供分类指标特征,机器学习为语料库提供崭新的计算机处理和研究方法,语料库技术和机器学习方法是本文研究译者文体并进行文本译者识别的技术路线。本文做了如下工作：(1)界定了典籍英译的概念,阐述了典籍英译的意义、主体和策略。设计并实现了一个计算机辅助典籍英译核心译者发现的系统。通过用《典籍英译研究》(第一二、三辑)作为原文进行实例分析和验证,发现了典籍英译领域的核心译者,得到了理想的效果。不仅为译者研究提供了客观发现核心译者的方法,也提供了了解译者的影响力和活跃度的参考。(2)探讨了译者定性研究的框架。把目标译者放入其教育背景、认知环境下去考察其翻译情况。从译者对原文本的选择和翻译策略出发,提出了本色译者的概念；从翻译过程考虑,探讨了译者研究的认知视角；从翻译策略的选择和译本的表现形式,研究了文化翻译观下的策略选择与中国英语现象；最后梳理了中国翻译伦理体系的发展,提出了回归翻译伦理的设想。(3)建立核心译者平行语料库和对比语料库,并在所建立的平行语料库和对比原料库的基础上从词汇、句子、语篇三个层次对译者文体进行了比较研究。通过对不同译者所翻译的《孔雀东南飞》的文本进行分析,比较研究了不同译者的文体。以典籍英译译者汪榕培教授的译本为主要研究对象,以英国著名汉学家Arthur Waley的译本和我国著名翻译家许渊冲的译本为参照,基于自建语料库从词汇、句子、语篇三个层次对译者文体进行了比较研究。(4)用基于语料库分析得到的代表译者文体的特征集作为文本表示的特征集,进行了文本译者识别的实验,提供了未知译者身份识别的方法。基于语料库分析的结果,提取了代表译者文体特征的三个层面的31维特征集,进行文本表示,建立向量空间模型,运用五种文本分类方法进行了三个阶段的实验。首先是运用基于单词频率的文本识别方法进行文本译者识别,获得了一定的准确率；之后扩大了译者文体特征的维数,进行了基于多层面特征的文本译者识别实验,识别的正确率有显著提高；最后基于信息增益特征选择进行实验,使实验复杂度降低,时间缩短,识别率提高。更多还原

【Abstract】 Studies on translators make up an important part in translation history. At present the focus of translation studies both at home and abroad is still confined within discussions on the standards and skills of translation. As to the subjects of translation,the translators themselves, the extent and depth of research are far from satisfactory or systematic. A new method is proposed here to combine studies on translators with computer technology. Developments in corpus linguistics and machine learning have provided scientific approach and technical support for the studies on translators and their styles.Corpus-based studies will discover and induce features for text categorization through machine learning, and machine learning will give in return new perspectives and technology for corpus-based studies on translators and their styles. Corpus technology and machine learning.are the two technical routes for the stylistic study of translators and the identification of their translatorship.The dissertation has done research in the following aspects:1.Defining the concept of translating the Chinese classics into English first, and discussing the meaning, the subject and strategies involved in this field.A computer-aided key translator discovery system has been developed. Analysis and testing have been carried out on the system with the three collections of Studies on Translation of Chinese Classics into English as the material and key translators have been discovered by the system with a very positive result, providing reference for the influence and activeness of the translators.2. Setting up a qualitative framework to study translators, placing the translators in the educational background and cognitive environment. This study has tentatively proposed a concept of individualized translators considering translators’ choice of source texts and the strategies they use. The cognitive perspective is proposed when examining the translation process and the use of China English is discussed when approaching translators from the cultural perspective. At last the ethical turn is suggested after combing the development of ethics in translation in China.3.Constructing both parallel and comparable corpora of translated poems by the key translators in the field of translating Chinese classics into English, and carrying-out statistic analyses on the three levels of words,sentences and discourse analysis. The dissertation did a comparative study of translators’ style reflected in the different translations done by three translators of China’s first long narrative poem:"Kongque dong nan fei". With Wang Rongpei’s version as the main object of study while using Waley’s and Xu Yuanchong’s versions for comparison, the present research did a multi-dimensional analysis of the translators’style at levels of words, sentences and discourse based on the self-constructed corpora..4.Carrying out translator identification experiments with the stylistic features induced from the corpus-based stylistic study of the former chapter, providing a method for the identification of translators for unidentified versions. With the translated text features as a 31-dimensioned feature set, a Vector Space Model (VSM) has been set up and three stages of experiments have been carried out with five different classifiers. First is the translators’ text categorization with the common classifier for writers’ text classification, achieving acceptable accuracy. Then comes the multi-layered translators’ text categorization with the enlarged attribute set of 31 dimensions,all with accuracy greatly improved.Finally we have done the categorization experiment with the attribute set after the Information Gain. The result achieved is that the complexity has been reduced, time consumed has been shortened and the accuracy of the identification has been obviously improved.更多还原