| The study on translator’s style is a heated topic both in corpus linguistics and corpus-based translation studies(CBTS),and it mainly deals with translator’s pattern detection and feature explanation process.With the development of corpus linguistics,a comparatively new research path has facilitated the discussion in translator style study.Therefore,Baker’s framework(2000)concerning the literary translator’s style investigation has been widely used in corpus-based translator studies,where some typical parameters such as TTR/STTR,average sentence length,and special verbs are frequently put into examination.There exists,however,some limitations in the framework,since such features might be pre-defined by the researcher and thus are not sufficient enough to summarise the translator’s style,whereas certain implicit yet distinctive features that might better illustrate it are,therefore,neglected.In this thesis,we would,first of all,clarify the definition of translator’s style and then present several factors which might influence a literary translator’s style with reference to Hu & Xie(2017).As a case study,it is argued that Zhang Guruo’s translator style can be unveiled by examining the following aspects:(1)Different translators of the same original text;(2)Zhang’s other translation works;(3)Chinese creative writing;(4)Source text style.Methodologically,in order to avoid the limitations of the corpus-based method,a quantitative investigation based on the machine learning approach is employed.To begin with,three self-built corpora were established.In terms of the machine learning process,we have refined the procedure applied in Lynch & Vogel(2018)and Zhan & Jiang(2017),by establishing a relatively comprehensive and in-depth feature set,which is composed of two levels,one is document-level features,from lexical,syntactical to discoursal aspects,another is N-gram level,which is a new attempt in English-Chinese translator style detection.The experiments mainly apply SVM classifier,together with Na?ve Bayes,Simple Logistic Regression and Decision tree as reference classifiers.Eventually,by utilising a set of distinctive features retrieved from the classifiers,a hierarchical cluster analysis is conducted so as to present a clear and intuitive picture of the classifying result through visualisation.The experiment results have shown the validity of the method.In experiment one,the average accuracy could reach above 90%,proving the classifier could distinguish Zhang’s translation of Tess from the other two versions.The second experiment yielded 85% accuracy on average,which indicates that certain features are universal in the collection of Zhang’s translations.And the final experiment has achieved the goal to distinguish Zhang’s translator style from the Chinese native writings with an average accuracy above 90%.Then the cluster analysis has given a perfect division where Zhang’s translations have been grouped within one cluster,thus proving the effectiveness of Document-level and N-gram features in distinguishing Zhang’s translator style.The qualitative analysis elaborated Zhang’s translator style with the help of a34-feature matrix.At the document-level,Zhang tends to use relatively more pronouns and less proper nouns,and the large-scale use of adverbs makes his translations more vivid.And Zhang’s endeavor in approaching native Chinese writing in his use of content words and function words is also evident.The relatively high ratio of semicolons and a low ratio of bei sentence marked Zhang’s style at the syntactical aspect,and his constant usage of dialectal expressions indeed mantled his translated discourses with a countryside touch and helped to portray characters in depth.At the N-gram level,Zhang’s usage of unigrams“儿”and“那”,bigrams like some collocations with“那”and other ones like这种情况”,“想要”,and“既是”,which are distinguishing his translations from other translations and the native Chinese works,all illustrate his preference and individual translator style in translation practice. |
PDF Full Text Request