語料庫驅動的譯者風格分析： 以張愛玲與金凱筠的文學翻譯為例

Corpus-driven Analysis of Translator Style: A Case Study on Eileen Chang and Karen S. Kingsbury

吳江美(Chiang-Mei Wu)

指導教授 ： 高照明

國立臺灣大學/文學院/翻譯碩士學位學程/碩士(2023年)

https://doi.org/10.6342/NTU202301412

摘要

在中文文學創作中，極少數作品享有翻譯為多重譯本的機會。中文作品英譯本單一的普遍現象，也令中譯英語言方向的譯者風格研究窒礙難行，研究材料與方法限制重重。本研究提出三層次的方法學架構，並選定分析張愛玲與金凱筠兩位譯者之作品，分別為《金鎖記》以及《傾城之戀》的英譯本，以及其中文原文。本研究運用最新人工智慧套件如vecalign句對齊、AWESOME align詞對齊工具，以及自動化程式編寫找出共同詞彙，並輔以PoS-gram分析句法結構以及可供下載之篇章分析工具，加以改善中譯英譯者風格研究之缺口，完成更具全面性，並且著重文本細部用詞與句構的語料庫翻譯研究。本研究以詞彙、句法及篇章三層架構分析語料，辨識翻譯風格差異，並歸結譯者背景對翻譯策略之影響。本研究發現張愛玲翻譯中文虛數詞以及特定副詞傾向使用固定的英文詞彙。反之，美籍譯者金凱筠則展現更多元的選詞。譯者風格差異也可能受到本身所處時代背景、意識形態、自譯與否等等因素的影響。本研究在辨識翻譯詞彙差異之後，分析譯者背景，討論前述因素對風格所造成的可能影響。透過實際譯例，本研究提出張愛玲自譯作品中可能展現的女性主義，剖析張對西力東漸的可能看法，論證張身為自譯者所展現的詞彙敏感度。綜而論之，本研究設計的方法學，借助預訓練模型的對齊工具，成功跨越不同來源文本譯作的研究困難。此為初步探析，建議日後研究依循此方法學，擴展語料庫，以期更完備地分析張與金的譯者風格差異。

關鍵字

語料庫風格研究 ； 張愛玲 ； 金凱筠 ； 自譯 ； 雙語語料對齊 ； 預訓練模型

並列摘要

In Sinophone literature, a limited number of literary works were translated multiple times, making it challenging to study translator’s styles due to the lack of research materials and methods. This study experimented with a multi-level methodology with Eileen Chang’s self-translation of The Golden Cangue and Karen S. Kingsbury’s translation of Love in a Fallen City. The research materials were translated respectively by two translators with diverse language and historical backgrounds, one of whom is a self-translator and the other is not. This study proposed novel methods empowered by state-of-the-art alignment tools, including vecalign for sentence alignment, AWESOME align for word alignment, and using basic Python coding to find matching lexical items. This study divided translator style into three levels: lexical, syntactic, and discourse. Tools such as alignment, PoS-gram, and computer-based software were used to investigate the three levels. With these tools, we provided translational evidence that helped identify stylistic differences between Chang and Kingsbury. Firstly, we found that Chang showed the tendency to keep to the same set of phrases when it comes to indefinite classifiers and certain adverbs, while Kingsbury exhibited a wider selection of words. Furthermore, we identified different word preferences that possibly fell under the influence of translator backgrounds. We found evidence of Chang’s authorial manipulation in several word choices, which may have revealed her take on feminism and the Western impact on Chinese culture. In conclusion, this study devised a methodology that leveraged the latest AI pre-trained models to explore translator style even when source texts and translators are different. This study suggests expanding corpora size to identify more patterns, exert a broader impact, and find more substantial proof for the translator style of Chang and Kingsbury.

並列關鍵字

corpus stylistics ； Eileen Chang ； Karen S. Kingsbury ； self-translation ； bilingual text alignment ； pre-trained alignment models

參考文獻

Abdul-Rauf, S., Fishel, M., Lambert, P., Noubours, S., & Sennrich, R. (2012). Extrinsic evaluation of sentence alignment systems.

Andriamanankasina, T., Araki, K., & Tochinai, K. (1999). Sub-sentential Alignment Method by Analogy. Paper presented at the Proceedings of the 13th Pacific Asia Conference on Language, Information and Computation.

Anthony, L. (2017). AntPConc (1.2.1) [Computer software]. Tokyo, Japan: Waseda University. Available from https://www.laurenceanthony.net/software/antpconc/

Anthony, L. (2022). AntConc (4.0.6) [Computer software]. Tokyo, Japan: Waseda University. Available from https://www.laurenceanthony.net/software/antconc/

Baker, M. (1993). Corpus Linguistics and Translation Studies Implications and Applications. Text and Technology: In honour of John Sinclair, 233.