Re-Looking Into Machine Translation Errors and Post-Editing Strategies in a Changing High-Tech Context

This article re-looks into machine translation (MT) errors and proposes a function-oriented MT post-editing (MTPE) typology in a new technological context. Driven by the technological advances of the neural machine translation (NMT) system over the past several years, the author thinks that we should re-examine MT errors created by NMT systems, and understand whether the NMT system can resolve the issues the rule-based MT (RBMT) and statistical MT (SMT) systems have encountered. A mixed-methods approach is used to complete this study, and technical texts, journalistic texts and web-based company texts are chosen as analytical materials. The three-phased procedure consists of (1) cross-checking the differences between source texts (STs), MT outputs and corresponding human translations (HTs) to identify MT errors, (2) proposing a three-tier MTPE typology to supplement the current binary MTPE typology and (3) exploring empirical and theoretical implications of this research. The findings differ from previous MTPE studies in three aspects: (1) amending linguistic, pragmatic and affective MT errors with the strategies of ＂accurate-enough editing,＂＂clear-enough editing＂ and ＂attractive-enough editing,＂ not the strategies of light editing and full editing; (2) replacing the existing editor-driven MTPE typology with a function-driven MTPE typology; and (3) using a progressive, flexible MTPE typology to meet the textual functions of different types of MT texts. Overall, this article re-examines MT errors and MTPE strategies, and raises an alternative MTPE typology from the perspective of textual functions in the framework of the NMT scenario. It expects to add some novel insights to contemporary MT studies.

關鍵字

MT errors ； MT post-editing (MTPE) strategies ； a cross-checking method ； a three-tier MTPE typology ； textual functions

並列摘要

本論文重新審視新型神經機器翻譯系統產生的翻譯錯誤，並提出功能導向的機譯後編輯類型學。由於近幾年來神經機器翻譯系統的技術大幅進步，作者認為有必要重新檢視神經機譯系統產生的翻譯錯誤類型，以了解是否有所突破與改進。本論文採用混成研究方法，並選擇技術文本、新聞文本及公司網頁為分析樣本。研究過程包含：（1）交叉分析比對原文、機器翻譯及人工譯文之差異，以辨識及歸納神經機譯錯誤類型；（2）提出「三層級的機譯後編輯類型學」，以補充目前二元對立局部與全部機譯後編輯類型學；（3）探討本研究在實證及理論層面所透露的意涵。最終分析結果證實本研究與過去機譯後編研究有三處差異：（1）本論文採用足夠正確、足夠清晰、足夠吸引的三層級機譯後編輯策略，來修正語言、語用及情感的機器翻譯錯誤；（2）本論文提出功能導向的機譯後編輯類型學，不是依照編輯的程度而定；（3）建議採用彈性、漸進式後編輯，俾各種文類的機器翻譯可達成不同的文本功能。簡言之，本研究在神經機譯科技的框架下，從文本功能視角來探究機譯的錯誤類型、後編輯策略，並提出三層級功能導向的機譯後編輯類型學，期能為當代翻譯研究提供一些嶄新的洞見。

並列關鍵字

機器翻譯錯誤；機器翻譯後編輯策略；交叉分析方法；三層級後編輯類型學；文本功能

參考文獻

Allen, J. H. (2003). Post-editing. In H. Somers (Ed.), Computers and translation: A translator’s guide (pp. 297-317). John Benjamins. https://doi.org/10.1075/btl.35

Castilho, S., Moorkens, J., Gaspari, F., Calixto, I., Tinsley, J., & Way, A. (2017). Is neural machine translation the new state of the art? The Prague Bulletin of Mathematical Linguistics, 108, 109-120. https://doi.org/10.1515/pralin-2017-0013

Flanagan, M., & Christensen, T. P. (2014). Testing post-editing guidelines: How translation trainees interpret them and how to tailor them for translator training purposes. The Interpreter and Translator Trainer, 8(2), 257-275. https://doi.org/10.1080/1750399X.2014.936111

Hu, K., & Cadwell, P. (2016). A comparative study of post-editing guidelines. Baltic J. Modern Computing, 4(2), 346-353. https://doi.org/10.13140/RG.2.1.2253.1446

Kliffer, M. D. (2008). Post-editing machine translation as an FSL exercise. Porta Linguarum, 9, 53-67. https://doi.org/10.30827/Digibug.31745

國際替代計量

Re-Looking Into Machine Translation Errors and Post-Editing Strategies in a Changing High-Tech Context

全文下載

主題瀏覽