Responsive image

 首頁 |   Language  : English

因學術研究等數據庫資料龐大,搜尋與運算需花費5至10秒,請您稍侯.

人文社會學院 > 翻譯學系(碩) > 會議論文 >


題名:機器譯文錯誤類型再探-以新聞文本中翻英譯文為例
作者:藍月素 邱東龍
會議名稱:2021年第25屆口筆譯教學國際學術研討會 (The 25th International Symposium on Translation and Interpretation)
發表頁數:
發表地點:東吳大學 (Online 發表)
發表年份:2021
發表月份:06
摘要:機器譯文錯誤類型再探-以新聞文本中翻英譯文為例 中文摘要 因應AI科技的發展,全球翻譯語言服務業者也將機器翻譯融入專業工作流程中,而且工作產值量不容忽視。在此趨勢下,在翻譯教學中訓練學生的機後編輯能力變成不可或缺的一環,甚至已有碩士學程設立「機器譯後編輯訓練模組」(如Universitat Autònoma de Barcelona等大學)培育學生譯後編輯能力,其中一項便是培訓評估機器譯文的品質與常見錯誤類型的能力。 本論文主要採用質性分析法,是2019年在研討會以”Non-binary Errors in Machine Translation of News Media Texts”為題發表之研究的延伸與改進,從2019年研究的語料-11篇紐約時報中文網國際縱覽網頁中最受歡迎文章-當中挑出六篇文章,做為分析與比較的文本,同樣用Google神經機器翻譯系統(GNMT),將這些新聞文本之人工中譯文回譯為英文,並以英文原文為評價指標,然後用適合各種語言配對且錯誤類型定義清楚的「MQM annotation guidelines 2015)」之機器翻譯錯誤類型模式,以分析與比較各篇文章2019年與2021年之英文回譯文的錯誤類型及質量上之差異,最後提出譯後編輯教學策略之建議。 研究發現,2021年的英文機器譯文在動詞時態(verb tense)、名字翻譯(terminology)、句子冗長(wordiness)上仍有很多錯誤,但較2019年的機器英文譯文改善了一些。在選字(word choice)上兩版本幾乎一樣,但是在誤譯(mistranslation)問題上,2021年的版本多了很多這方面的錯誤。國內在中英機器翻譯的相關研究大都集中在技術型文本,且多為英譯中的研究;本論文則研究機器英譯新聞文本譯文,研究結果有助於了解GNMT在中譯英翻譯上的進展,也會對譯後編輯教學有啟示作用。 Error Typology in MT Revisited: Chinese to English Translation Output of Journalistic Texts Abstract Translation services around the world have incorporated the use of machine translation (MT) in their professional workflow as a response to the development of artificial intelligence. The quantity of work accomplished with MT cannot be ignored. As a result, post-editing competency training has become indispensable in the translation instruction. Some postgraduate programs have even set up post-editing modules (e.g. Universitat Autònoma de Barcelona, Ghent University, and Université de Genève) to cultivate the skills for their students’ competency. One of such skills is the ability to evaluate the quality of MT output and identify the common errors found in the raw MT output. This paper is a follow-up of a study titled “Non-binary Errors in Machine Translation of News Media Texts” published at a conference in 2019. The authors use qualitative analysis to track the changes in the quality of the MT output produced by Google Neural Machine Translation (GNMT) in the past two years. The study selects six articles from the eleven most-clicked articles on the New York Times Chinese language websites used in 2019 for analysis and comparison. GNMT is used to back-translate the manually translated Chinese versions into English, and the non-translated texts are used as an evaluation reference. The discrepancies in terms of quality and quantity of the machine translated texts between 2019 and 2021 are analyzed according to the MT error typology in MQM Annotation Guidelines 2015. This paper will also give suggestions on teaching post-editing skills. This research found that though the English back-translations from 2021 still contain man errors in verb tenses, terminology, and wordiness, they are still better compared to the translations from 2019. The 2021 and 2019 versions still have very similar mistakes in word choice, but 2021 versions contain more errors in mistranslation. The Chinese/English machine translation studies done in Taiwan primarily focus on technical texts and mostly on Chinese texts translated from English. This research focuses on machine translated English output of journalistic texts. The error types in accuracy and fluency analyzed in these texts will help see the development of GNMT in the Chinese to English translation. It will also serve in the teaching of post-editing skills.
關鍵字:機後編輯能力、新聞文本、Google神經機器翻譯系統(GNMT)、「MQM annotation guidelines (2014)」、精確度及流暢度
英文關鍵字:post-editing competency, journalistic texts, Google Neural Machine Translation (GNMT), MQM annotation guidelines (2014), accuracy and fluency
全文下載:未授權提供下載
Mail:tschuel@mail.cjcu.edu.tw