Sentence Length and Translation: A Comparative Review of Human, NMT, and LLM Translations

다국어 초록 (Multilingual Abstract)

This paper aims to investigate the handling of long sentences in Korean-to-English translation by human translators, large language models (LLMs), and neural machine translation (NMT). Using reliable human translations in the business reports genre as a reference, the article analyzes human, NMT, and LLM translations through three analytic phases: a quantitative comparison, a qualitative analysis, and retranslation after pre-editing. The analysis found that the sentence length in the original source texts negatively correlates with translation quality in MT outputs, and that NMT's tendency to preserve the original sentence boundary often led to omissions and incomplete translations. However, retranslation after pre-editing effectively fixed both issues. The findings in this article contribute to broadening the literature on machine translation by highlighting the need to focus on the linguistic characteristics of MT, exploring different translation tendencies of the LLM and NMT models, and enhancing the representativeness of test corpora.

참고문헌 (Reference)

1 The jamovi project., "jamovi (Version 2.4) [Computer Software]"

2 Ragni, V., "What has changed with neural machine translation? a critical review of human factors" 30 (30): 137-158, 2022

3 Luo, J., "Universals in machine translation?" 27 (27): 31-58, 2022

4 Graham, Y., "Translationese in machine translation Evaluation"

5 임진, "Translation universals in translated CEO letters in sustainability reports" 20 (20): 131-162, 2019

6 Lommel, A., "Translation Quality Assessment" pringer Cham 109-127, 2018

7 Lyu, Q., "Translating radiology reports into plain language using ChatGPT and GPT-4 with prompt learning: results, limitations, and potential" 6 (6): 2023

8 김경숙, "The influence of case markers on the machine-translation of Korean proverbs into English" 26 (26): 139-157, 2018

9 Zhang, M., "The effect of translationese in machine translation test sets" 1 : 73-81, 2019

10 이승찬 ; 최예슬, "The development of a web-based automatic text complexity measurement program" 30 (30): 163-180, 2019

11 Smith, D. R., "The Lexile scale in theory and practice:final report for NIH grant ID-19448" International Reading Association 1989

12 Baker, M., "Terminology, LSP and translation" John Benjamins 175-186, 1996

13 Blum-Kulka, S., "Strategies in Interlanguage Communication" Longman 119-139, 1983

14 이영희, "Spotting non-nativeness in L2 texts : a statistical approach to translationese" 45 (45): 367-388, 2019

15 이상빈, "Review of literature on machine translation : based on arts and humanities journals covered in KCI Journal Database(from 2011 to early 2020)" 22 (22): 75-104, 2020

16 Chang, M., "Research on Text Difficulty Evaluation for Teaching Reading in Korean" Korea University 2012

17 Yim, J., "NMT vs. LLM - focused on long-sentence translation" 92-104, 2023

18 이창수, "Machine learning classification of literary translation samples by human and machine translators" 22 (22): 199-217, 2021

19 Exchange, "KRX Series: KTOP30"

20 Castilho, S., "Is neural machine translation the new state of the art?" 108 : 109-120, 2017

21 Jiao, W., "Is ChatGPT a good translator? yes with GPT-4 as the engine"

22 이정화 ; 차경환, "Human interpretation and machine translations based upon interviews with Director Joon-ho Bong" 23 : 204-219, 2023

23 Gao, Y., "How to design translation prompts for ChatGPT : an empirical study" arXiv

24 Bizzoni, Y., "How human is machine translationese? comparing human and machine translations of text and speech" 280-290, 2020

25 Bisiada, M, "From Hypotaxis to Parataxis: An Investigation of English-German Syntactic Convergence in Translation" University of Manchester 2013

26 박옥수, "Error analysis and criterion for correcting error of machine translation in terms of source language : based on the syntactic characteristics of non-literary texts" 44 : 151-171, 2018

27 박옥수, "Error analysis according to the typological characteristics of source text in Korean-English machine translation" 41 : 155-183, 2017

28 Kincaid, J. P., "Derivation of New Readability Formulas (Automated Readability Index, Fog Count and Flesch Reading Ease Formula) for Navy Enlisted Personnel (RESEARCH BRANCH REPORT 8-75;Naval Technical Training)" Defense Technical Information Center 1975

29 Graesser, A. C., "Coh-Metrix : analysis of text on cohesion and language" 36 (36): 193-202, 2004

30 Papineni, K., "BLEU: a method for automatic evaluation of machine translation" 311-318, 2002

31 Choi, M, "Assessing Source Text Difficulty for Interpreter Education: With a Focus on Textual Factors of English Source Texts in English-Korean Consecutive Interpreting" Ewha Womans University 2013

32 이정화 ; 차경환, "An analysis of Korean-English translation errors in Google Translate" 89 : 221-257, 2019

33 이소현, "A study on the difficulty distribution of texts for Korean education using the text difficulty measurement program" 68 : 425-450, 2020

34 이창수, "A study of lexical usage differences between human and machine translation in English translations of Korean newspaper editorials" 22 (22): 245-262, 2020

35 이정화 ; 차경환, "A study of error types in Korean-English translation from Korean spoken language to Papago" 23 (23): 56-65, 2022

36 Flesch, R., "A new readability yardstick" 32 (32): 221-233, 1948

37 Dale, E., "A formula for predicting readability : instructions" 27 (27): 37-54, 1948

38 이창수, "A follow-up study of stylistic differences between human and machine translation with ChatGPT added in the mix" 24 (24): 539-561, 2023

39 고광윤 ; 이영희, "A corpus-based study of translation universals in English translations of Korean newspaper texts" 45 : 109-143, 2016

40 이지은 ; 최효은, "A case study on the quality of machine translation and the source text difficulty : a case of a Korean-English statutory translation by Google Translate" 28 (28): 77-101, 2023

41 이지은 ; 최효은, "A case study on the evaluation of Korean-English legal translations by generic and custom neural machine translation engines" 25 (25): 75-98, 2023

42 이지은 ; 최효은, "A case study of Korean-English machine translation of dual subject sentences : a comparison of statutory translations by Google Translate and Naver Papago" 12 (12): 211-241, 2022

43 최효은 ; 이지은, "A Study on the evaluation of Korean-English patent machine translation-focusing on KIPRIS K2E-PAT translation" 19 (19): 139-178, 2017

44 이준호, "A Study on performance evaluation of an specialized machine translation engine in the legal domain : focusing on the Korean-to-English translation of legal contracts" 12 (12): 169-192, 2022

45 Koo, "A Study on Measuring Text Difficulty for Korean Reading Education" Catholic University of Korea. 2013

46 구민지, "A Study of text difficulty analysis on the Korean reading materials" 22 (22): 27-48, 2011