The Feasibility of Chinese–English Machine Translation Applied to Academic Texts: Using Thesis Abstracts from National Digital Library of Theses and Dissertations (NDLTD) in Taiwan

Abstract

In 2016, Google upgraded its translation system Google Translate (GT) from statistical-based machine translation (SMT) to neural-based machine translation (NMT), the latest machine translation (MT) technology so far, resulting in faster translation speed and higher accuracy. However, according to Angkana Tongpoon-Patanasorn and Karl Griffith (2020), most previous studies focused on statistical-based GT, while few tested the quality of GT’s latest NMT system. This study explores the feasibility of applying GT to Chinese–English academic texts and analyzes whether the English translation produced by GT meets the requirements of English academic writing (AW). The original data used in this study is comprised of Chinese abstracts collected from the website of National Digital Library of Theses and Dissertations in Taiwan. Specifically, twenty most clicked thesis abstracts from each of the five most clicked disciplines, i.e., Engineering, Business and Management, Society and Behavior, Education, and Humanities, are selected, amounting to a total of 100 abstracts. Through a qualitative coding method and content analysis, the results show that three AW features can be found in the English translation output, including this-format, mid-positioning of adverbs, and passive voice. In terms of verb choice, nearly half of the abstracts show use of phrasal verbs, indicating GT’s inability to constantly adopt single-word verbs when translating academic texts from Chinese into English. Also, GT performs best on abstracts from Engineering discipline. By exploring the feasibility of applying GT to Chinese–English academic texts, this study may help contribute to a better understanding of and facilitating use of MT in academic circles.