Post-editing neural machine translation output of wealth management text : challenges and the way forward

• Main Author: Chan, Elaine Choon Ling
• Other Authors: -
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Humanities::Language
• Online Access: https://hdl.handle.net/10356/148585
• Institution: Nanyang Technological University

The prevalence of artificial intelligence (AI) and its power to transform almost anything bring to the forefront the hard truth that no sensible decision can be made without taking AI into account. The use of AI in neural machine translation (NMT) revolutionised translation process with unprecedented standards of efficiency and productivity. While post-editing (PE) on NMT output is necessary for quality control––which applies to human translation output just the same–– research and developments are just scratching the surface in the PE field. More than rectifying superficial MT output errors, the potential value-add which PE offers to the future of language service providers is far-reaching as they pivot to PE-MT business model. Centred on a wealth management text, this paper aims to examine PE as the fundamental auxiliary to MT in navigating societal nuances and cultural undertones. The study also focuses on the Chinese to English (C-E) language pairs which has seen lesser of the limelight than other prominent language pairs in PE exploration. Excerpts from wealth management textbook Zhongguo Caifu Guanli Guwen Yingxiao Shizhan 中国财富管理顾问营销实战 was fed into an online MT tool, Youdao Translate to obtain the raw TT which was then comparatively analyzed against the TT done purely by human translator sans machine-assisted translation tools. The manual human translation (HT) considered theoretical strategies mindful of cultural and pragmatic divergence. A quantitative analysis of consolidated MT output errors and non-errors revealed shortcomings in processing semantics and pragmatics categories. PE would then bridge this gap and boost reader experience with a human touch. The paper concludes that the future of the translation industry, especially in Asia, requires more supported development of PE; with post-editors playing a more significant role to improve MT systems and provide a smoother translation process.