Human vs. AI: An Assessment of the Translation Quality Between Translators and Machine Translation

Human vs. AI: An Assessment of the Translation Quality Between Translators and Machine Translation

Hanji Li (Dalian University of Technology, Dalian, China) and Haiqing Chen (Dalian University of Technology, Dalian, China)
DOI: 10.4018/IJTIAL.2019010104
OnDemand PDF Download:
No Current Special Offers


As one of the most important applications of AI, machine translation has always been the hot topic among scholars in linguistics, computer science, cognitive science and other areas. This article made an assessment of translations of 4 selected major online machine translation platforms from perspectives of efficiency, operating mode and condition. The outputs of machine and human were compared by employing new “6-4” table and comprehensive error rate. The assessment shows that although the quality of machine translation is improving, the gap still exists between the quality of machine translation and human translation. Based on the research findings, the author predicts that machine translation cannot possibly replace human translation and the two will continue to coexist in the foreseeable future.
Article Preview

1. Introduction

Artificial intelligence (AI) seeks to make computers do the sorts of things that minds can do (Boden, 2016, p. 1). Since the Dartmouth Conference in 1956, it had made great progress in so many areas which has already been developed into the third generation that shows its new paradigms in learning methods and technical applications. Due to the close relationship between language and mind, Natural Language Processing (NLP) is considered to be the key for AI. As the earliest and the most common application of NLP, machine translation has always been the hot topic among scholars in linguistics, computer science, cognitive science and other areas with the development of technologies and the growing demand of users. AI, NLP and machine translation are closely related together which relationship is shown in Figure 1.

Figure 1.

The relationship between AI, NLP and machine translation


Machine translation system has also shifted from a rule-based method to a statistical-based method. Currently, with the continuous development in cognitive science, the machine translation based on artificial neural network has become the mainstream and shows its strong performance. Baidu’s neural machine translation system was officially launched in May 2015. On September 27, 2016, Google announced that Google Neural Machine Translation (GNMT) lowered the error rate ranged from 55% to 85% between several key languages than Phrase-Based Machine Translation (PBMT), which results is shown as Figure 2 (Le & Mike, 2016).

Figure 2.

The comparison between the translation quality of GNMT, PBMT and human translation


According to the data, the translation quality of Google Translate using GNMT has been significantly improved compared with the previous PBMT. The translation quality in language combinations such as English-Spanish, English-French and other languages is very close to the level of human translation. Although the quality of translation between English and Chinese is still far from human translation, it has made great progress.

Norbert Wiener (1988, p. 184) uses the example of Prometheus in ancient Greek mythology to show that technology is a “double-edged sword” with both positive and negative effects, which is particularly evident in the era of AI. With the continuous integration of translation technology and AI, users have benefited a lot from the improvement of working efficiency and quality. By contrast, an increasing number of translators are beginning to worry about their future. According to predictions, only 10% of tasks will require manual translation in the future, 70% of documents will require machine translation, and the other 20% tasks need translators to do some post-editing work (Du et al., 2013, pp. 1-8). Ray Kurzweil, the proponent of Singularity, predicts that the quality of machine translation will reach the level of human translation by 2029. Frans De Laet (2017), honorary advisor to the International Federation of Translators, points out that there are two opinions in the translation industry for the use of technology. Some believe that machine translation will take place of human translators; while others believe that it is impossible.

At the moment of the fiery development of AI as well as machine translation, we should ask ourselves some questions. What is the difference between human and machine, or between translators and machine translation. Can machine help human to better their work, or can machine translation improve user’s productivity? What is the future of AI or machine translation? This paper selects 4 widely used online machine translation platforms and evaluates their translations as opposed to human translation in order to answer the above-mentioned questions in the era of AI.

Complete Article List

Search this Journal:
Volume 4: 1 Issue (2022): Forthcoming, Available for Pre-Order
Volume 3: 2 Issues (2021): 1 Released, 1 Forthcoming
Volume 2: 2 Issues (2020)
Volume 1: 2 Issues (2019)
View Complete Journal Contents Listing