In a recent battle of man vs. computer, human translators proved they still outperform Machine Translation (MT) systems notwithstanding great progress in artificial intelligence technology.
In a competition held in South Korea on Tuesday, four professional translators competed against and outdid three MT programs, producing drastically more accurate translations, according to South Korea’s Yonhap News Agency.
The competition was hosted by the International Interpretation Translation Association and had both the MT systems and human linguists translate random English articles into Korean and vice versa. Both sides had 50 minutes to complete the work, which was evaluated by two professional translators.
According to Yonhap, the human translators scored an average of 25 out of 30 on their Korean-to-English translations compared to MT’s score of between 10 and 15. The MT systems used were provided by Google, Naver Inc. and Systran International.
The MT technology translators squared off against were Neural Machine Translation (NMT) applications, which have gained much attention in the language world lately. NMT system creators and others believe that the new technology will outpace the competence of human translators.
Difficulty Understanding Context
According to Yonhap, the NMT systems had difficulty comprehending the context of what they translated. This shortcoming has long been an issue for MT applications, which are not as capable as humans in deciphering nuanced content.
Organizers of the competition said the structure of NMT’s translations were “grammatically awkward.” Given current technologies, it’s common to have even the most robust MT systems produce translations that seem mechanical or clumsy.
Even though the results weren’t exactly favorable for MT, the outcome of the competition doesn’t mean the technology should be written off. Progress in the field of NMT has been tremendous recently, and MT remains a great way to produce “gist” translations.
The Future of Machine Translation
Most in the language industry won’t see the competition’s results as surprising. Practitioners realize MT has a long way to go before it reaches its full potential. And, according to some, it may happen faster than we think.
In the Yonhap article, Systran Director Kim Yoo-seok is quoted as saying, "I would say the (MT) technology is at the stage of elementary school at the moment but it will improve to the level of high school and college in just one or two years.”
That’s an interesting analogy to use, and Yoo-seok’s comment gives a glimpse into how quickly language technologies are advancing.
For now, the translation industry will have to wait to see MT’s potential fully realized and the impacts the technology will have on the profession.
Until then, human translators seem to have the advantage.
