Google, Naver in all-out war for accurate translation service in South Korea
Artificial intelligence increases accuracy in full sentence translations, ‘and the figure is expected to reach higher as more data piles up over time’
By Lee Min-hyung
With artificial intelligence (AI) taking the technology world by storm, global IT firms are identifying intelligent translation services as their next key growth area.
The paradigm shift in translation comes as conventional statistical machine translation (SMT) has so far failed to generate satisfactory outcomes, due to its heavy reliance on computerised statistics which have not been able to translate languages in a more human-like, natural way.
However, AI is now serving as the “brain” for the next-generation neural machine translation (NMT) services, with global portal giants ― such as Google, Baidu and Naver ― launching their own platforms equipped with the AI-based translation technology.
The NMT-based services enable users to translate different languages more naturally, as the AI system analyses and translates the entire sentence, unlike the previous phrase-based translation system.
Another striking feature of the NMT is that it can learn by itself and generate more human-like translation outcomes through big data, whereas the SMT requires manual input to add more data to improve its functions.
Naver vying with Google in Korean translation
Two companies ― Google and Naver ― have so far adopted the NMT in the Korean-English translation, gearing up to gain an upper hand in the infant AI industry here.
In October, Naver, the nation’s No. 1 portal service operator, launched its own NMT service, Papago. Google also unveiled its upgraded translation service the following month, equipped with its AI software.
“It has been a decade since SMT was introduced, but its accuracy has reached only 30 per cent,” said an Naver spokesman. “But NMT, which debuted less than a year ago, has achieved more than 60 per cent accuracy so far, and the figure is expected to reach higher as more data piles up over time.”
Conventional translation industry players are expected to face tough times unless they embrace the new AI-based translation, according to the official.
“We call NMT the brain for the next translation industry, as it can learn languages by itself with enough big data, which has never been the case for SMT,” he said.
The company said its competitive edge comes from its wider range of content, which can be used as a textbook for its NMT-based Papago service.
“We operate such services as Webtoon, or digital comics service, and V app, the personal broadcasting service of celebrities,” said the company official. “These services are popular especially with young people, who help Papago learn more vivid, colloquial Korean expressions in everyday life, which we believe is the best big data textbook to train our Korean translation service, compared with Google Translate.”