AI (Artificial Intelligence) is the loser in translation competition… literature section is for ‘human’
Jin Hyun Jin | 2jinhj@ | 2017-02-22 15:38:06
‘Battle’ between four translators and three AI translator machines
Human being has won the battle against AI translator machine, which was organized for the first time in Korea. Although many experts predicted that the machine cannot be as good as human being, this time the limit of machine was clear especially in literature field.
‘Human against AI (Artificial Intelligence) translation battle’ has been held in Sejong University under cooperation of Sejong University and International Translation& Interpretation Association on February 21.
This competition was held due to concerns that consequently AI translation machines’ advanced function will take away human translation jobs
The competition was among four professional translators of International Interpretation & Translation Association and Google Translator, Naver Translator-Papago, Systran.
According to experts, the three machine translators have been known to have 85 ~ 90% translation accuracy compared to human translators, using the Artificial Neural Network Machine Translation (NMT) technology, which learns many translation results by themselves.
The competition was conducted by translating the four prints of literature and non-literary (technology, business, current affairs) into Korean and English and vice versa. Four human translators translated passages one per person and Google • Naver • Sistran Machine Translator simultaneously translated all four passages. The results showed only the translation completion rankings, without opening human translator and machine translators’ name.
Professor Kwak Jung Chul of the University of Interpretation and Translation of Hankuk University of Foreign Studies, who was in charge of the judge at the competition, said that ‘The human translators have won and human translators took 49 points while one AI translator got 28 points and other two got below 20 points out of total 60 points.’ Professor Kwak mentioned that he has selected four passages, which have never translated before.
According to International Interpretation & Translation Association, New York Times columnist-Thomas Friedman’s ‘Thank you for being late’ has been used as literature English-Korean translation passage.
If we compare the translation result, the original sentence "Apple would do the apps. A year later, though, he changed his mind; that fund was launched, and the mobile phone app industry exploded" has been translated to Korean by human translators correctly.
However, AI translator machines processed wrong translation of original text to Korean. Professor Kwak mentioned that ‘AI translator cannot replace human translators for a while’.
The head of International Interpretation & Translation Association-Kang Dae Yeong stated that ‘It is suspicious that machine capability has been improved as human’. He added that ‘Human and machines will be in cooperation to complement some parts to increase work effectiveness’.
By Jin Hyun Jin 2jinhj@
[ copyright ⓒ The Digitaltimes ]
Human being has won the battle against AI translator machine, which was organized for the first time in Korea. Although many experts predicted that the machine cannot be as good as human being, this time the limit of machine was clear especially in literature field.
‘Human against AI (Artificial Intelligence) translation battle’ has been held in Sejong University under cooperation of Sejong University and International Translation& Interpretation Association on February 21.
This competition was held due to concerns that consequently AI translation machines’ advanced function will take away human translation jobs
The competition was among four professional translators of International Interpretation & Translation Association and Google Translator, Naver Translator-Papago, Systran.
According to experts, the three machine translators have been known to have 85 ~ 90% translation accuracy compared to human translators, using the Artificial Neural Network Machine Translation (NMT) technology, which learns many translation results by themselves.
The competition was conducted by translating the four prints of literature and non-literary (technology, business, current affairs) into Korean and English and vice versa. Four human translators translated passages one per person and Google • Naver • Sistran Machine Translator simultaneously translated all four passages. The results showed only the translation completion rankings, without opening human translator and machine translators’ name.
Professor Kwak Jung Chul of the University of Interpretation and Translation of Hankuk University of Foreign Studies, who was in charge of the judge at the competition, said that ‘The human translators have won and human translators took 49 points while one AI translator got 28 points and other two got below 20 points out of total 60 points.’ Professor Kwak mentioned that he has selected four passages, which have never translated before.
According to International Interpretation & Translation Association, New York Times columnist-Thomas Friedman’s ‘Thank you for being late’ has been used as literature English-Korean translation passage.
If we compare the translation result, the original sentence "Apple would do the apps. A year later, though, he changed his mind; that fund was launched, and the mobile phone app industry exploded" has been translated to Korean by human translators correctly.
However, AI translator machines processed wrong translation of original text to Korean. Professor Kwak mentioned that ‘AI translator cannot replace human translators for a while’.
The head of International Interpretation & Translation Association-Kang Dae Yeong stated that ‘It is suspicious that machine capability has been improved as human’. He added that ‘Human and machines will be in cooperation to complement some parts to increase work effectiveness’.
By Jin Hyun Jin 2jinhj@