Man vs. Machine: Humans Triumph in Translation Competition
The Human vs. Artificial Intelligence Challenge, held at Sejong University in Seoul on 21st Feburary, saw human translators face off against machine translation engines. The event gained a lot of attention, and was covered by dozens of news outlets. This comes less than a year after a Google AI system won 4 out of 5 games of “Go” against a level 9 grandmaster at a competition held in the same city. However, this time the machines weren’t so lucky.
The event, which was organised by the International Interpretation and Translation Association (IITA) and Sejong University, pitted 3 human professional translators against different three machine translation engines. The machine translation side was represented by Google Translate, Never’s Papago and a Systran translation engine. The judging panel was made up of three experts from the IITA, who were also responsible for selecting the source texts. This panel, which included IITA chairperson Gwak Jung-Cheol, scored the translations on a scale ranging from 0 to 30 points.
Both human and machine translators were subjected to the same tasks. Each was made to undertake literary translation and a non-literary translations. The humans were given 50 minutes to complete each task, having access to online resources. On average, the humans scored 24.5 out of 30 points. This represented a clear victory over the machine translation engines, which gained an average score of 10 out of 30.
What lessons can be learned from this competition? Well, it is clear that machine translation software is not capable of replacing human translators just yet. However, Shin Seok-hwan of Saltlux, an artificial neural network technology, emphasised how the results demonstrate the usefulness of collaboration between human and machine translators. Indeed, this is already affecting the translation industry, as demonstrated by the growth of post-editing as a discipline. According to Kim Yu-soek of translation company Systan, advances in neural network machine translation represent an opportunity for further collaboration between human and machine translators. Therefore, it cannot be denied that the future of translation involves both humans and machines.