Japanese Automatic Voice Translation to Impress at Olympic Games 2020

tokyo-2020-logo.jpg

Japanese Automatic Voice Translation to Impress at Olympic Games 2020

In 2020, Japan will host the Olympic Games in Tokyo. The country, that is known for its technological progressiveness, is hoping to impress the tourists that will visit the event with a very advanced translation device!

According to Tatsuyuki Kobori in an article he wrote at The Asahi Shimbun, Japan plans to launch a new technological device at the 2020 Olympic Games.

This new translation tool will translate Japanese voice recordings into English without a delay.

This will especially come in handy for the media, who, with this new device, will more or less have an interpreter in their back pockets.

The research team for the new device was led by Satoshi Nakamura, who teaches at the Graduate School of Information Science in Nara. The team’s aim for the new tool was to enable simultaneous interpreting for long speeches and news reports, Nakamura says. Even though the commercial launch of the translation device will probably be in 2020, details about the new technology will be made public at the Acoustical Society of Japan conference which started this week.

According to Kobori, the biggest downside to the existing Japanese-English translation software is the time lag that occurs. This delay can be attributed to the fact that the two languages have very different sentence structures. In Japanese, listeners have to pay close attention to the entire sentence the speaker utters, as the verb comes at the very end of a sentence. Thus, the longer the sentence, the longer an automated translation user has to wait to hear the final product.

Nakamura and his team started their research two years ago. They first built a corpus of 500,00 Japanese sentences and their English translations. In addition, they also created a 2.4 million word big corpus consisting of Japanese-English word translations. After analysing the syntax of this data, the researchers created a programme that is able to determine whether translation can already begin without knowing the verb.

And, Kobori says, with success! According to the researchers, the time lag, that usually consisted of about five seconds, has been reduced to one or two seconds. The team also tested the software during a lecture and stated that it was as accurate and fast as an experienced simultaneous interpreter. At the moment, the researchers are working on similar software that can interpret English into Japanese.

According to the team, the software still has room for improvement, however; if voice recognition is fine-tuned, the produced translations will be even more accurate and will be carried out even quicker! The Kwintessential team for one can’t wait for the 2020 games to begin!

Katia Reed
No Comments

Sorry, the comment form is closed at this time.