
An image of Naver's Papago translation service / Courtesy of Naver
By Baek Byung-yeul
Naver said Wednesday that it has upgraded its image translation service for three languages ― English, Chinese and Korean ― enabling users to receive a more accurate interpretation of text captured in images.
The company uses artificial intelligence (AI) technology for the translation app Papago, and said it had enhanced the image translation service by utilizing deep learning technology.
“Thanks to the upgraded deep learning technology-based translation engine, the service now better recognizes text in images and its accuracy has been improved by 57 percent,” a company official said.
Naver, which applied the upgraded translation engine for Japanese first, said many users were satisfied with the service.
“The upgraded translation feature was applied to Japanese first and we saw the use of image translations triple year-on-year,” the official said.
The image translation service has been an essential feature for people travelling outside the country as they can easily translate menus at local restaurants, as well as signs. Naver said offering an improved image translation service was an important mission for the company because more than a quarter of all Papago users frequently use it.
The company will continue to upgrade the image translation service to reduce inconvenience when users travel overseas.
“We will keep trying to advance the service enabling it to even translate images of hand-written messages,” Shin Joong-hui, who supervises the Papago service, said. “We hope overseas tourists traveling to Korea or Koreans visiting other countries will be able to overcome the language barrier by using Papago.”
Naver launched the AI-based translation service in July 2017, and it has become the leading mobile translation service here with more than 10 million users. Powered by neural machine translation technology, it provides translations for 13 languages.
Citing its own data, Naver said the Papago's translation quality of the four most-used languages ― Korean, English, Chinese and Japanese ― was 27 percent more accurate than other translation services on average.